[{"content":"How AI computing became an infrastructure business, who supplies the value chain, and what determines delivery and investment returns. Sources current to October 1, 2026. This report covers the chips, servers, data centers and cloud services used for AI training and inference. Financial figures retain each company\u0026rsquo;s reporting period and business scope. Supporting materials appear in a separate appendix outside Parts 1 to 4.\nAI services depend on a chain of suppliers and operators that turn equipment into usable computing capacity. This report follows an order through that chain, traces how the industry developed, identifies representative companies, and examines the constraints on delivery and investment returns.\nPart 1: Story In September 2026, Dell reported results for the fiscal quarter ended July 31. AI-optimized server revenue was $16.4 billion, new orders totaled $60.9 billion, and the quarter ended with a $95 billion backlog.[23] Revenue represents business recognized during the quarter, orders reflect newly booked demand, and backlog is the stock still awaiting delivery. Converting that backlog into revenue requires production and delivery to be completed.\nMuch has to happen between an order and a working installation. Server vendors need accelerators, memory and networking components to arrive, then must test the assembled systems. Customers need sites with suitable power and cooling. A delay at one stage can push back the deployment schedule. For the customer, receiving the equipment is only one step in the project.\nOnce a cluster is running, the operating questions change. Users care less about how many chips it contains than whether its answers are useful and arrive quickly enough. Service providers must handle more paid requests while meeting those expectations, because depreciation, electricity bills and staffing costs continue. The capabilities promised at purchase have to be delivered in everyday operation.\nAI infrastructure connects businesses that can otherwise look quite different. Chip design, manufacturing, system delivery and cloud services all respond to the same underlying demand. Understanding the industry means following an order through to the point where computing capacity is used and customers continue paying for it. The formation of this value chain begins with much earlier changes in research and engineering.\nPart 2: Industry History 1. From research questions to commercial experiments In 1943, McCulloch and Pitts proposed a mathematical model of neural activity. In 1950, Alan Turing examined machine intelligence and introduced the \u0026lsquo;imitation game\u0026rsquo;. The term \u0026lsquo;artificial intelligence\u0026rsquo; appeared in the 1955 proposal for the Dartmouth research project, which took place in the summer of 1956. The first question was whether ideas about intelligence could be turned into programs a computer could execute.[1]\nEarly programs made progress on well-defined problems but often failed when faced with the complexity of the real world. In the 1970s, disappointed research expectations were followed by cuts in funding. Expert systems, which rose to prominence in the 1980s, encoded domain knowledge as rules but were subsequently constrained by maintenance costs and limited adaptability. Specialized Lisp computers also faced competition from cheaper general-purpose workstations. Both downturns exposed the distance between a technical demonstration and a commercially sustainable application.[2]\nIn 1997, IBM\u0026rsquo;s Deep Blue defeated Garry Kasparov in a six-game match. It combined specialized search chips, parallel computing, evaluation functions and databases of chess games to solve a clearly defined problem. The principle of designing a system around a task can still be seen in later AI infrastructure.[3]\n2. GPUs and cloud computing improve access GPUs can perform many similar operations at once, making them well suited to much of the computation in neural networks. NVIDIA introduced the CUDA architecture in 2006 and made related development tools available in 2007. Researchers gained not only faster chips but also programming tools they could put to work. Around the same time, AWS launched its S3 storage and EC2 computing services in 2006, allowing developers to rent infrastructure.[4][5]\nAlexNet provided a clear example of the change in 2012. The individual network described in the paper took about five to six days to train on two GTX 580 GPUs. The team\u0026rsquo;s ensemble of models achieved a top-5 test error rate of 15.3% in the ImageNet competition, compared with 26.2% for the runner-up. This error rate measures the share of cases in which the correct category did not appear among the five highest-ranked answers. GPUs, labeled data, network design and training methods all contributed to the breakthrough.[6]\n3. Larger training runs change system design In 2018, OpenAI analyzed large training runs carried out since 2012. It found that the amount of compute used in the largest runs in its sample had doubled roughly every 3.4 months, increasing by more than 300,000 times overall. This was a historical change in resources devoted to training. It was not a growth rate for model capabilities, and it cannot simply be extrapolated to the present.[7]\nHardware and algorithms were both changing. NVIDIA introduced Tensor Cores with its Volta architecture in 2017 to accelerate matrix operations. Google\u0026rsquo;s first-generation TPU, deployed from 2015, was designed primarily for inference: applying an already-trained model to new inputs. The 2017 Transformer paper demonstrated greater training parallelism in its machine-translation experiments and provided an architectural foundation for later language models.[8][9][10]\nGPT-3, introduced in 2020, had 175 billion parameters. ChatGPT\u0026rsquo;s release in 2022 brought conversational models to a much wider audience. Demand acquired an additional dimension: systems had to handle a continuous stream of user requests as well as complete training runs.[11]\n4. Computing becomes a data center engineering problem Meta\u0026rsquo;s 2024 Llama 3 report described using up to approximately 16,000 H100 GPUs to train its 405-billion-parameter model and discussed networking, storage and fault recovery in detail. As the number of machines grows, communication delays and downtime become more costly.[12]\nCooling must keep pace with increasing system density. The GB200 NVL72 uses a liquid-cooled rack, while the H100 product range still includes air-cooled configurations. The choice depends on power consumption and deployment conditions. Software also changes how efficiently equipment is used: the PagedAttention study demonstrated that better GPU memory management could improve inference throughput under the conditions tested.[13][14]\nPower supply further constrains where a project can be built and when it can enter service. In its 2026 report, the International Energy Agency estimated global data center electricity consumption at approximately 485 terawatt-hours in 2025 and forecast approximately 950 terawatt-hours in 2030. These figures include non-AI uses. Operators must then answer a set of practical questions: when will the equipment go live, how many customers can it serve, and how long will it take to recover the investment?[15]\nPart 3: Value Chain INDUSTRY MAP\nThe AI computing infrastructure value chain Follow the capabilities needed to turn chips into usable computing services.\nFit to view Expand all Collapse Full screen Read the full text outline AI computing infrastructureUpstream · Design and componentsAccelerators and softwareGPUs, development tools and model execution Representative companiesNVIDIA AMD Huawei Cloud provider chipsSpecialized accelerators for cloud workloads Representative companiesAWS · Trainium Google · TPU Fabrication and advanced packagingWafer production and system integration Representative companiesTSMC ASE HBM memoryData capacity and transfer bandwidth Representative companiesSK hynix Micron Samsung Midstream · Systems and facilitiesNetworking and interconnectsCommunication between servers and computing nodes Representative companiesNVIDIA Broadcom Servers and integrationAssembly, testing and delivery of complete systems Representative companiesDell Supermicro Power and coolingPower distribution, thermal management and site readiness Representative companiesVertiv Schneider Electric Downstream · Computing servicesCloud computing servicesCapacity, service availability and cluster operations Representative companiesAWS Microsoft Azure Google Cloud CoreWeave Customer usesModel training Inference and business applications DEX editorial map based on Part 3 and the supporting appendix. Company examples are illustrative and may span several stages. Branches group business roles; they do not establish supplier contracts, market shares or an exhaustive list.\nSources: NVIDIA · computing and networking business · TSMC · CoWoS advanced packaging · Dell · AI server revenue and delivery pipeline · IEA · energy and infrastructure constraints. Reviewed 2026-10-01.\n1. Working back from the customer\u0026rsquo;s bill The value chain ends with the user of computing capacity. A business might rent GPUs by the hour or pay for model services according to usage. At sufficient scale, it may build its own cluster. Cloud providers collect service fees and spend a portion on servers, networking, electricity and operations. Server suppliers buy accelerators, memory and other components, while chip designers purchase manufacturing services from foundries and packaging providers. This is a simplified account of a typical division of work. Actual contracts may also involve large customers buying components directly and commissioning their integration.\nThe same end-customer demand generates revenue for several companies along the chain. Chip sales, server sales and cloud revenue therefore cannot simply be added together to calculate market size. When examining a particular stage, identifying the customer, the product or service delivered and the point at which revenue is recognized is more useful than starting with a single figure for the entire industry.\nThe table identifies representative companies by what they deliver. A company may participate at several stages; the list is not a market-share ranking.\nValue chain stage Representative companies Business role and sources Accelerators and software NVIDIA , AMD , Huawei NVIDIA supplies GPUs and CUDA; AMD offers Instinct products; Huawei provides the Ascend platform and CANN.[4][18][19][27] Cloud provider chips AWS , Google Develop Trainium and TPU respectively, configuring compute for internal and customer workloads.[21][22] Fabrication and advanced packaging TSMC , ASE TSMC provides wafer fabrication and CoWoS; ASE supplies advanced packaging including 2.5D/3D integration.[16][29] HBM memory SK hynix , Micron , Samsung Supply high-bandwidth memory, providing accelerators with data capacity and transfer bandwidth.[17][30][31] Networking and interconnects NVIDIA , Broadcom Supply data center networking products and chips that connect servers and computing nodes.[18][20] Servers and integration Dell , Supermicro Integrate accelerators and other components into AI servers, with air-cooled and liquid-cooled system formats.[23][32] Power and cooling Vertiv , Schneider Electric Supply power, thermal management and supporting data center infrastructure.[24][33] Cloud computing services AWS , Microsoft Azure , Google Cloud , CoreWeave Deliver computing capacity and related services to customers and operate the underlying clusters.[5][25][26][34] 2. Chip design and software Accelerator design determines how computing units, on-chip memory and interconnects work together. Software determines whether models can use those resources. CUDA offers one way to understand this relationship: development tools, libraries and accumulated code affect the engineering work required to move to different hardware.[4] On that basis, this report argues that a chip should be assessed not only on performance and price, but also on the time needed to migrate models, debug them and achieve stable operation.\nGeneral-purpose GPUs can accommodate changing models and customer requirements. Large cloud providers are also in a position to develop specialized chips for more clearly defined workloads. AWS announced general availability of Trainium3 UltraServers in December 2025. Google introduced the TPU 8t and 8i for different workloads in April 2026. The former was an availability announcement; the latter was a product announcement. Neither establishes how many systems have actually been deployed.[21][22]\nThis creates two different commercial choices: selling chips and using chips to provide services. Whether an in-house chip pays off depends on how widely development costs can be spread, as well as performance, utilization and operating costs after migration. Cloud providers have an incentive to reduce the cost of serving each unit of demand. External customers are more concerned with whether their workloads run successfully.\n3. Manufacturing advanced packaging and memory Once a chip is designed, it must pass through wafer fabrication, packaging and testing. Advanced packaging is taking on more work. TSMC\u0026rsquo;s CoWoS, for example, connects logic chips and high-bandwidth memory within the same packaging system. In its second-quarter 2026 earnings call, TSMC said packaging capacity remained tight. More wafer supply does not automatically mean complete accelerators can be delivered on time.[16]\nHBM is high-bandwidth memory placed close to the computing chip. Capacity determines how much data it can hold, while bandwidth affects how quickly data moves. One cannot substitute for the other. Micron\u0026rsquo;s published specifications for its 12-high HBM4 product, for example, list 36 GB of capacity and bandwidth exceeding 2.8 TB/s. These are component specifications; they cannot be converted directly into the amount of text a server generates per second.[17]\nConstraints at this stage travel through the supply chain. Delays in computing chips, HBM or packaging can each hold up system shipments. Capacity expansion, meanwhile, requires upfront spending. The corresponding additional revenue generally cannot be recognized until the new capacity is ready and products have been delivered. Assessing business conditions therefore requires looking at demand, construction progress and yields together, rather than relying on expansion announcements alone.\n4. Networks and server integration Large training clusters must exchange data between computing nodes while continuously reading training material from storage.[12] Assessing a system therefore requires measuring how much time is spent on communication and data access. A single chip\u0026rsquo;s peak computing performance does not capture these issues.\nServer vendors and system integrators combine accelerators, CPUs, memory, networking and cooling into deliverable products. They also handle testing, deployment and after-sales service. Revenue can grow rapidly, but the funding needed for procurement, inventory commitments and delivery obligations rise with it.\n5. Power cooling and cloud operations Data centers need power distribution, backup power, cooling equipment and routine maintenance. Buying servers does not by itself provide usable computing capacity: sites, grid connections and installation must also be ready. Vertiv supplies infrastructure including power and thermal-management systems. It reported approximately $3.274 billion in sales in the second quarter of 2026. That figure indicates the scale of one supplier\u0026rsquo;s business; it cannot all be classified as revenue from liquid cooling for AI.[24]\nOnce systems are operating, utilization and pricing become central. Training customers may occupy large amounts of equipment for a defined period, while inference services must respond to changing request volumes and latency requirements. Under the same requirements for answer quality and waiting time, the number of requests each machine can handle is a measure of efficiency closer to what operators need. Software optimization and standardized testing provide ways to make such comparisons.[14][28]\nThe operating economics can be expressed as a straightforward calculation: service revenue must cover equipment depreciation, financing, electricity, premises, networking and operating costs. Even as new chips become faster, older equipment may still serve less demanding tasks. But falling rental prices or low utilization will lengthen the payback period. End customers\u0026rsquo; willingness to pay determines whether expansion further up the chain can produce recurring revenue.\nPart 4: Industry Challenges 1. Supply constraints and delivery schedules Advanced manufacturing, packaging, memory and grid connections constrain different stages of delivery. TSMC\u0026rsquo;s disclosure of tight packaging capacity shows that constraints can arise after wafer fabrication. The power bottlenecks discussed by the IEA show that construction obstacles remain even after equipment leaves the factory.[16][15] This report\u0026rsquo;s assessment is that a shortage at a particular stage may give its suppliers greater pricing power, but that advantage must be reassessed as supply expands and customer demand changes.\nCapacity takes time to build, while chip generations and customer workloads can change during construction. A shortage today does not guarantee strong utilization once an expansion is complete. Buyers need to assess delivery dates and readiness across the whole project; suppliers need to judge whether demand will persist when their new capacity comes online.\n2. Revenue growth and investment returns The table below selects public figures from different stages of the value chain to show where demand is turning into revenue. Companies differ in fiscal periods, product scope and revenue recognition. These figures therefore do not rank shares of a single market. All amounts are in billions of US dollars and represent disclosed quarterly revenue; orders and forecasts are excluded.\nCompany / business Period ended Quarterly revenue ($bn) Scope and source NVIDIA Data Center 2026-07-26 89.0 Includes computing and networking; not GPU revenue alone.[18] AMD Data Center 2026-06-27 6.7 Includes EPYC CPUs and Instinct GPUs.[19] Broadcom AI semiconductors 2026-08-02 16.7 Includes custom accelerators and AI networking.[20] TSMC , company-wide 2026-06-30 40.20 Includes non-AI demand, such as smartphones.[16] Dell AI-optimized servers 2026-07-31 16.4 Recognized server revenue.[23] Vertiv , company-wide 2026-06-30 Approx. 3.274 Includes infrastructure and services for non-AI uses.[24] Google Cloud 2026-06-30 24.768 Includes revenue from GCP, Workspace and other products.[25] CoreWeave , company-wide 2026-06-30 2.575 Company revenue, including cloud services.[26] NVIDIA offers a view of revenue concentration within one company. In the quarter ended July 26, 2026, Data Center revenue was approximately $89.0 billion against total company revenue of $96.221 billion. Data Center therefore represented approximately 92.5% of NVIDIA\u0026rsquo;s revenue, with other businesses accounting for the remaining 7.5%. These calculated percentages describe NVIDIA\u0026rsquo;s revenue mix; Data Center includes computing and networking, and the figures do not measure AI revenue alone or NVIDIA\u0026rsquo;s share of the industry.[18]\nREVENUE MIX · FY2027 Q2\nNVIDIA revenue mix Quarter ended July 26, 2026. Data Center as a share of NVIDIA\u0026#39;s total quarterly revenue. This measures one company\u0026#39;s revenue composition, not its industry market share or AI-only sales.\nWorldwide company revenue · Share of NVIDIA revenue (%)\nFull screen View the data table NVIDIA revenue mix · FY2027 Q2 · Share of NVIDIA revenue (%) Business groupingShare Data Center92.5%All other businesses7.5% Download data (CSV) Calculated from approximately $89.0 billion in Data Center revenue and $96.221 billion in total revenue. Data Center share = 89.0 / 96.221 × 100, rounded to one decimal place; the other category is the residual to 100%. Data Center includes computing and networking and is not synonymous with AI. The source rounds Data Center revenue, so the percentages are approximate.\nSource: NVIDIA — Financial results for second quarter fiscal 2027 (2026-08-26). Reviewed 2026-10-01.\nDifferences in scale do not establish which business model is more profitable. Chip design, wafer fabrication, complete-system delivery and cloud operations carry different costs. Assessing the quality of a business also requires examining gross profit, capital expenditure, cash flow and asset utilization. In particular, business revenue that includes CPUs or enterprise productivity services cannot all be counted as AI accelerator or model-service revenue.\nContract value is no substitute for profitability. In the second quarter of 2026, CoreWeave reported revenue of $2.575 billion alongside a GAAP net loss of $0.626 billion. Its approximately $104 billion quarter-end revenue backlog included remaining performance obligations and estimated future revenue under other contracts. Realization remained subject to conditions including delivery and service availability.[26] These figures place growth and cost pressures on the same set of accounts.\n3. Platform choice and software migration The first approach is to build a hardware and software platform around accelerators. NVIDIA\u0026rsquo;s business now spans computing and networking, while AMD supplies both data center CPUs and accelerators. The second is for cloud providers to develop chips around their own workloads, then supply computing capacity to internal operations or external customers. Products from AWS and Google illustrate this approach. The third is to provide custom accelerators and networking chips for large customers. Broadcom\u0026rsquo;s disclosed AI semiconductor revenue falls into this category.[18][19][20][21][22]\nAll three approaches can appear on a single customer\u0026rsquo;s purchasing list. An established platform is attractive when models are still changing and there is little time for migration. Specialized designs become more attractive when workloads are stable enough and usage is sufficiently large. This report therefore expects competition to center on workloads, software compatibility and total cost of use. A single chip metric is unlikely to determine the outcome.\nCompetition in China’s market also involves software. Huawei\u0026rsquo;s documentation positions CANN, its heterogeneous computing architecture, between AI frameworks and Ascend hardware, covering functions such as compilation, operators and runtime execution.[27] Assessing a platform therefore requires looking beyond chip specifications to whether existing models can be migrated, which operators need rewriting and how easily the tools support debugging.\n4. Hardware specifications and actual service costs A useful system comparison starts with the same task. The model, precision, quality target and latency requirements need to be specified before throughput is compared. MLPerf’s inference benchmarks use defined scenarios and quality requirements for this reason.[28] A chip specification or a result from one workload cannot establish the operating performance of every deployment.\nFor operators, the next step is to connect those measurements to the bill. The same number of installed machines can produce different costs per completed request if utilization, downtime or software efficiency differs. Evaluation therefore needs to include deployment work, power, maintenance and the amount of capacity customers actually use.\n5. Whether paid usage can sustain expansion On the demand side, watch whether paid usage keeps pace with capacity additions. On the supply side, track whether advanced packaging, memory and power projects are delivered on schedule. For operating performance, examine utilization, cash flow and investment payback periods. These measures are connected: delivery delays can hold back real customer demand, while fully installed equipment can still face falling prices and low utilization.\nAI infrastructure now links model development with manufacturing, engineering and ongoing operations. Companies that deliver reliably, enable customers to use their products or services successfully, and cover costs while generating recurring cash flow at their own stage of the chain have an opportunity to turn a wave of purchasing into a lasting business. That is this report\u0026rsquo;s overall assessment of the value chain; actual delivery and financial results will be needed to test it.\nAppendix: Supporting Materials Source numbers correspond to the citations in the article. The appendix is separate from Parts 1 to 4.\n[1] Early AI documents McCulloch \u0026amp; Pitts (1943) — A Logical Calculus of the Ideas Immanent in Nervous Activity\nTuring (1950) — Computing Machinery and Intelligence\nMcCarthy et al. (1955) — Dartmouth research proposal\nDartmouth — Artificial Intelligence Coined at Dartmouth\nLocation: Turing, pp. 433–434; opening of the Dartmouth proposal; Dartmouth history page. Original papers and institutional records.\n[2] AI winters and expert systems Ted E. Senator (2026) Implications for AI Research: Applying Lessons from the Expert Systems Boom and Bust to the Current Large-Language Model Boom Location: PDF pp. 1–2. A retrospective scholarly paper.\n[3] Deep Blue Campbell, Hoane \u0026amp; Hsu (2002) — Deep Blue Location: IBM paper abstract. An account by the project researchers.\n[4] GPUs and CUDA NVIDIA CUDA Programming Guide — Introduction\nNVIDIA CPU vs GPU — What’s the Difference\nLocation: guide introduction and official historical account. Distinguishes the 2006 architecture introduction from the 2007 software release.\n[5] Early AWS services AWS Our Origins Location: paragraphs on the launch of S3 and EC2. Official historical account.\n[6] AlexNet Krizhevsky, Sutskever \u0026amp; Hinton (2012) — ImageNet Classification with Deep Convolutional Neural Networks Location: Sections 3 and 6, Table 2. Training hardware refers to an individual network; the competition score is for an ensemble.\n[7] Historical training compute Amodei \u0026amp; Hernandez / OpenAI (2018) — AI and Compute Location: opening and Overview. Original 2018 analysis of its historical sample.\n[8] Volta and Tensor Cores NVIDIA (2017) NVIDIA Launches Revolutionary Volta GPU Platform Official announcement, 2017-05-10. Used for the launch date and matrix-computation function.\n[9] First-generation TPU Jouppi et al. (2017) — In-Datacenter Performance Analysis of a Tensor Processing Unit Location: paper abstract. Supports deployment in 2015 and its inference role.\n[10] Transformer Vaswani et al. (2017) — Attention Is All You Need Location: abstract and Section 4. Findings relate to the paper’s machine-translation experiments.\n[11] GPT-3 and ChatGPT Brown et al. (2020) — Language Models are Few-Shot Learners\nOpenAI (2022) — Introducing ChatGPT\nLocation: GPT-3 abstract and ChatGPT launch page. A model paper and the product announcement of 2022-11-30, respectively.\n[12] Llama 3 training cluster Meta Llama Team (2024) — The Llama 3 Herd of Models Location: Sections 3.3.1–3.3.4. An engineering example from a particular training project.\n[13] Cooling configurations NVIDIA — GB200 NVL72 product information\nNVIDIA — H100 product specifications\nLocation: GB200 NVL72 rack description and H100 Form Factor specification. Evidence for the named product configurations.\n[14] PagedAttention Kwon et al. (2023) — Efficient Memory Management for Large Language Model Serving with PagedAttention Location: abstract and evaluation. Throughput findings depend on the tested models, workloads and latency constraints.\n[15] Data center electricity IEA (2026) — Key Questions on Energy and AI: Executive Summary Institutional report, 2026-04-16; executive summary. 485 TWh is a 2025 estimate and 950 TWh a 2030 forecast; both include non-AI uses.\n[16] TSMC: fabrication and packaging TSMC — CoWoS technology\nTSMC — 2Q26 results presentation (2026-07-16)\nTSMC — 2Q26 earnings-call transcript\nLocation: technology page; presentation p. 4; call transcript p. 10. Revenue is company-wide; the packaging constraint is management’s assessment at that time.\n[17] HBM product specifications Micron — HBM4 Location: specifications for the 12-high product. Capacity and bandwidth are supplier component specifications, not measured server performance.\n[18] NVIDIA quarterly revenue NVIDIA — Q2 fiscal 2027 results (2026-08-26) Location: Data Center discussion and the quarter ended 2026-07-26. The revenue covers computing, networking and other products in that business.\n[19] AMD quarterly revenue AMD — Q2 2026 financial results (2026-08-04) Location: Data Center discussion; quarter ended 2026-06-27. The segment includes server CPUs and GPUs.\n[20] Broadcom quarterly revenue Broadcom — Q3 fiscal 2026 results (2026-09-02) Location: AI semiconductor revenue discussion; quarter ended 2026-08-02. Uses reported revenue, not guidance for the following quarter.\n[21] AWS custom silicon AWS — Amazon EC2 Trn3 UltraServers availability (2025-12-02) Official availability announcement. Supports availability, not deployment volume or market share.\n[22] Google TPU products Google — Introducing TPU 8t and TPU 8i (2026-04-22) Official product introduction. Supports workload specialization, not general availability or installed volumes.\n[23] Dell: revenue, orders and backlog Dell — Q2 fiscal 2027 financial results (2026-09-01) Location: AI server commentary and the quarter ended 2026-07-31. Revenue, orders and backlog are cited as separate measures.\n[24] Vertiv: power and thermal management Vertiv — Q2 2026 results, SEC Exhibit 99.1 (2026-07-29) Location: net sales and company description; quarter ended 2026-06-30. Figures are company-wide.\n[25] Google Cloud revenue scope Alphabet — Q2 2026 earnings release (2026-07-22)\nAlphabet — Form 10-Q, quarter ended 2026-06-30\nLocation: release segment-revenue table; Google Cloud in the 10-Q revenue-recognition discussion. Includes cloud services, subscriptions and product sales.\n[26] CoreWeave: growth and operating costs CoreWeave — Q2 2026 earnings release (2026-08-11) Location: highlights, revenue-backlog definition and income statement; quarter ended 2026-06-30. Net loss is GAAP; backlog is not cash received.\n[27] Ascend software architecture Huawei — CANN Community Edition 8.5.0 documentation Location: CANN architecture and functionality. Technical documentation, not comparable market-share evidence.\n[28] Conditions for inference comparisons MLCommons — MLPerf Inference: Datacenter Location: benchmark description, scenarios and quality requirements. Used for comparison principles, not vendor rankings.\n[29] ASE advanced packaging ASE — VIPack™ Location: VIPack overview and six packaging technology pillars, especially the passages on 2.5D/3D architectures and HBM interconnects.\n[30] SK hynix high bandwidth memory SK hynix Begins Volume Production of Industry’s First HBM3E (2024-03-19) Location: Opening paragraph of the 19 March 2024 release, the HBM definition and the discussion of AI processor–memory interconnections.\n[31] Samsung high bandwidth memory Samsung Semiconductor — HBM Location: HBM overview describing TSV stacking, AI training and HPC; used to establish business scope, not customer qualification or market share.\n[32] Supermicro GPU servers Supermicro — GPU Servers for AI, Deep / Machine Learning \u0026amp; HPC Location: The Liquid Cooled GPU Systems and Air Cooled GPU Systems categories under Supermicro GPU Servers.\n[33] Schneider Electric data center infrastructure Schneider Electric — AI Data Centers e-guide (2026-02-11) Location: Opening portfolio description on the guide download page; version 1.5, document 998-2372985_AI_Ready_DC.\n[34] Microsoft Azure accelerated computing Microsoft Azure — Virtual Machine series Location: The N Family — GPU accelerated virtual machines section, including the roles of the ND, NC and NV series.\nBrand icon sources and licenses are listed in the asset credits.\n","date":"2026-10-01T21:45:00+08:00","image":"/post/ai-computing-infrastructure/cover.png","permalink":"/post/ai-computing-infrastructure/","title":"AI and Computing Infrastructure: Industry History, Value Chain, and Challenges"},{"content":"Send a message to DEX Research. Corrections, collaboration, feedback, and inquiries are welcome. Get in touch Use the form below to send a message. I read every submission and will reply when appropriate.\nTypical topics:\nReport corrections or source suggestions Dataset feedback Research collaboration Sponsorship / partnership inquiries General questions about the methodology Name Email Message Send Message You can also reach me directly:\nEmail: dex222444@gmail.com Proton Mail: qizhangdong325@proton.me ","date":"2026-09-30T00:00:00Z","permalink":"/contact/","title":"Contact \u0026 Messages"},{"content":"An evidence-based guide to semiconductor history, design and manufacturing, market structure, AI demand, policy, and supply-chain risks. Scope and Data Notes In this report, the “semiconductor industry” includes chip design, electronic design automation (EDA) software and semiconductor intellectual property (IP), manufacturing equipment and materials, wafer fabrication, packaging and testing, and major end uses. Historical events are dated to when they occurred. Company financials, capacity, and process developments generally reflect information through December 31, 2025; policy information is updated through September 27, 2026. Market-share figures are cited only when the source specifies the market boundary, time period, and measurement basis. Definitions of “foundry,” “AI accelerator,” and “advanced process” vary across organizations, so figures using different definitions should not be compared directly.\nExecutive Summary Semiconductors are not a single market. They are a cross-border industrial network spanning design tools, architecture and circuit IP, manufacturing equipment, critical materials, wafer fabrication, packaging and testing, and end systems. Development cycles, capital needs, and business models differ substantially across these activities. Software and IP businesses depend on research, ecosystem compatibility, and long-term licensing relationships. Equipment and materials require lengthy process qualification. Wafer fabrication relies on sustained capital investment, yield learning, and capacity utilization. Packaging and testing are expanding from conventional back-end services into system-level integration. A single “smile curve” or industry-wide gross-margin range cannot adequately describe this structure.\nThree forces have shaped the industry over the past eight decades. The first is progress in devices and manufacturing, from the transistor, integrated circuit, and planar process to FinFETs, gate-all-around transistors, and extreme ultraviolet (EUV) lithography. The second is a change in industrial organization. The rise of the dedicated foundry allowed design and manufacturing to be performed by different companies, supporting a specialized ecosystem of fabless designers, foundries, outsourced assembly and test providers (OSATs), and tool suppliers. The third is a shift in demand, from mainframes and consumer electronics to personal computers, mobile communications, cloud computing, automotive electronics, and today\u0026rsquo;s AI infrastructure.\nBy the end of 2025, AI training and inference demand was driving investment in advanced logic, advanced packaging, high-bandwidth memory (HBM), and high-speed interconnects. That growth, however, was not reaching every semiconductor category equally. Mature-node chips, analog and power devices, consumer memory, and industrial semiconductors remained subject to their own inventory cycles and end-market demand. Governments were also using incentives, research programs, and export controls to strengthen supply security. As a result, supply-chain decisions increasingly balance cost and scale against compliance, regional capacity, and resilience.\n1. How the Industry Developed 1.1 From Vacuum Tubes to Silicon Transistors Electronic computers of the 1940s relied heavily on vacuum tubes. Tubes could amplify and switch electrical signals, but their size, power consumption, heat, and limited service life constrained miniaturization and reliability. In 1947, a Bell Laboratories team developed the point-contact transistor, establishing solid-state devices as a promising alternative.H18\nEarly transistors were mainly made of germanium. In 1954, Morris Tanenbaum at Bell Laboratories produced a silicon transistor, while a Texas Instruments team led by Gordon Teal produced the first commercial silicon transistors. Silicon eventually prevailed not only because of its high-temperature performance and availability, but also because its interface with silicon dioxide permits a stable insulating layer and repeatable, scalable manufacturing processes.H01\nIn 1957, eight engineers left Shockley Semiconductor Laboratory to establish Fairchild Semiconductor. Fairchild and the companies that grew out of it became an important part of Silicon Valley\u0026rsquo;s semiconductor startup network. The defensible conclusion is that this event accelerated the circulation of technical talent, venture capital, and new firms; it was not the sole origin of Silicon Valley\u0026rsquo;s entrepreneurial culture.\nOne of the key Bell Laboratories patents associated with the point-contact transistor is John Bardeen and Walter Brattain\u0026rsquo;s US 2,524,035, Three-Electrode Circuit Element Utilizing Semiconductive Materials.H15 Its patent grant date is distinct from the 1947 laboratory demonstration.\n1.2 Integrated Circuits, the Planar Process, and Moore\u0026rsquo;s Law Replacing vacuum tubes with individual transistors did not solve the problems of connecting large numbers of components or manufacturing them at scale. Integrated circuits and the planar process emerged in the late 1950s. Through oxidation, photolithography, diffusion, and metal interconnection, the planar process made it possible to form and connect multiple devices on the surface of a single silicon wafer. It laid the foundation for high-volume monolithic integrated circuits.\nEarly integrated circuits took different technical approaches. Jack Kilby\u0026rsquo;s relevant patent is US 3,138,743, Miniaturized Electronic Circuits. Robert Noyce\u0026rsquo;s is US 2,981,877, Semiconductor Device-and-Lead Structure.H16H17\nIn 1965, Gordon Moore used the limited data then available to predict that the number of components on an integrated circuit would roughly double every year for the next decade. In 1975, he revised the cadence to approximately every two years.H02 What became known as “Moore\u0026rsquo;s Law” was both an empirical observation and a reference point for coordinating technology roadmaps across design, equipment, materials, and manufacturing. It does not imply that the price of every chip automatically falls. Whether the cost per function declines also depends on die area, yield, design complexity, packaging, and utilization.\n1.3 Microprocessors, Memory Competition, and US–Japan Adjustments Intel introduced the 4004 in 1971. Developed for a calculator, it was a commercial four-bit microprocessor containing approximately 2,300 transistors.H03 Its significance lay in showing that a general-purpose programmable processor could be sold as a standardized product. The personal-computer market subsequently emerged through the combined development of eight- and 16-bit processors, memory, software, and complete computer systems. The 4004 alone did not “directly launch the PC era.”\nFrom the late 1970s through the 1980s, DRAM became a focal point of competition between Japanese and US companies. Japan\u0026rsquo;s Ministry of International Trade and Industry supported a VLSI research program, while manufacturers\u0026rsquo; production capabilities, quality control, and domestic electronics demand also contributed to their growth.H14 A historical study by the US International Trade Commission reports that Japanese firms\u0026rsquo; share of the global DRAM market rose from less than 30% in 1978 to nearly 75% in 1986.H04 Those dated figures are more precise than a general claim of “nearly 80% in the mid-1980s.”\nFacing price competition in memory, Intel exited DRAM around 1985 and redirected resources to microprocessors. Its consumer-facing Intel Inside cooperative marketing program formally began in 1991.H05 The 1986 US–Japan Semiconductor Agreement primarily addressed access to the Japanese market and anti-dumping concerns. Later arrangements referred to an industry expectation that foreign suppliers would reach a 20% share of the Japanese market, not a binding floor reserved for US chips.H06\n1.4 Dedicated Foundries and Vertical Specialization Vertically integrated manufacturers dominated the industry\u0026rsquo;s early years, often handling product definition, design, wafer fabrication, packaging, and testing within one company. It would nevertheless be inaccurate to say that all companies followed the integrated device manufacturer (IDM) model. Specialization expanded as process development and fab construction became more expensive.\nTSMC was founded in 1987 and built its business around a dedicated foundry model: it manufactured customers\u0026rsquo; designs without selling its own branded chips.H07 This model enabled design companies to bring products to market without building advanced fabs, while foundries aggregated demand from multiple customers to spread process R\u0026amp;D and capacity investment. It created more room for fabless companies such as Qualcomm, NVIDIA, and Broadcom. AMD moved toward a fabless model much later, after spinning off manufacturing assets to GlobalFoundries in 2009–2010.H08\n1.5 Immersion Lithography, FinFETs, and EUV In the early 2000s, the industry faced growing pressure to improve the resolution of 193 nm argon-fluoride (ArF) lithography. A 157 nm exposure path had been explored, but it posed challenges for materials and optical systems. Immersion lithography placed ultrapure water between the projection lens and wafer, increasing numerical aperture and improving resolution and depth of focus while retaining the 193 nm light source. The wavelength remained 193 nm; resolution improved through the larger numerical aperture. In 2003, TSMC ordered the industry\u0026rsquo;s first immersion lithography tool from ASML.H19 Commercial production still required collaborative work across fabs, optics, light sources, photoresists, and research institutions.H09S01\nIn transistor architecture, Hitachi researchers demonstrated a precursor to the FinFET in 1989. In the late 1990s, a University of California, Berkeley team involving Chenming Hu further developed and named the FinFET.H10H13 Intel began high-volume production of its 22 nm tri-gate transistor in 2012.H11 FinFET is therefore best understood as the product of sustained research and industrialization by multiple teams, rather than the invention of a single researcher.\nEUV lithography uses 13.5 nm light. ASML delivered its first production-oriented EUV system in 2013, and customers gradually adopted EUV for advanced logic and memory production later in the 2010s. The first High-NA EUV system was delivered in 2023.H12 Prices, configurations, and revenue-recognition practices differ significantly across system generations; any quoted equipment price must specify the model, year, currency, and accounting basis.\n1.6 Mobile Computing, AI, and Heterogeneous Integration Smartphones increased demand for highly integrated, low-power systems on a chip (SoCs), radio-frequency front ends, image sensors, and mobile memory. They also helped drive advanced manufacturing from planar transistors toward FinFETs. More recently, generative AI has shifted attention toward parallel computing, HBM, high-speed networks, and advanced packaging. GPUs are well suited to massively parallel workloads, but CPUs, GPUs, purpose-built accelerators, and network processors generally work together in a system. “GPUs replace CPUs” is too simple a description.\nAs the cost of designing and producing a single large die rises, chiplets and advanced packaging have become important ways to scale systems. A chiplet architecture does more than mechanically split a large logic chip. It assigns compute, input/output, cache, or analog functions to separately designed and manufactured dies, then integrates them through standard or proprietary interconnects. “More than Moore” is broader: it also encompasses the extension of functions such as sensing, radio frequency, power, optoelectronics, and heterogeneous materials. It should not be treated as synonymous with chiplets.\n2. The Semiconductor Value Chain INDUSTRY MAP\nThe semiconductor value chain See how design, production capabilities and end markets fit together.\nFit to view Expand all Collapse Full screen Read the full text outline SemiconductorsUpstream · Design \u0026amp; inputsChip designCPUs, GPUs and accelerators Analog, power and RF EDA \u0026amp; reusable IPDesign and verification tools Processor and interface IP Equipment \u0026amp; materialsLithography and process tools Wafers, gases and chemicals Midstream · ProductionWafer fabricationDedicated foundries Integrated device manufacturers Assembly \u0026amp; packagingConventional packaging Advanced packaging and chiplets TestingWafer probing Final test and reliability Downstream · End marketsComputingData centers and AI PCs and smartphones Connected systemsAutomotive Telecommunications Specialized applicationsIndustrial and energy Defense and aerospace DEX editorial map based on the accompanying report. Examples are illustrative, not exhaustive or ranked. Companies can operate across several stages; connections show categories, not verified supplier contracts.\nSources: Semiconductor industry primer — production stages. Reviewed 2026-09-29.\n2.1 Chip Design, EDA, and Semiconductor IP Chip design begins with product requirements and system architecture, then proceeds through logic design, functional verification, synthesis, placement and routing, timing closure, physical verification, and tape-out preparation. EDA software links design rules, foundry process design kits, and manufacturing constraints. Its value comes from algorithms, complete tool flows, process compatibility, and years of accumulated validation data.\nSemiconductor IP consists of designed and verified modules that can be reused in a chip, including processor cores, memory controllers, PCIe, DDR, USB, SerDes, and security blocks. An instruction set architecture (ISA) must be distinguished from processor IP. Arm licenses both architectures and processor-core IP. RISC-V is an open-standard ISA, not a processor core that can be manufactured directly; companies must still develop or license a specific implementation.S02 x86 is a proprietary ISA ecosystem, with Intel and AMD as its principal product suppliers.\nDigital devices include CPUs, GPUs, microcontrollers, FPGAs, SoCs, network processors, and AI accelerators. Analog and mixed-signal chips manage power, data conversion, amplification, and sensor interfaces. RF and optoelectronic devices handle wireless transmission and reception, filtering, power amplification, and conversion between electrical and optical signals. These categories differ in design cycle, software dependence, product life, and process needs. An advanced node is not the only measure of a chip\u0026rsquo;s value.\n2.2 Manufacturing Equipment and Materials Front-end equipment includes lithography, etching, thin-film deposition, ion implantation, thermal processing, cleaning, chemical-mechanical polishing (CMP), metrology, and defect inspection systems. Back-end equipment includes thinning and dicing, die attach, bonding, molding, probe systems, automatic test equipment (ATE), and sorting equipment. Atomic layer deposition (ALD) is especially useful for thickness control and conformal coverage in high-aspect-ratio structures. “High selectivity” applies to particular selective deposition processes, not to every ALD tool.\nCritical materials include silicon wafers and compound-semiconductor substrates, photoresists, masks, electronic gases, wet chemicals, deposition precursors, CMP consumables, sputtering targets, package substrates, and bonding materials. Purity specifications vary by material and process; they cannot all be summarized as “nine nines.” In gas classification, phosphine, arsine, and diborane can be used for doping. Nitrogen trifluoride is primarily used for chamber cleaning and some etching processes, rather than as a typical dopant gas.\nQualification for high-volume production often takes substantial time. A supplier must demonstrate more than the specifications of a single tool or material batch: customers also need stable performance across lots, defect control, service capability, and compatibility with their process platform. This joint optimization helps explain high concentration in some niches. It does not mean that every segment has only one supplier.\n2.3 Wafer Fabrication Wafer manufacturers are commonly divided into IDMs and foundries. An IDM sells its own products and performs at least some manufacturing; a dedicated foundry primarily manufactures customer designs. In practice, the boundary is not absolute. Some IDMs offer foundry services to external customers, while some systems companies take a direct role in chip design and supply-chain management.\nA typical front-end process repeatedly applies film formation, photoresist coating, exposure, development, etching, ion implantation, thermal processing, cleaning, and CMP to form transistors and multiple interconnect layers on a wafer. After front-end fabrication, a foundry delivers a processed wafer or diced dies, not a “bare wafer.” A bare wafer is generally a substrate on which device structures have not yet been formed.\nProcess-node names identify generations of manufacturing platforms; they no longer correspond to a single directly measurable physical dimension. “2 nm” or “Intel 18A” therefore does not mean that every transistor feature measures 2 nm or 1.8 nm. Process capability should be assessed through transistor architecture, density, performance, power, yield, design rules, and production status. Intel 18A uses RibbonFET gate-all-around transistors and PowerVia backside power delivery. In 2025, Intel disclosed that the first 18A client product had entered production and that it planned to begin high-volume production that year.S03S06\n2.4 Packaging and Testing Conventional packaging protects the die, provides electrical and mechanical connections, and supports assembly into a system. Advanced packaging also enables dense interconnects, more bandwidth, power management, and heterogeneous integration. Flip-chip packaging connects a die to its substrate through bumps. In 2.5D packaging, a silicon interposer or redistribution structure can connect multiple side-by-side dies. In 3D packaging, dies are stacked using hybrid bonding, through-silicon vias (TSVs), or other vertical interconnects. CoWoS is a 2.5D and related advanced-packaging platform; it should not be conflated with every form of 3D stacking.\nHBM typically stacks multiple DRAM dies connected by TSVs and integrates them with a logic chip for high bandwidth. 3D NAND, by contrast, stacks memory cells vertically within a NAND device. It is a device structure and manufacturing process, not a synonym for TSV-based die stacking.S04\nTesting includes wafer-level probing, final testing after packaging, and reliability evaluation for particular uses. Automotive integrated circuits commonly undergo failure-mechanism-based stress tests and customer qualification under specifications such as AEC-Q100. These specifications are not equivalent to a “certification certificate” issued by a single organization.S05\n2.5 End Markets Data centers use CPUs, GPUs and dedicated accelerators, HBM, network switches and optical interconnects, power-management devices, and security chips. Smartphones and PCs balance performance, energy use, wireless connectivity, and cost. Automotive electronics encompass microcontrollers, analog and power devices, cockpit and driver-assistance processors, sensors, and battery-management chips; they also require lengthy qualification and supply commitments. Industrial, renewable-energy, telecommunications, and defense applications place particular weight on reliability, long-term supply, environmental tolerance, or specific security requirements.\nAdvanced nodes are not essential for every application. Power management, analog, RF, sensors, and many automotive and industrial products continue to use mature processes extensively. An industry assessment should therefore track both advanced-node investment and mature-node inventories, utilization, and replacement demand.\n3. Market Structure 3.1 Concentration and Interdependence Some semiconductor segments are highly concentrated because R\u0026amp;D is expensive, customer qualification takes time, process knowledge is difficult to replicate quickly, and software and hardware ecosystems raise switching costs. That does not establish that the top three suppliers hold 70%–90% in “almost every” subsector. A sound market-share statement first defines the market—for example, complete EUV systems, discrete data-center GPUs, DRAM, foundry services, or OSAT—and then specifies the geography, period, and whether it measures revenue or shipments.\nRegional specialization likewise cannot be reduced to closed “blocs.” US companies are strong in EDA, processor and accelerator design, and parts of the equipment market. Europe is prominent in lithography, optics, and certain automotive and industrial chips. Japan has important materials, equipment, and image-sensor companies. South Korea has large-scale memory producers. Taiwan plays a central role in foundry and packaging. Mainland China is expanding its mature-process, packaging, equipment, and materials capabilities. Cross-border investment, customer relationships, and supply dependencies remain extensive.\n3.2 Chip Design and AI Computing General-purpose processors, mobile SoCs, analog chips, and AI accelerators each have different competitive structures. NVIDIA leads in data-center GPUs and their software ecosystem, but a claim that it holds 80%–90% of “AI training and inference chips” lacks a consistent market boundary. In its review of NVIDIA\u0026rsquo;s proposed acquisition of Run:ai, the European Commission estimated that NVIDIA\u0026rsquo;s shipment share of the defined global discrete data-center GPU market had exceeded 80%–90% in several preceding years and stood at 70%–80% in the first half of 2024. The decision also cautioned that shipment estimates inferred from revenue and average selling prices were less reliable.M01 The case illustrates why market share must be reported with its product scope, date, and method.\nAMD and Intel offer GPUs or other accelerators, while cloud providers develop in-house or custom ASICs such as TPUs and Trainium. In-house chips can improve performance, cost, or supply control for specific workloads, but they do not automatically displace commercial GPUs. Their results depend on software tools, utilization, model fit, networking, and deployment scale.\n3.3 Foundries and Advanced Processes MARKET SHARE · Q4 2025\nGlobal wafer foundry revenue share Wafer foundry revenue under TrendForce\u0026#39;s market definition. Samsung excludes System LSI. This is not total semiconductor revenue or the expanded Foundry 2.0 market.\nWorldwide · Share of foundry revenue (%)\nFull screen View the data table Global wafer foundry revenue share · Q4 2025 · Share of foundry revenue (%) CompanyShare TSMC70.4%Samsung Foundry7.1%SMIC5.2%UMC4.2%GlobalFoundries3.8%Other foundries9.3% Download data (CSV) Five largest suppliers shown. Other foundries = 100% minus the five published shares and includes both other ranked and unranked suppliers. Percentages retain the source rounding. Historical quarter; not full-year 2025, all chip sales, or the broader Foundry 2.0 definition.\nSource: TrendForce — AI Demand Drives 4Q25 Global Top 10 Foundries Revenue Up 2.6% QoQ; Samsung Gains Share and Tower Moves Up in Rankings (2026-03-12). Reviewed 2026-09-29.\nTSMC is the leading dedicated foundry. In its 2025 annual report, the company defined “Foundry 2.0” broadly to include logic wafer fabrication, packaging, testing, masks, and non-memory IDM activity, and estimated that market at US$305 billion in 2025. This is substantially broader than conventional dedicated foundry services; a Foundry 2.0 share should not be directly compared with a third-party pure-foundry share. TSMC also reported that its 3 nm process accounted for 24% of its own wafer revenue in 2025 and that its 2 nm process entered volume production in the fourth quarter of that year.M02 Those figures describe TSMC\u0026rsquo;s revenue mix and manufacturing progress, not the entire industry\u0026rsquo;s 3 nm or 2 nm market share.\nSamsung operates in memory, logic products, and foundry services, so its process investment must be considered alongside both internal IDM demand and external foundry customers. Intel offers manufacturing and packaging to external customers through Intel Foundry; the scale of 18A production and external customer adoption should be updated against subsequent earnings reports and product deliveries. SMIC, UMC, and GlobalFoundries also have different product mixes, process platforms, customer industries, and expansion priorities.\n3.4 Memory and Advanced Packaging The major DRAM suppliers include Samsung Electronics, SK hynix, and Micron. NAND participants also include Kioxia, Western Digital/SanDisk-related operations, and Solidigm. Memory is highly cyclical: prices respond to inventory, capital expenditure, product transitions, and end-market demand. HBM growth has prompted suppliers to invest in advanced DRAM, TSVs, and packaging. Any assertion that “two companies hold the overwhelming majority of HBM” should identify the quarter and whether it measures revenue or bit shipments, while accounting for changes at suppliers including Micron.\nFoundries, memory makers, IDMs, and OSATs all participate in advanced packaging. ASE, Amkor, and JCET are major OSAT providers, while TSMC, Samsung, and Intel combine advanced packaging with front-end processes. Control is not simply shifting in one direction from OSATs to foundries. Platforms compete and overlap in interposers, hybrid bonding, HBM integration, testing, and volume delivery.\n3.5 Equipment, Materials, and Profitability ASML is currently the only company able to supply complete EUV lithography systems commercially. DUV, metrology, inspection, and other manufacturing-equipment markets have different competitors. ASML\u0026rsquo;s 2025 annual report records €32.7 billion in total net sales, a gross margin of 52.8%, and revenue recognition for 48 EUV systems during its 2025 fiscal year.M03 These figures illustrate the scale and technical barriers of the EUV business. They do not support a claim that every equipment monopoly earns a 60%–80% gross margin.\nTSMC reported a gross margin of 59.9% for 2025. Revenue recognition, depreciation, and cost structures differ among EDA, IP, equipment, materials, foundries, and packaging and test providers.M02 Profitability should therefore be analyzed using specific companies and a consistent fiscal year and accounting basis. At minimum, software licenses, equipment sales, materials, manufacturing, and testing should be distinguished rather than assigned fixed margins across the value chain.\n4. Principal Risks and Constraints 4.1 Industry Cycles and Concentrated AI Demand AI infrastructure is increasing demand for advanced logic, HBM, advanced packaging, networks, and power devices, but semiconductors remain subject to inventory and capital-spending cycles. If cloud providers slow investment, relevant suppliers could face order revisions and lower utilization. If AI-related revenue and computing demand continue to grow, advanced capacity could remain tight in the near term. These are conditional scenarios, not grounds for declaring an inevitable “ROI cliff” or “profit collapse.”\nAI capacity does not crowd out every traditional chip category in equal measure. Advanced GPUs and automotive microcontrollers generally use different nodes and production lines, limiting direct substitution. HBM expansion may redirect some DRAM resources, but consumer-memory prices also depend on inventory, demand, and suppliers\u0026rsquo; capital discipline. Risk analysis needs to distinguish products and processes.\n4.2 Industrial Policy, Export Controls, and Regionalization The US CHIPS and Science Act provided the Department of Commerce with US$50 billion for manufacturing incentives, R\u0026amp;D, and related programs. That figure represents statutory program funding, not cash already paid to companies.R01 The European Chips Act took effect in September 2023. The EU set a policy goal of raising its share of the global semiconductor market to 20% by 2030; that number is a target, neither an achieved share nor a firm forecast.R02 In June 2026, the European Commission proposed a Chips Act 2.0 to build on the original law. The proposal should be distinguished from the 2023 act already in force.R05\nExport controls are changing customer screening and delivery procedures for equipment, software, HBM, and advanced computing chips. In January 2025, the US Bureau of Industry and Security updated advanced-computing controls and foundry due-diligence requirements; related rules also changed definitions of advanced-node integrated circuits and the Entity List.R03 Businesses consequently face licensing, end-user, resale, technical-service, and geographic compliance risks. Policies can change, so a rule in force at one point should not be treated as a permanent industrial boundary.\nRegional incentives can add local capabilities and geographic redundancy, but they can also raise construction costs, reduce utilization, intensify competition for talent, and complicate cross-border operations. Whether a project amounts to “duplicative capacity” depends on actual demand, its technology generation, and long-term utilization. Not every localization project can be assumed in advance to destroy economies of scale.\n4.3 Technical Complexity and Recovery of Capital Investment Advanced processes face short-channel effects, interconnect delay, power density, heat, stochastic defects, and growing design complexity. Gate-all-around transistors, backside power delivery, EUV, High-NA EUV, and advanced packaging offer new ways to scale, while increasing R\u0026amp;D, equipment, mask, design-migration, and yield-ramp costs. A node label is not a physical-limit gauge. The characters “2 nm” alone cannot establish that quantum tunneling has become the decisive obstacle for every product.\nInvestment in a fab or critical tool varies substantially with the project boundary, cleanroom, equipment mix, capacity, and location. A claim that a fab costs US$20–30 billion, or that a certain tool has a particular price, should identify a specific project or system, announcement date, currency, and whether infrastructure is included. Totals from different projects should not be substituted for one another.\n4.4 Supply Concentration and Operational Continuity Concentrated supply does create single-point risks. ASML is currently the sole commercial supplier of complete EUV systems, and some equipment subsystems, material formulations, and advanced-packaging capacities are concentrated among a small number of firms. Yet photoresists, metrology and inspection, and most process materials generally have multiple suppliers—even if an alternative cannot be qualified and substituted quickly. More useful risk measures include qualification time for replacements, inventory coverage, geographic concentration, capacity flexibility, and the cost of an alternative process. A blanket claim that there is “no second option anywhere in the world” obscures these differences.\nCompanies can reduce exposure through developing second sources, stocking critical spare parts, diversifying production sites, signing long-term purchase agreements, and conducting joint qualification. Because semiconductor tools and materials must be qualified against specific processes, establishing an alternative often takes months or longer. That work should begin during normal operations, before a supply interruption.\n4.5 Electricity, Water, and Infrastructure Advanced fabs require reliable electricity, ultrapure water, gases, and waste-treatment systems. AI data centers are increasing demand for high-density computing, cooling, and grid connections. The International Energy Agency estimates that data centers used about 415 TWh of electricity worldwide in 2024, or about 1.5% of global electricity consumption. In its 2025 base case, the IEA projects roughly 945 TWh by 2030.R04 These are global model estimates; they do not mean that every regional grid will reach its limits at the same time.\nPower constraints vary sharply by location, depending on grid-connection queues, generation mix, transmission and distribution capacity, and data-center clustering. Semiconductor companies should evaluate power reliability, water availability, extreme weather, and carbon costs when selecting sites. Data-center customers should also incorporate server utilization, model efficiency, and cooling methods into capacity planning.\n4.6 Talent and Organizational Capability Semiconductor production requires specialists in devices, materials, chemistry, mechanics, optics, software, quality, and equipment maintenance. New fabs need more than additional graduates: they need experienced teams that can introduce processes into volume production, improve yield, and keep tools running. Talent risk should not be described as an equally “severe shortage” in every region. It should be measured by project location, job category, hiring time, and turnover.\nIf industrial policy subsidizes buildings and machines without vocational training, research platforms, supplier engineering capacity, and arrangements for international talent mobility, new capital may be slow to turn into stable output. Companies should include training periods, succession for critical roles, replication across fabs, and supplier-service capacity in their expansion plans.\nConclusion Technological accumulation, specialization, and cross-border collaboration define the semiconductor industry. Advances in transistor architecture, lithography, materials, design tools, fabrication, and packaging depend on one another. No single segment determines the outcome on its own. Dedicated foundries separated design and manufacturing, while advanced packaging is bringing front-end and back-end work closer together again. AI has increased demand for advanced computing, but has also deepened dependence on HBM, interconnects, electricity, and software ecosystems.\nAssessing a company or market requires comparable definitions and data. A market-share figure needs a time period, geography, product boundary, and measurement basis. An equipment price needs a model and year. A process node cannot be read as a literal physical dimension. Gross margins should be compared only under consistent accounting conventions. Unpublished yields, customer confidence, or future capacity should not be presented as established facts.\nOver the next several years, competition will center on four capabilities: advancing device and system technologies; maintaining efficient volume production despite heavy capital spending; building auditable cross-border supply networks; and securing power, talent, and critical materials. Regionalization will increase the weight of compliance and redundancy in investment decisions without fully replacing global specialization. Companies that assess technology roadmaps, customer demand, capital discipline, and supply security together will be better placed to navigate both growth and cyclical volatility.\nReferences Historical Sources [H01] Computer History Museum, “1954: Silicon Transistors Offer Superior Operating Characteristics.” https://www.computerhistory.org/siliconengine/silicon-transistors-offer-superior-operating-characteristics/\n[H02] Intel, “Moore’s Law.” https://www.intel.com/content/www/us/en/newsroom/resources/moores-law.html\n[H03] Intel, “The Chip that Changed the World.” https://newsroom.intel.com/opinion/the-chip-that-changed-the-world\n[H04] U.S. International Trade Commission, “The South Korea-Japan Trade Dispute in Context: Semiconductor Manufacturing, Chemicals and Concentrated Supply Chains.” https://usitc.gov/sites/default/files/publications/332/working_papers/semiconductor_working_paper_corrected_103119.pdf\n[H05] Intel, “Ingredient Branding: End User Marketing and Intel Inside.” https://www.intel.com/content/www/us/en/history/virtual-vault/articles/end-user-marketing-intel-inside.html\n[H06] Office of the United States Trade Representative, “1996 National Trade Estimate—Japan: Semiconductors.” https://ustr.gov/archive/Document_Library/Reports_Publications/1996/1996_National_Trade_Estimate/1996_National_Trade_Estimate-Japan.html\n[H07] TSMC, “2025 Annual Report—About TSMC.” https://investor.tsmc.com/static/annualReports/2025/english/index.html\n[H08] AMD, “AMD Reports Fourth Quarter and Annual Results,” January 21, 2010. https://ir.amd.com/financial-information/sec-filings/content/0001193125-10-009806/dex991.htm\n[H09] ASML, “How Immersion Lithography Saved Moore’s Law,” 2023. https://www.asml.com/en/company/stories/2023/how-immersion-lithography-saved-moores-law\n[H10] University of California, Berkeley EECS, “History.” https://eecs.berkeley.edu/about/history/\n[H11] Intel, “Moore’s Law: Fun Facts.” https://www.intel.com/content/www/us/en/history/history-moores-law-fun-facts-factsheet.html\n[H12] ASML, “EUV Lithography Systems.” https://www.asml.com/en/products/euv-lithography-systems\n[H13] IEEE Technology Navigator, “FinFETs.” https://technav.ieee.org/topic/finfets/\n[H14] Ministry of Economy, Trade and Industry of Japan, “2018 White Paper on International Economy and Trade—VLSI Project History.” https://www.meti.go.jp/report/tsuhaku2018/2018honbun/i2220000.html\n[H15] John Bardeen and Walter H. Brattain, US 2,524,035, “Three-Electrode Circuit Element Utilizing Semiconductive Materials.” https://patents.google.com/patent/US2524035A/en\n[H16] Jack S. Kilby, US 3,138,743, “Miniaturized Electronic Circuits.” https://patents.google.com/patent/US3138743A/en\n[H17] Robert N. Noyce, US 2,981,877, “Semiconductor Device-and-Lead Structure.” https://patents.google.com/patent/US2981877A/en\n[H18] Computer History Museum, “1947: Invention of the Point-Contact Transistor.” https://www.computerhistory.org/siliconengine/invention-of-the-point-contact-transistor/\n[H19] ASML, “TSMC Selects ASML for Industry’s First Immersion Tool Order,” December 3, 2003. https://www.asml.com/en/news/press-releases/2003/tsmc-selects-asml-for-industry-first-immersion-tool-order\nTechnology and Value-Chain Sources [S01] ASML, “Lenses and Mirrors—Lithography Principles.” https://www.asml.com/technology/lithography-principles/lenses-and-mirrors\n[S02] RISC-V International, “About RISC-V.” https://riscv.org/about/\n[S03] Intel, “Intel 18A Process Technology Simply Explained,” January 30, 2025. https://newsroom.intel.com/intel-foundry/intel-18a-process-technology-simply-explained\n[S04] Samsung Semiconductor, “3D V-NAND Flash Memory.” https://semiconductor.samsung.com/support/tools-resources/dictionary/semiconductor-glossary-3d-v-nand-flash-memory/\n[S05] Automotive Electronics Council, “AEC-Q100: Failure Mechanism Based Stress Test Qualification for Integrated Circuits,” documents index. https://www.aecouncil.com/AECDocuments.html\n[S06] Intel, “Postcard from Intel Technology Tour Arizona: Panther Lake Draws in Cameras and Crowds,” October 10, 2025. https://www.intel.com/content/www/us/en/newsroom/news/client-computing/postcard-itt-panther-lake-draws-cameras-and-crowds.html\nMarket and Company Sources [M01] European Commission, Case M.11766, NVIDIA/Run:ai merger decision, 2024 market evidence. https://ec.europa.eu/competition/mergers/cases1/202516/M_11766_10599589_2740_3.pdf\n[M02] TSMC, “2025 Annual Report.” https://investor.tsmc.com/static/annualReports/2025/english/index.html\n[M03] ASML, “2025 Annual Report.” https://www.asml.com/en/investors/annual-report/2025\nPolicy and Risk Sources [R01] U.S. Department of Commerce, “Semiconductor Industry—CHIPS for America.” https://www.commerce.gov/issues/semiconductor-industry\n[R02] European Commission, “European Chips Act.” https://digital-strategy.ec.europa.eu/en/policies/european-chips-act\n[R03] U.S. Bureau of Industry and Security, “Commerce Strengthens Restrictions on Advanced Computing Semiconductors,” January 15, 2025. https://www.bis.gov/press-release/commerce-strengthens-restrictions-advanced-computing-semiconductors-enhance-foundry-due-diligence-prevent\n[R04] International Energy Agency, “Energy and AI,” April 10, 2025. https://www.iea.org/reports/energy-and-ai\n[R05] European Commission, “Proposal for the Chips Act 2.0,” June 3, 2026. https://digital-strategy.ec.europa.eu/en/library/proposal-chips-act-20\n","date":"2026-09-28T00:00:00Z","image":"/post/semiconductor-industry-report/cover.webp","permalink":"/post/semiconductor-industry-report/","title":"Semiconductors and Chips: Industry History, Value Chain, Markets, and Risks"},{"content":"A sourced starting map of AI inputs, model platforms and applications, with business-model questions and metrics to track.  From computing inputs to models and paid workflows. artificial intelligence models inference enterprise automation 人工智能 Enterprise AI software and services; excludes a standalone valuation of chipmakers or cloud infrastructure. Who gets paid when an AI capability becomes a useful workflow? Global business-model lens; US framework and vendor examples. Follow customer revenue after inference, support and integration costs. A growing user count alone does not establish attractive unit economics. Compute \u0026 data Cloud providers, data owners and tooling suppliers Supply the resources used to build and run AI systems. Compute consumption, data licensing or tooling contracts. Can data rights and serving costs support the intended use? Models \u0026 deployment Model developers and deployment platforms Turn resources into usable inference and evaluation services. Usage-based APIs, capacity commitments or licenses. Does reliability hold under real workloads, not just benchmarks? Applications \u0026 adoption Software vendors, integrators and enterprise teams Embed a capability into a customer's operating process. Subscriptions, implementation projects or usage fees. Will verified savings translate into repeat paid use? Which workflows have measurable time savings? Do pilots become recurring paid deployments? Model errors, data rights and accountability may block adoption. Falling model prices may help customers without improving every supplier's margin. Cost per completed task Include inference, retries and human review; compare like-for-like tasks. Paid retention Track continued paid use, not registrations or free trials. Task success rate Use a defined test set and disclose human intervention. Amazon Bedrock publishes input/output token pricing for several models, providing a concrete example of usage-based inference billing. NIST's AI RMF is voluntary guidance for managing risk across AI design, development, use and evaluation. AWS Amazon Bedrock Pricing Undated; live pricing page Company documentation Billing-model example only; no price or market-size estimate is reproduced. NIST AI Risk Management Framework AI RMF 1.0 released 2023-01-26 Government framework Risk-management scope; not a revenue forecast or an official value-chain taxonomy. AI risks and industry turbulence","date":"2026-09-22T00:00:00Z","permalink":"/industry-breakdowns/ai-software/","title":"AI Software \u0026 Services — Industry Breakdown"},{"content":"Explore battery materials, systems and project operation, with research questions and clearly scoped IEA evidence.  From minerals and cells to installed storage systems. battery lithium stationary grid minerals cells 储能 电池 Battery supply chains, with a stationary-storage operating lens. EV batteries and grid-storage revenue pools are not interchangeable. Who earns the return: material suppliers, manufacturers or operators? Global supply chain; project economics depend on local power markets. Do not equate cheaper cells with better project returns. Connection, installation, financing, degradation and market rules belong in the same model. Materials \u0026 components Miners, refiners and component suppliers Provide battery-grade inputs and components. Material and component sales under spot or contract terms. Do quality, supply security and chemistry choices align? Cells \u0026 systems Cell makers, pack makers and system integrators Turn components into a tested, controllable storage system. Cell/system sales, integration and warranty services. Can manufacturing quality and warranty obligations be sustained? Projects \u0026 operation Developers, utilities and storage operators Connect and operate assets to serve a particular power market. Contracted or market-based power-system services. Will local revenues cover degradation and full project costs? Where does variable generation create a need for flexibility? Can storage participate and get paid in the relevant local market? Safety, degradation and warranties may change lifecycle costs. Supply concentration and changing project revenues can offset lower hardware prices. Installed system cost Use consistent currency, date and system boundary; distinguish USD/kW from USD/kWh. Usable capacity \u0026 efficiency Measure at the same operating conditions and asset age. Contracted revenue share Separate contracted cash flows from merchant assumptions. The IEA describes concentration across battery supply chains and explains storage's role in flexibility and grid stability. Its 2024 scenarios are not treated here as current forecasts. International Energy Agency Batteries and Secure Energy Transitions — Executive summary 2024-04-25 Intergovernmental analysis Supply-chain structure and power-system applications; historical scenarios, not fresh market estimates.","date":"2026-09-22T00:00:00Z","permalink":"/industry-breakdowns/battery-storage/","title":"Batteries \u0026 Energy Storage — Industry Breakdown"},{"content":"Connect drug discovery, clinical evidence and commercialization through a carefully scoped US regulatory reference.  From scientific discovery to evidence and commercialization. biotech pharmaceuticals drugs clinical trials medicine 生物制药 医药 Innovative drug development. The US pathway is an illustrative regulatory reference; individual products and other jurisdictions require separate analysis. How does a scientific asset become an approved and commercially viable product? US regulatory reference; not a universal approval pathway. Separate scientific progress, approval and commercial success. Model development spending and potential dilution before assuming future product revenue. Discovery \u0026 preclinical Research teams, biotechs and research suppliers Develop candidates and investigate their properties before human studies. Research services or licensing arrangements to investigate. Does the evidence justify further development and funding? Clinical evidence \u0026 review Sponsors, trial providers and regulators Generate human evidence and evaluate a submission for approval. Trial-service contracts or milestone arrangements to investigate. Are safety, efficacy and manufacturing evidence sufficient? Manufacture \u0026 access Manufacturers, commercial teams and health-system buyers Supply approved products while maintaining quality and safety monitoring. Product sales, royalties or manufacturing services to investigate. Does approval translate into supply, access and viable net revenue? What unmet need would a successful product address? Which evidence would change clinical or purchasing decisions? Clinical failure and manufacturing problems can change the entire thesis. Approval alone does not establish reimbursement or commercial demand. Evidence milestones State trial phase, population and endpoints; avoid treating a milestone as guaranteed. Cash runway Test spending assumptions and funding needs under delays. Net realized revenue Distinguish actual sales from headline prices and potential markets. FDA's drug-development overview describes discovery, preclinical research, clinical research, review and post-market safety monitoring. The three cards group those activities for readability. The 1962 U.S. Drug Amendments strengthened the evidence of effectiveness required for new-drug approval. This is historical U.S. law, not a worldwide approval pathway. US Food and Drug Administration The Drug Development Process Undated overview; reviewed on the date shown above Regulator explainer US development and review stages; not clinical advice or evidence for a particular drug. U.S. Congress / GovInfo Drug Amendments of 1962, Public Law 87-781 1962-10-10 Original statute · PDF Historical U.S. effectiveness requirement and regulatory powers; not a global or modern market-size estimate. Pharmaceutical industry report","date":"2026-09-22T00:00:00Z","permalink":"/industry-breakdowns/biopharmaceuticals/","title":"Biopharmaceuticals — Industry Breakdown"},{"content":"Understand security signals, products and operations, and separate business-model interpretation from NIST framework evidence.  From security inputs to products and operational response. security identity threat intelligence detection response zero trust 网络安全 Enterprise security products and services. The commercial layers below are DEX's map, not the NIST CSF functions or a compliance checklist. Does the customer buy another tool, or a better security outcome? Global enterprise lens; NIST provides the risk-management reference. Test renewal quality and delivery cost alongside product claims. Security spending is not itself evidence that customer risk has fallen. Signals \u0026 foundations Telemetry, identity and intelligence providers Supply the information and controls used by security teams. Data feeds, infrastructure or identity-service contracts. Are the signals relevant, complete and lawful to use? Security products Endpoint, cloud, network and identity vendors Turn inputs into prevention, detection and investigation tools. Subscriptions, licenses or consumption-based fees. Can tools integrate without overwhelming operators? Operations \u0026 response Internal teams, managed providers and responders Operate controls and coordinate incident handling and recovery. Managed-service contracts, retainers or response projects. Can the customer turn alerts into timely, effective action? Which assets and workflows create an uncovered exposure? Would integration or managed delivery solve a staffing bottleneck? Product overlap and tool consolidation may affect renewals. Failures in the vendor's own systems can undermine trust. Net revenue retention Use a consistent cohort and vendor definition. Response effectiveness Define incident severity and measurement boundaries before comparing. Service delivery cost Include analyst time and infrastructure per customer. NIST CSF 2.0 organizes cybersecurity outcomes under Govern, Identify, Protect, Detect, Respond and Recover. These are outcomes, not six sequential commercial stages. NIST SP 800-207 sets out zero-trust architectural principles; it is not a market-sizing or vendor-share study. Menlo Ventures' 2022 market map is an illustrative snapshot of product categories and vendors, not a current estimate of spending or company shares. NIST The NIST Cybersecurity Framework (CSF) 2.0 2024-02-26 Government framework · PDF Risk-management functions only; does not establish vendor revenue, effectiveness or market size. NIST SP 800-207: Zero Trust Architecture 2020-08 Government publication Zero-trust architecture concepts; not an estimate of industry revenue. Menlo Ventures Cybersecurity Market Map 2022-09 Investor research · PDF Illustrative historical vendor categories; not a verified 2026 revenue dataset. Cybersecurity industry report","date":"2026-09-22T00:00:00Z","permalink":"/industry-breakdowns/cybersecurity/","title":"Cybersecurity — Industry Breakdown"},{"content":"Understand six industries through their value chains, business models, research questions and primary-source evidence. A visual starting point for understanding how an industry works. Follow the value chain, examine how money moves, and identify what evidence would change your view.\n","date":"2026-09-22T00:00:00Z","permalink":"/industry-breakdowns/","title":"Industry Breakdowns"},{"content":"Map industrial robot components, integration and lifecycle services, with payback questions and metrics to investigate.  From components to integrated, supported automation. robots automation industrial integration actuators cobots 机器人 Industrial robot systems and their integration. This is a subset of the explorer's Industrial + Service Robotics row; not a humanoid or service-robot forecast. Can a robot deliver a repeatable outcome in a real operating environment? Global industrial lens; ABB is one vendor example, not a market ranking. Evaluate the complete installed cell and support burden, not just the robot's purchase price. Pilot demonstrations are not recurring production deployments. Components \u0026 control Motion, sensing and control suppliers Provide the hardware and software building blocks. Component sales and software licenses. Can parts satisfy precision, durability and integration requirements? Robots \u0026 integration Robot manufacturers and system integrators Combine machines, tooling and software around a defined task. Equipment sales and engineering projects. How much customization is required for each deployment? Operation \u0026 lifecycle Factory operators and service providers Commission, maintain and improve the deployed system. Maintenance, training and modernization contracts. Does uptime and throughput justify the full lifecycle cost? Which repetitive tasks have stable requirements and measurable payback? Can integration be reused across multiple customer sites? Customization and commissioning delays can erode project margins. Reliability, operator training and safe integration need on-site validation. Installed-system payback Include integration, downtime, maintenance and training. Productive uptime Measure time actually completing the intended task. Repeat deployments Separate pilots from repeat orders and operating installations. ABB's robotics services cover installation, support, maintenance and modernization, illustrating commercial activity beyond the initial equipment sale. ABB Robotics Services Undated; product/service page Company documentation One vendor's service offerings; not independent evidence of achieved returns or industry-wide revenue mix.","date":"2026-09-22T00:00:00Z","permalink":"/industry-breakdowns/robotics/","title":"Robotics — Industry Breakdown"},{"content":"Follow semiconductor design, wafer fabrication and packaging, and examine the economics and evidence behind each stage.  From chip design to fabrication, packaging and testing. chips integrated circuits foundry design packaging 半导体 芯片 Semiconductor production, with equipment and materials as enabling inputs. Not a ranking of individual companies or process nodes. Which production constraint determines where value is captured? Global production lens; industry-body primer. Separate design economics from capital-intensive manufacturing. A shortage in one process or package does not imply a shortage across all chips. Design Chip designers and design-tool/IP suppliers Specify the device and prepare it for production. Chip sales, design services or intellectual-property licensing. Can the design meet performance, power and cost requirements? Wafer fabrication Foundries, integrated manufacturers and input suppliers Fabricate circuits on wafers using specialized facilities. Manufacturing services or internally produced chip sales. Are yield and utilization sufficient to cover fixed costs? Assembly, test \u0026 packaging Packaging/test providers and integrated manufacturers Separate, test and package devices for customers. Packaging and testing services or bundled chip sales. Can packaging capacity and reliability meet product requirements? Which applications are creating durable design wins? Where do customer requirements exceed available production capability? Capacity additions can arrive after the demand cycle changes. Geographic dependencies and customer concentration require separate analysis. Manufacturing yield Good devices as a share of output, for a specified product/process. Capacity utilization Track actual loading, not announced capacity alone. Inventory \u0026 lead time Read together to distinguish shortages from order inflation. SIA's primer distinguishes design, front-end fabrication and back-end assembly, test and packaging. The three-stage map here follows that production sequence. Semiconductor Industry Association Semiconductor Industry Primer: The Stages of Production and Business Models 2015-02-25 Industry-body explainer Production stages; foundational reference, not current market sizing. Semiconductor industry report","date":"2026-09-22T00:00:00Z","permalink":"/industry-breakdowns/semiconductors/","title":"Semiconductors — Industry Breakdown"},{"content":"Explore a provisional 50-industry screening list, inspect sourced baselines and their limits, and test CAGR scenarios.  Explore a provisional 50-industry screening list and test CAGR scenarios. The original market ranges, growth rates and projections remain unverified research leads. The semiconductor and cybersecurity rows now include separately sourced historical baselines; the biopharmaceutical row includes broader prescription-medicine market context. Each has its own year, geography, definition and source. These additions do not validate the original estimates or make different market definitions comparable. Select a row to inspect the evidence and its limits. Categories and maturity labels are editorial classifications.\nUnderstand the industry behind the numbers Explore 6 scoped value-chain maps, business models, risks and source-backed evidence anchors. Coverage is a subset of the dataset.\nAI Software \u0026amp; ServicesSemiconductorsBatteries \u0026amp; Energy StorageCybersecurityRoboticsBiopharmaceuticals Browse all industry breakdowns → Download CSV Read the report Research methodology File updated September 2026. Added source records were reviewed on September 30, 2026; this is separate from the measurement year and publication date. Original numeric market sizes are in USD billions where available. Filters and scenarios still use the original indicative estimates, not the separately sourced baselines. The CAGR filter uses a range midpoint only when both bounds exist; a one-sided value is treated as a lower bound.\nThe full screening list is available below. Numeric coverage appears when the interactive controls load.\nSearch Category All AI Foundation Digitization Energy Hard Tech Bio \u0026 Health Manufacturing Maturity All Early Growth Mature Minimum indicative CAGR Any 10%+ 20%+ 25%+ Full industry list\nSearch, sorting and scenarios require JavaScript; the list remains readable below.\nProvisional industry screening list. Use an industry's Select button to inspect its data and prepare a scenario. # Industry Technology Market ~2025 CAGR Category 1 AI Software \u0026amp; Services\nOriginal ~2025 estimates unverifiedView industry breakdown\nRead report Large models / Agentic AI, Generative AI ~$230–390B 25%–35% AI Foundation 2 Semiconductors\nSourced baseline · 2025\nOriginal ~2025 estimates unverifiedView industry breakdown\nRead report\nView industry mind map\nView share chart Advanced nodes (2nm/3nm), HBM, AI accelerators ~$630–775B 10%–18% AI Foundation 3 Cloud Services\nOriginal ~2025 estimates unverified Hyperscale data centers, AI training/inference cloud ~$220B baseline 12%–17% AI Foundation 4 Electric Vehicles (EVs)\nOriginal ~2025 estimates unverified Solid-state batteries, 800V platforms, domain controllers ~$900–990B 10%–15% Energy 5 Batteries \u0026amp; Energy Storage\nOriginal ~2025 estimates unverifiedView industry breakdown Solid-state, sodium-ion, flow batteries ~$98B (batteries) 12%–14% Energy 6 Shared Autonomous Vehicles\nOriginal ~2025 estimates unverified L4/L5 autonomy, Robotaxi, vehicle-road coordination Early stage High (4%–20%) Hard Tech 7 Space Economy\nOriginal ~2025 estimates unverified Reusable rockets, LEO constellations, satellite internet ~$550–626B 7%–10% Hard Tech 8 Cybersecurity\nSourced baseline · 2024\nOriginal ~2025 estimates unverifiedView industry breakdown\nRead report\nView industry mind map\nView share chart Zero Trust, AI security, quantum-safe cryptography ~$160–240B 8%–20% Digitization 9 Robotics (Industrial \u0026#43; Service)\nOriginal ~2025 estimates unverifiedView industry breakdown Humanoid robots, collaborative robots, embodied AI ~$50–90B 14%–20% Hard Tech 10 E-commerce\nOriginal ~2025 estimates unverified Social commerce, instant retail, cross-border platforms Mature large-scale 7%–9% Digitization 11 Digital Advertising\nOriginal ~2025 estimates unverified Programmatic, AI targeting, retail media networks ~$320–580B range 8%–20% Digitization 12 Streaming Video\nOriginal ~2025 estimates unverified Cloud gaming, interactive streaming, AI content Mature growth Mid-high Digitization 13 Video Games\nOriginal ~2025 estimates unverified Cloud gaming, metaverse games, AI NPCs Hundreds of $B Mid-high Digitization 14 Modular Construction\nOriginal ~2025 estimates unverified Prefabricated modules, digital construction Early expansion High Manufacturing 15 Future Air Mobility\nOriginal ~2025 estimates unverified eVTOL, urban air mobility Early commercialization 10%–20% Hard Tech 16 Obesity \u0026amp; Related Drugs\nOriginal ~2025 estimates unverified GLP-1 agonists, innovative weight-loss drugs ~$24–100B range 9%–35% Bio \u0026amp; Health 17 Nuclear Fission Power\nOriginal ~2025 estimates unverified Small modular reactors (SMR), advanced reactors ~$18B baseline 7%–13% Energy 18 Industrial \u0026amp; Consumer Biotech\nOriginal ~2025 estimates unverified Synthetic biology, enzyme engineering, bio-based materials Early to growth High Bio \u0026amp; Health 19 Integrated Circuits (China focus)\nOriginal ~2025 estimates unverified Domestic advanced packaging, Chiplet, automotive chips Rapid China growth 20%\u0026#43; AI Foundation 20 Aerospace (China focus)\nOriginal ~2025 estimates unverified Domestic large aircraft, satellites, LEO constellations Rapid expansion High Hard Tech 21 Biopharmaceuticals\nSourced context · 2025\nOriginal ~2025 estimates unverifiedView industry breakdown\nRead report\nView industry mind map\nView share chart Cell \u0026amp; gene therapy, antibody drugs, AI drug discovery Mature high growth 15%–25% Bio \u0026amp; Health 22 Low-Altitude Economy\nOriginal ~2025 estimates unverified Drone logistics, eVTOL, low-altitude air traffic management Early boom in China 30%\u0026#43; Hard Tech 23 New-Type Energy Storage\nOriginal ~2025 estimates unverified Electrochemical, compressed air, flywheel storage Rapid China growth 25%\u0026#43; Energy 24 Intelligent Robots\nOriginal ~2025 estimates unverified Humanoid, industrial intelligent, service robots Rapid expansion High Hard Tech 25 Quantum Technology\nOriginal ~2025 estimates unverified Quantum computing, communication, sensing Early R\u0026amp;D / commercialization High AI Foundation 26 Bio-manufacturing\nOriginal ~2025 estimates unverified Synthetic bio-manufacturing, cell factories Early High Bio \u0026amp; Health 27 Green Hydrogen\nOriginal ~2025 estimates unverified Water electrolysis, hydrogen trucks, green hydrogen chemicals Early commercialization High Energy 28 Nuclear Fusion Energy\nOriginal ~2025 estimates unverified Tokamak, inertial confinement, demo reactors R\u0026amp;D stage Extremely high (long-term) Energy 29 Brain-Computer Interface\nOriginal ~2025 estimates unverified Invasive / non-invasive BCI, neuromodulation Early clinical / commercial High Bio \u0026amp; Health 30 Embodied AI\nOriginal ~2025 estimates unverified Multimodal models \u0026#43; robot bodies, physical-world interaction Early breakout Extremely high Hard Tech 31 6G Communications\nOriginal ~2025 estimates unverified Terahertz, integrated space-air-ground networks R\u0026amp;D / standards High (long-term) Digitization 32 Data Centers \u0026amp; Computing Power\nOriginal ~2025 estimates unverified Liquid cooling, AI servers, intelligent computing centers High-speed expansion High AI Foundation 33 Renewable Energy (Solar \u0026#43; Wind)\nOriginal ~2025 estimates unverified Perovskite cells, large offshore turbines, IBC modules Large-scale mature Mid-high Energy 34 Hydrogen Full Value Chain\nOriginal ~2025 estimates unverified Green hydrogen production, storage/transport, fuel cells Early to growth High Energy 35 CCUS\nOriginal ~2025 estimates unverified Carbon capture, utilization, geological storage Early High Energy 36 Smart Manufacturing / Industry 4.0\nOriginal ~2025 estimates unverified Digital twins, industrial AI, flexible production lines ~$550–650B (2025) ~19% Manufacturing 37 Fintech\nOriginal ~2025 estimates unverified Digital payments, embedded finance, blockchain settlement Mature high growth Mid-high Digitization 38 Silver Economy / Smart Elderly Care\nOriginal ~2025 estimates unverified Age-friendly products, telemedicine, rehab robots Rapid China growth 15%–20% Bio \u0026amp; Health 39 Gene \u0026amp; Cell Therapy\nOriginal ~2025 estimates unverified CRISPR gene editing, CAR-T, stem-cell therapy Rapid commercialization 20%\u0026#43; Bio \u0026amp; Health 40 Synthetic Biology\nOriginal ~2025 estimates unverified Genetic circuit design, microbial cell factories Early expansion High Bio \u0026amp; Health 41 Advanced Materials\nOriginal ~2025 estimates unverified Graphene, superconducting materials, next-gen semiconductors Early to growth High Manufacturing 42 Satellite Internet\nOriginal ~2025 estimates unverified LEO constellations, high-throughput satellite communications Rapid deployment High Hard Tech 43 Autonomous Driving \u0026amp; Connected Vehicles\nOriginal ~2025 estimates unverified LiDAR, domain controllers, vehicle-road-cloud integration Growth stage High Hard Tech 44 Medical AI \u0026amp; Digital Health\nOriginal ~2025 estimates unverified AI diagnosis, telemedicine, wearable monitoring Rapid expansion High Bio \u0026amp; Health 45 Green Buildings \u0026amp; Energy Efficiency\nOriginal ~2025 estimates unverified BIPV, smart building systems BIPV ~$17B (2024) ~20% Manufacturing 46 Logistics Automation \u0026amp; Drone Delivery\nOriginal ~2025 estimates unverified Warehouse robots, last-mile delivery drones Rapid expansion High Manufacturing 47 Digital Twin \u0026amp; Industrial Metaverse\nOriginal ~2025 estimates unverifiedRead report\nView industry mind map\nView share chart Factory digital twins, virtual-physical simulation Early to growth High Manufacturing 48 Privacy Computing \u0026amp; Data Elements\nOriginal ~2025 estimates unverified Federated learning, multi-party computation, data trading Early expansion 20%\u0026#43; Digitization 49 Advanced Connectivity\nOriginal ~2025 estimates unverified 5G-Advanced, 6G, space-air-ground-sea integration 5G mature, 6G R\u0026amp;D High Digitization 50 Sustainable Food \u0026amp; Alternative Proteins\nOriginal ~2025 estimates unverified Cultured meat, plant-based proteins, precision fermentation Early commercialization High Bio \u0026amp; Health Scenario calculator Select an industry to inspect its fields and load indicative numbers, or enter your own assumptions.\nStarting value (USD billion) Annual growth rate (%) Years Enter values to calculate\nSelect an industry to inspect its scope and source limitations. The result is a mathematical scenario, not a market forecast.\nSource review notes The original screening estimates are unchanged. The following records are shown separately so their boundaries remain visible:\nIndustry row Sourced figure What it establishes Semiconductors USD 795.6 billion in 2025 Annual semiconductor product sales in WSTS\u0026rsquo;s March 6, 2026 release. This is above the original screening range and does not establish its CAGR or forecast. Cybersecurity USD 193.408 billion in 2024 Gartner\u0026rsquo;s historical estimate of worldwide information security end-user spending, from its July 29, 2025 release. The source\u0026rsquo;s later-year numbers are forecasts. End-user spending is not necessarily vendor revenue. Biopharmaceuticals USD 1,667.671 billion in 2025 Broader global prescription-medicine sales at ex-manufacturer prices, reported by EFPIA / IQVIA MIDAS, Key Data 2026, page 14. Includes medicines beyond biological products and does not measure the biopharmaceutical-only market. The downloadable CSV includes separate baseline_* columns for these records. Empty original-source fields still mean the original estimate has not been verified. A sourced baseline or broader-market context is not a forecast validation.\n","date":"2026-09-21T00:00:00Z","permalink":"/industries/","title":"Industries"},{"content":"A provisional 50-industry screening list inspired by global arenas and China policy priorities. Numeric estimates await row-level source verification. Interactive database: filter the 50 industries and run a CAGR calculator on the Industries page.\nOver the past two decades, a small set of industries captured an outsized share of global growth and market-value creation. McKinsey Global Institute calls them “arenas of competition”: sectors that combine high growth with intense competitive dynamism.\nMcKinsey Global Institute\u0026rsquo;s 2024 analysis identifies 18 potential future arenas and models $29–48 trillion in combined 2040 revenue across them. Those figures describe McKinsey\u0026rsquo;s collective scenario, not the 50 rows below.\nAt the same time, China\u0026rsquo;s NDRC described six emerging pillar industries and six future industries in March 2026. It estimated that output related to the six emerging pillars alone could exceed RMB 10 trillion by 2030; this is not a forecast for all 12 categories together.\nThis report brings those signals together into a provisional screening list of 50 industries—with representative technologies, indicative figures where available, and commercial or policy notes.\nEvidence status: All original screening ranges, growth rates and projections below remain unverified. The Industries page and downloadable CSV now carry separate, source-checked historical baselines for semiconductors and cybersecurity, plus broader prescription-medicine context for biopharmaceuticals. Each added record has its own year, scope and source; it does not verify the original figures in this report. Empty original-source fields mean unverified, not zero or not applicable. Selected contextual reading links also do not substantiate market size or CAGR. See the methodology for the verification standard.\n1. McKinsey’s Future Arenas McKinsey groups high-growth, high-dynamism industries into themes such as:\nAI foundation — AI software and services, semiconductors, cloud Digitization — e-commerce, digital advertising, cybersecurity, streaming, games Electrification — EVs, batteries, nuclear fission Hard tech — robotics, space, shared autonomous vehicles, future air mobility, modular construction New bio-frontiers — obesity drugs, industrial and consumer biotech These arenas are defined by steep S-curves, shifting market shares, and outsized contribution to GDP growth.\n2. China’s Policy Pillars China’s NDRC has prioritized:\nSix emerging pillars: integrated circuits, aerospace, biopharmaceuticals, low-altitude economy, new-type energy storage, intelligent robots Six future industries: quantum technology, bio-manufacturing, green hydrogen and fusion, brain-computer interface, embodied AI, 6G Together they form a clear directional map for the next decade of industrial policy and capital allocation.\n3. How the List of 50 Was Built The 50 entries below expand the McKinsey arenas and Chinese pillars into a broader working list, adding adjacent high-growth segments (data centers, fintech, silver economy, advanced materials, etc.) so readers can scan both global and China-focused opportunity spaces in one place.\nRank Industry Representative Technology Market Size (~2025) Projected Size Est. CAGR Key Notes 1 AI Software \u0026amp; Services Large models / Agentic AI, Generative AI ~$230–390B $1.5–2.4T (2030–32) 25%–35% Enterprise AI spend surging; Gartner ~$1.5T AI spend in 2025 2 Semiconductors Advanced nodes (2nm/3nm), HBM, AI accelerators ~$630–775B $1.5–3.2T (2030) 10%–18% AI servers driving growth; BofA sees $3.2T TAM by 2030 3 Cloud Services Hyperscale data centers, AI training/inference cloud ~$220B baseline $1.6–3.4T (2040) 12%–17% Infrastructure demand accelerating with AI 4 Electric Vehicles (EVs) Solid-state batteries, 800V platforms, domain controllers ~$900–990B $1.1–3.7T (2030–35) 10%–15% China leads volume; IEA tracks global adoption 5 Batteries \u0026amp; Energy Storage Solid-state, sodium-ion, flow batteries ~$98B (batteries) $810B–1.1T (2040) 12%–14% China new-type storage is a pillar industry 6 Shared Autonomous Vehicles L4/L5 autonomy, Robotaxi, vehicle-road coordination Early stage $610B–2.3T (2040) High (4%–20%) Waymo, Cruise, Tesla advancing commercialization 7 Space Economy Reusable rockets, LEO constellations, satellite internet ~$550–626B $1–1.8T (2035–40) 7%–10% Launch cost decline; commercial services dominant 8 Cybersecurity Zero Trust, AI security, quantum-safe cryptography ~$160–240B $590B–1.2T (2040) 8%–20% AI threats + quantum computing driving demand 9 Robotics (Industrial + Service) Humanoid robots, collaborative robots, embodied AI ~$50–90B $110–205B (2030) 14%–20% China intelligent robots is a pillar industry 10 E-commerce Social commerce, instant retail, cross-border platforms Mature large-scale $14–20T (2040) 7%–9% Emerging markets + new categories 11 Digital Advertising Programmatic, AI targeting, retail media networks ~$320–580B range $2.1–2.9T (2040) 8%–20% AI optimization + privacy shifts 12 Streaming Video Cloud gaming, interactive streaming, AI content Mature growth Multi-hundred $B Mid-high Content + advertising models converging 13 Video Games Cloud gaming, metaverse games, AI NPCs Hundreds of $B Continued strong growth Mid-high ~40% of global population may be gamers by 2030 14 Modular Construction Prefabricated modules, digital construction Early expansion High potential High Labor shortages + efficiency drivers 15 Future Air Mobility eVTOL, urban air mobility Early commercialization $75–340B (2040) 10%–20% China low-altitude economy is a pillar industry 16 Obesity \u0026amp; Related Drugs GLP-1 agonists, innovative weight-loss drugs ~$24–100B range $120–280B (2040) 9%–35% Semaglutide-class drugs driving rapid growth 17 Nuclear Fission Power Small modular reactors (SMR), advanced reactors ~$18B baseline $65–150B (2040) 7%–13% Energy security + low-carbon dual drivers 18 Industrial \u0026amp; Consumer Biotech Synthetic biology, enzyme engineering, bio-based materials Early to growth High potential High Non-medical biotech; sustainable materials 19 Integrated Circuits (China focus) Domestic advanced packaging, Chiplet, automotive chips Rapid China growth Part of 6 pillars \u0026gt; RMB 10T by 2030 20%+ NDRC emerging pillar industry 20 Aerospace (China focus) Domestic large aircraft, satellites, LEO constellations Rapid expansion Part of 6 pillars High Large aircraft + commercial space 21 Biopharmaceuticals Cell \u0026amp; gene therapy, antibody drugs, AI drug discovery Mature high growth Continued expansion 15%–25% Aging + innovation; China pillar industry 22 Low-Altitude Economy Drone logistics, eVTOL, low-altitude air traffic management Early boom in China China may exceed RMB 3.5T by 2035 30%+ Strong policy support; NDRC pillar 23 New-Type Energy Storage Electrochemical, compressed air, flywheel storage Rapid China growth Part of 6 pillars 25%+ Mandatory storage pairing with wind/solar 24 Intelligent Robots Humanoid, industrial intelligent, service robots Rapid expansion Part of 6 pillars High NDRC priority; embodied AI synergy 25 Quantum Technology Quantum computing, communication, sensing Early R\u0026amp;D / commercialization ~$85B by 2035 possible High China future industry; technology breakthrough phase 26 Bio-manufacturing Synthetic bio-manufacturing, cell factories Early High-potential future industry High China future industry; chemical substitution 27 Green Hydrogen Water electrolysis, hydrogen trucks, green hydrogen chemicals Early commercialization High growth High China future industry; dual-carbon goals 28 Nuclear Fusion Energy Tokamak, inertial confinement, demo reactors R\u0026amp;D stage Long-term huge potential Extremely high (long-term) China future industry 29 Brain-Computer Interface Invasive / non-invasive BCI, neuromodulation Early clinical / commercial High-potential future High China future industry 30 Embodied AI Multimodal models + robot bodies, physical-world interaction Early breakout High-potential future Extremely high China future industry; humanoid synergy 31 6G Communications Terahertz, integrated space-air-ground networks R\u0026amp;D / standards Commercial after 2030 High (long-term) China future industry; IoT foundation 32 Data Centers \u0026amp; Computing Power Liquid cooling, AI servers, intelligent computing centers High-speed expansion Continued high growth High AI training/inference demand; China computing power network 33 Renewable Energy (Solar + Wind) Perovskite cells, large offshore turbines, IBC modules Large-scale mature Multi-trillion $ install base Mid-high Core of global energy transition 34 Hydrogen Full Value Chain Green hydrogen production, storage/transport, fuel cells Early to growth High growth High Key path for trucking, steel, chemicals decarbonization 35 CCUS Carbon capture, utilization, geological storage Early High growth potential High Industrial decarbonization \u0026amp; negative emissions 36 Smart Manufacturing / Industry 4.0 Digital twins, industrial AI, flexible production lines ~$550–650B (2025) ~$1.6T (2030) ~19% Core of China’s manufacturing upgrade 37 Fintech Digital payments, embedded finance, blockchain settlement Mature high growth Continued expansion Mid-high Global digital payments \u0026amp; financial inclusion 38 Silver Economy / Smart Elderly Care Age-friendly products, telemedicine, rehab robots Rapid China growth Trillion-RMB market 15%–20% Aging population rigid demand 39 Gene \u0026amp; Cell Therapy CRISPR gene editing, CAR-T, stem-cell therapy Rapid commercialization High growth 20%+ Precision medicine breakthroughs 40 Synthetic Biology Genetic circuit design, microbial cell factories Early expansion High potential High Materials, food, energy applications 41 Advanced Materials Graphene, superconducting materials, next-gen semiconductors Early to growth China ~RMB 1.2T by 2030 possible High Future materials priority direction 42 Satellite Internet LEO constellations, high-throughput satellite communications Rapid deployment High growth High Global coverage \u0026amp; emergency communications 43 Autonomous Driving \u0026amp; Connected Vehicles LiDAR, domain controllers, vehicle-road-cloud integration Growth stage High growth High Synergy with Robotaxi \u0026amp; low-altitude economy 44 Medical AI \u0026amp; Digital Health AI diagnosis, telemedicine, wearable monitoring Rapid expansion High growth High Aging + efficiency dual drivers 45 Green Buildings \u0026amp; Energy Efficiency BIPV, smart building systems BIPV ~$17B (2024) BIPV ~$42B (2029) etc. ~20% Dual-carbon \u0026amp; building energy control 46 Logistics Automation \u0026amp; Drone Delivery Warehouse robots, last-mile delivery drones Rapid expansion High growth High E-commerce \u0026amp; instant delivery drivers 47 Digital Twin \u0026amp; Industrial Metaverse Factory digital twins, virtual-physical simulation Early to growth High potential High Smart manufacturing \u0026amp; training applications 48 Privacy Computing \u0026amp; Data Elements Federated learning, multi-party computation, data trading Early expansion High growth 20%+ Data element marketization \u0026amp; compliance 49 Advanced Connectivity 5G-Advanced, 6G, space-air-ground-sea integration 5G mature, 6G R\u0026amp;D Long-term high growth High IoT \u0026amp; low-latency application foundation 50 Sustainable Food \u0026amp; Alternative Proteins Cultured meat, plant-based proteins, precision fermentation Early commercialization High potential High Population + climate pressure food innovation 4. What the List Highlights AI foundation cluster (AI software, semiconductors, cloud) is a useful set of related research questions about demand, infrastructure, and value capture. China\u0026rsquo;s policy priorities identify areas for further investigation; the list does not establish their investment outcomes. Electrification and hard tech (EVs, batteries, robotics, space, future air mobility) require separate market definitions before their estimates can be compared. Market sizes and CAGRs vary significantly across sources due to differing definitions; always cross-check the latest primary reports. 5. Background Sources (Not Row-Level Citations) McKinsey Global Institute, Growth industries and the next big arenas of competition (2024): future-arena framework and collective scenario. NDRC, economic press conference (March 2026, Chinese): names of the six emerging pillars and six future industries; aggregate output scenario for the pillars. IEA, Global EV Outlook 2025: EV adoption context; not a citation for the EV market-size row. OECD, The Space Economy in Figures (2023): definitions and measurement context; not a citation for the space market-size row. NIST SP 800-207, Zero Trust Architecture (2020): provided primary context for the cybersecurity category; not a citation for row 8\u0026rsquo;s market-size or CAGR. FDA, Frances Oldham Kelsey and thalidomide: provided primary context for pharmaceutical regulation; not a citation for row 21\u0026rsquo;s growth range. List file updated September 2026; the date does not establish when the figures were sourced or checked. Verify every relevant estimate with its original report before use.\n","date":"2026-09-19T00:00:00Z","image":"/post/50-high-potential-industries/cover.jpg","permalink":"/post/50-high-potential-industries/","title":"50 High-Potential Industries (2025–2040): Technologies, Markets, and Growth"},{"content":"A source-scoped look at U.S. technology job-cut announcements, AI-attributed reasons, and the limits of causal interpretation. Part 1: Story U.S. technology employers announced substantial job cuts in 2026 even as many companies invested in AI infrastructure. In Challenger, Gray \u0026amp; Christmas\u0026rsquo;s August report, technology accounted for 155,126 announced U.S. cuts from January through August, while employers across all industries cited AI in 116,175 announced cuts over that period. The two figures have different denominators and must not be added or treated as confirmed completed layoffs.\nAI was the most-cited monthly reason for five months beginning in March, but restructuring led in August; the cited reason is what an employer reports, not an independent finding that AI automated every affected role.\nAt the same time, employers announced hiring plans, including in technology. The public narrative split into two camps: one said AI was replacing white-collar work at scale; the other emphasized restructuring and shifts in investment. Neither claim can be established solely from the stated reasons in layoff announcements.\nBoth stories contain partial truths. The useful question is not whether AI “causes” layoffs in a single slogan, but how AI risk narratives, industry restructuring, and labor markets are interacting in practice.\nPart 2: What “AI Risk” Means in This Context 1. Capability Risk vs. Labor Risk Public debate often collapses several different risks into one phrase:\nCapability / safety risk — models that act beyond intended control, amplify errors, or enable large-scale misuse. Economic / labor risk — displacement of tasks and roles, slower hiring for entry-level knowledge work, wage pressure in AI-exposed occupations. Business / capital risk — firms betting so heavily on AI infrastructure that they must cut elsewhere to fund the bet. The relative contribution of these mechanisms cannot be read directly from public layoff totals. Safety debates continue in parallel; labor attribution requires company- and task-level evidence.\n2. The Productivity Promise and the Evidence Gap Some executives anticipate productivity gains from AI, but output must be measured against outcomes rather than only activity, such as code changes or messages handled. The earlier specific account of Meta\u0026rsquo;s internal metrics and proposed team cuts could not be confirmed in a primary publication and has been removed.\nThat pattern matters: announcing AI-driven efficiency is easier than proving it. Until measurement improves, “AI” can function both as a genuine operating shift and as a narrative cover for cost control.\n3. Who Feels the Pressure First Stanford Digital Economy Lab\u0026rsquo;s research points to uneven impact, including weaker employment and hiring for younger workers in AI-exposed occupations; this is an observed association under its methods, not proof that AI alone caused every difference:\nEntry-level and young workers in AI-exposed white-collar roles show weaker employment growth than peers in less-exposed jobs. Pay and job quality may change even when employment does not; the direction and size depend on the occupation and study design. Experienced workers in the same fields often look more resilient so far—suggesting augmentation and selective hiring rather than blanket replacement. Goldman Sachs Research has estimated that on the order of 6%–7% of U.S. workers could be “displaced” over a decade in the sense of needing new employment because of automation—material, but not an overnight wipeout of half the white-collar workforce.\nPart 3: Industry Structure of the Layoff Wave 1. Scale of the Cuts Through August 2026, Challenger counted 155,126 U.S. technology-sector announced cuts, up 52% from the corresponding first eight months of 2025. Across all U.S. industries it counted 116,175 announced cuts citing AI, about 22% of total announced cuts. The technology-sector and AI-reason series overlap but are not interchangeable; these are announcements, not verified individual separations.\nThe previously listed company-specific counts mixed reporting periods, headcount changes and layoff announcements without links to comparable primary disclosures. They are omitted until each example can be checked against the relevant company\u0026rsquo;s dated statement or filing.\n2. Three Forces Running at Once Restructuring and changing demand. Hiring and revenue trajectories differ by firm; a particular reduction does not require an AI explanation.\nCapital reallocation to AI infrastructure. Data centers, chips and power compete for investment budgets, but a direct budget-for-jobs substitution should be documented at the company level.\nTask-level automation. Coding assistants, customer-support bots and internal tools can change staffing needs; displacement and complementary hiring vary across tasks and firms.\n3. Hiring and Firing in the Same Industry Challenger recorded 119,825 announced U.S. hiring plans across all sectors through August 2026, including 19,751 in technology. These are planned hires, not verified jobs filled, and they cannot be subtracted mechanically from announced cuts. The broader economy\u0026rsquo;s net employment effect from AI remains contested.\nPart 4: Challenges and Open Questions 1. Attribution Problem When a company cites AI in a layoff memo, outsiders cannot easily separate:\nroles genuinely automated, roles cut to fund GPUs and data centers, roles that would have been cut in any efficiency drive. Over-attributing everything to AI inflates fear; under-attributing it ignores a real shift in bargaining power and skill demand.\n2. Entry-Level Pipeline Risk If firms use AI to skip junior headcount, they also thin the pipeline that produces senior talent. That is a structural risk for the industry itself: fewer apprenticeships in code, analysis, and operations today mean a thinner expert layer in five to ten years.\n3. Public Expectation vs. Measured Outcome Expectations about AI\u0026rsquo;s job impact and measured employment are different types of evidence. Stanford\u0026rsquo;s research reports no widespread economy-wide displacement in its sample while identifying more specific early-career risks. The gap between expectation and measured outcome warrants continued investigation.\n4. What to Watch Next Whether AI-attributed cuts stay elevated or fade as the “easy” post-pandemic layers are already gone. Whether productivity metrics (revenue per employee, feature velocity, support resolution) catch up with headcount rhetoric. Whether entry-level hiring in AI-exposed fields stabilizes or keeps lagging. How regulation and public opinion respond if displacement concentrates in visible white-collar cohorts. Part 5: Key Takeaways U.S. technology-sector job-cut announcements increased through August 2026 year-on-year in Challenger\u0026rsquo;s series; AI is frequently named across industries, but its exact causal share cannot be derived from announcement reasons. AI risk in the labor market is currently more about task reallocation, slower junior hiring, and narrative cover for restructuring than about a single switch that deletes half of white-collar work overnight. Hiring and cutting can coexist: planned technology hires appear alongside announced cuts, with no guarantee that plans became actual jobs. The hardest problems ahead are measurement (what did AI actually replace?), the entry-level pipeline, and the political gap between public fear and still-mixed macroeconomic data. Sources and scope Challenger, Gray \u0026amp; Christmas: August 2026 job-cuts report — primary source for announced U.S. job cuts and hiring plans through August, including industry and stated-reason breakdowns. It does not establish the causal impact of AI or actual completed layoffs. Stanford Digital Economy Lab: Canaries in the Coal Mine? — research on AI-exposed occupations and early-career workers; the paper explicitly says it finds no widespread economy-wide displacement. Goldman Sachs Research: AI and the U.S. labor market — the 6%–7% over roughly a decade figure is a scenario for workers potentially needing new employment, not a count of current layoffs. Reviewed September 2026. Announcement series, survey research, and forecast scenarios answer different questions and should not be combined into a single count.\n","date":"2026-09-19T00:00:00Z","permalink":"/post/ai-risks-layoffs-industry-turbulence/","title":"AI Risks, Industry Turbulence, and the Layoff Wave: What the Numbers Actually Show"},{"content":"A deep dive into the 150-year evolution of the pharmaceutical industry, core market structures, and future challenges. Part 1: Story In 1960, Frances Oldham Kelsey joined the U.S. Food and Drug Administration (FDA) as a medical reviewer.\nOne of her first assignments was an application to market thalidomide in the United States. The sedative was already used in other countries, including by some pregnant patients.\nKelsey and colleagues found the evidence of safety inadequate and asked the applicant for more information, including about reported nerve damage. The FDA did not approve the application.\nThe drug was subsequently linked to severe birth defects in thousands of children internationally. The United States did not approve its commercial sale, although some U.S. patients had received it through investigational distribution.\nThe episode helped build support for the 1962 Kefauver–Harris Drug Amendments. U.S. law had already required evidence of safety for new drugs; the amendments added a requirement to demonstrate effectiveness and strengthened oversight of clinical investigations.\nThe law is a milestone in U.S. drug regulation, but it did not alone determine global R\u0026amp;D costs or create the industry\u0026rsquo;s market structure. Today\u0026rsquo;s long development cycles reflect scientific uncertainty, testing, manufacturing, regulation, and commercialization together.\nPart 2: Industry History The modern pharmaceutical industry developed across more than a century of chemical synthesis, biological research, large-scale manufacturing, clinical testing, and regulation. The timeline below is a simplified guide, not a claim that innovation or manufacturing was confined to the United States and Europe.\n1. Late 19th Century - 1930s: From Dye Workshops to Chemical Synthesis (Disorderly Emergence) Plant Extraction and Accidental Discovery.\nIn 1899, the German dye giant Bayer launched Aspirin.\nModern pharmaceuticals were born in the chemical and textile industries. Scientists, while developing synthetic dyes, discovered that certain coal tar derivatives and azo dyes had bactericidal or analgesic effects. At that time, German dye giants such as IG Farben naturally entered the pharmaceutical industry. At this time, the industry had almost no concept of \u0026ldquo;clinical trials,\u0026rdquo; and drug sales were similar to folk remedies.\n2. 1940s - 1960s: Mass Production of Antibiotics and the Iron Curtain of Regulation (Foundational Period) Bacterial Culture and Large-Scale Random Screening.\nDuring World War II, Pfizer helped scale penicillin production using deep-tank fermentation; in 1962, the U.S. passed the Kefauver–Harris Drug Amendments amid the thalidomide crisis.\nLarge-scale antibiotic production helped develop fermentation and purification capabilities. The 1962 amendments reinforced the evidence required for new-drug approval in the United States; they did not by themselves eliminate smaller manufacturers or establish an oligopoly.\n3. 1970s - 1990s: The Era of Molecular Biology and Blockbuster Drugs (The Golden Age of Profits) Target-Based Drug Discovery.\nThe lipid-lowering drug Lipitor and the antidepressant Prozac were launched.\nWith a better understanding of receptors and enzymes, target-based discovery became more important alongside screening and empirical methods. Medicines for common chronic diseases supported a blockbuster model, although sales and margins differ markedly by product and company.\n4. 2000s - 2010s: The Biomolecular Revolution and Restructuring of Specialization Recombinant DNA technology and monoclonal antibodies (mAbs).\nProducts such as Humira illustrated the commercial potential of biologics. Researchers also debated the long-run productivity of drug R\u0026amp;D, sometimes calling the observed trend \u0026ldquo;Eroom\u0026rsquo;s Law.\u0026rdquo;\nBiotechnology firms such as Genentech helped bring biologics into mainstream development. Many large firms now combine in-house R\u0026amp;D with licensing and acquisitions; smaller firms and large manufacturers can each participate at multiple stages of development.\n5. 2020s to Present: Multimodal, Precision Medicine, and Computational Drug Development Programmable drugs (mRNA, ADC, CGT) and AI-driven computing (AIDD).\nmRNA vaccines were rapidly developed and launched during the COVID-19 crisis; AI structural biology tools such as AlphaFold have achieved widespread penetration; ADCs (antibody-drug conjugates) have become the mainstay of precision oncology.\nComputational methods increasingly complement laboratory and clinical work. In some disease areas, biomarker-guided treatments supplement broad-market products; computational predictions do not replace experimental validation.\nSimilarly, the pharmaceutical industry, like its history, is extremely complex. Let\u0026rsquo;s break it down simply:\nPart 3: Industry Structure INDUSTRY MAP\nThe pharmaceutical value chain Follow the path from research and production support to medicines, distribution and payment.\nFit to view Expand all Collapse Full screen Read the full text outline PharmaceuticalsUpstream · R\u0026amp;D supportInstruments \u0026amp; reagentsThermo Fisher Scientific Danaher Illumina APIs \u0026amp; intermediatesActive ingredients Chemical intermediates Research \u0026amp; manufacturing servicesCROs CDMOs / CMOs Technology platformsDrug-discovery software Gene-editing tools Midstream · MedicinesBig PharmaPfizer Eli Lilly AstraZeneca BiotechnologyBioNTech Moderna Generic medicinesTeva Sandoz Sun Pharma Specialty medicinesSanten Jazz Pharmaceuticals Downstream · AccessDistributionMcKesson Cencora Cardinal Health Care \u0026amp; retail channelsHospitals Retail and specialty pharmacies Payers \u0026amp; reimbursementPublic health systems Insurers and PBMs DEX editorial map based on the accompanying report. Examples are illustrative, not exhaustive or ranked. Companies can operate across several stages; connections show categories, not verified supplier contracts.\nSources: Biopharmaceutical industry breakdown and source notes. Reviewed 2026-09-29.\nThe pharmaceutical industry is a complex and unique industry characterized by high technological barriers, high compliance thresholds, long cycles, high profit margins, and high risks. The entire industry chain can be clearly divided into three core segments: upstream (R\u0026amp;D support), midstream (pharmaceutical companies and product portfolios), and downstream (distribution, channels, and payers).\nI. Upstream: R\u0026amp;D and Production Support (\u0026ldquo;Water Sellers\u0026rdquo; and Infrastructure) The upstream provides pharmaceutical companies with comprehensive services from target discovery and experimental consumables to contract manufacturing. It is a relatively risk-dispersed and cash-flow-stable \u0026ldquo;water-selling\u0026rdquo; segment. It is divided into four complex parts:\nScientific Reagents and High-End Instruments\nCore Functions and Roles: Providing gene sequencers, high-resolution mass spectrometers, culture media, biochips, and laboratory consumables; it is the \u0026ldquo;arsenal\u0026rdquo; of pharmaceutical R\u0026amp;D. Representative companies/institutions: Thermo Fisher Scientific, Danaher, Merck KGaA, Illumina Active Pharmaceutical Ingredients (APIs) \u0026amp; Intermediates\nCore Functions and Roles: Providing active pharmaceutical ingredients (APIs) and key chemical intermediates. Divided into bulk APIs (such as vitamins and antibiotics) and specialty/high-difficulty APIs. Representative companies/institutions: Zhejiang Medicine, Huahai Pharmaceutical, Lonza, Teva (API division) CXO (Contract Research and Development Organization)\nCore Functions and Roles: CRO (Contract Research Organization) assists in preclinical and clinical trials; CDMO/CMO (Contract Development and Manufacturing Organization) assists in process development and commercial mass production. Representative companies/institutions: WuXi AppTec, Pharmaron, Tigermed, Lonza, Catalent AI Computing Power and Cutting-Edge Technology Platform\nCore Functions and Roles: Provides gene editing tools, protein structure prediction, and AI drug molecular screening platform (AIDD). Representative companies/institutions: Schrödinger, Recursion Pharma, Insilico Medicine II. Midstream: Pharmaceutical Companies and Product Matrix (Core Value Creators and Risk Bearers) The midstream is the main body of the pharmaceutical industry, bearing the most significant financial risks in drug development and being the primary beneficiaries of patent monopoly premiums. It can be broken down into four camps based on business models:\nBig Pharma (Multinational Traditional Giants)\nBusiness Model: In-house research + external mergers and acquisitions (M\u0026amp;A) / licensing introduction. Core Barriers: Extensive global clinical trial compliance capabilities, a global commercial sales network, and abundant cash flow. Representative companies: Pfizer, Eli Lilly, Novartis, Merck (MSD), AstraZeneca. Biotech (Innovative Biotechnology Companies)\nBusiness Model: Focuses on specific cutting-edge targets or new technology platforms (such as ADC, mRNA, CAR-T). They typically lack a mature sales force, and after reaching Phase II clinical trials, they often choose to sell their rights to a Big Pharma or be directly acquired. Core Barriers: The R\u0026amp;D efficiency of top scientists and patent protection. Representative companies: BioNTech, Moderna, BeiGene, Sarepta. Generic Pharma (Generic Drug Companies)\nBusiness Model: Involves rapidly following up on original drug patents after they expire (Paragraph IV challenge), achieving large-scale substitution at extremely low cost. Core Barriers: Extremely high production cost control, ability to tackle complex formulations, and rapid supply chain response capabilities. Representative companies: Teva, Sandoz, Viatris, Sun Pharma. Specialty Pharma\nBusiness Model: Avoid the fierce competition of Big Pharma in areas like oncology and cardiovascular diseases, and instead focus on niche markets such as ophthalmology, dermatology, central nervous system (CNS), or rare diseases. Representative companies: Santen (ophthalmology), Jazz Pharmaceuticals (rare diseases/sleep disorders). III. Downstream: Distribution, Channels, and Payers (Commercial Monetization and Value Loop) After drugs are produced, they must be delivered to end users through an efficient supply chain and monetized through a specific payment system.\nPharmaceutical Distributors\nRole: Warehousing, distribution, financing, and channel management between pharmaceutical companies and end users. Business characteristics: Often relies on scale and efficient working-capital turnover; margins differ by contract and reporting basis. Representative companies: Large U.S. distributors include McKesson, Cencora (formerly AmerisourceBergen), and Cardinal Health. Chinese giants: Sinopharm, Shanghai Pharmaceuticals, China Resources Pharmaceutical, and Jointown Pharmaceutical Group. End Channels\nHospital Market (HCOs): Public/private hospitals. The core market for the vast majority of prescription drugs, injectables, and critical care medications. Retail \u0026amp; DTP Pharmacies: Chain pharmacies, DTP (Direct to Patient) high-value prescription pharmacies, and online pharmacy platforms (such as JD Health and Meituan Pharmacy). Payers \u0026amp; Access – The Industry\u0026rsquo;s \u0026ldquo;Life and Death Power\u0026rdquo;\nPublic financing and procurement (varies by country): In China, the National Healthcare Security Administration (NHSA) plays a central role in reimbursement and procurement; the National Medical Products Administration (NMPA) regulates medicines. European payment systems differ substantially by country. Multi-payer model (US market): Commercial insurance companies (UnitedHealth, Anthem, etc.) and PBMs (Pharmacy Benefit Managers, such as CVS Caremark, Express Scripts) hold the power of life and death over whether a drug can be included in the reimbursement list. PBMs, by controlling a massive number of patient payments, force pharmaceutical companies to offer substantial rebates. IV. The Underlying Operating Mechanism of the Industry Chain: \u0026ldquo;Patent Cliff\u0026rdquo; and Value Cycle The flow of funds and value throughout the pharmaceutical industry chain is driven by the lever of \u0026ldquo;patent protection.\u0026rdquo;\nNew drug R\u0026amp;D investment -\u0026gt; regulatory approval and a period of market exclusivity where applicable -\u0026gt; patent or exclusivity expiry -\u0026gt; potential generic or biosimilar competition -\u0026gt; reinvestment in development. Prices and the duration of effective exclusivity vary by product and jurisdiction.\nBiotech companies undertake high-risk breakthroughs in the early stages; CXOs earn fixed service fees; Large firms may use capital and commercial networks to launch medicines across markets during their remaining effective exclusivity; After patent expiration, generic drug companies and distributors quickly take over the market, significantly reducing healthcare costs, while Big Pharma uses its profits to seek the next biotech target, completing the cycle. Of course, high returns often come with high risks. This complex competitive landscape limits drug development efficiency and triggers various commercial struggles and public health crises.\nWhere medicines are sold The regional sales mix below describes the destinations of prescription medicine sales. It does not rank pharmaceutical companies.\nMARKET SHARE · Full year 2025\nPrescription medicine sales by region Global retail and hospital prescription medicine sales at ex-manufacturer prices. Regional sales destinations, not manufacturers\u0026#39; headquarters or company market shares.\nWorldwide, grouped by sales market · Share of prescription medicine sales value (%)\nFull screen View the data table Prescription medicine sales by region · Full year 2025 · Share of prescription medicine sales value (%) Sales regionShare North America (US \u0026amp; Canada)54.7%Europe23.7%China6.5%Japan3.5%Latin America4.2%Africa, other Asia \u0026amp; Australia7.4% Download data (CSV) Europe includes Belarus, Turkey, Russia and Ukraine. Other Asia excludes China and Japan. Values retain the publisher\u0026#39;s one-decimal rounding. This is a geographic sales split, not a vendor ranking or total healthcare spending.\nSource: EFPIA / IQVIA MIDAS — The Pharmaceutical Industry in Figures — Key Data 2026 (2026). Reviewed 2026-09-29.\nPart 4: Industry Issues and Challenges 1. The \u0026ldquo;Eroom\u0026rsquo;s Law\u0026rdquo; of R\u0026amp;D Efficiency R\u0026amp;D productivity is difficult to measure: The term \u0026ldquo;Eroom\u0026rsquo;s Law\u0026rdquo; describes a historical observation that research spending rose faster than some measures of output; its rate depends on the period and method used.\nScientific and clinical uncertainty: Some targets are difficult to validate, trials are expensive, and projects can fail at any stage. A single \u0026ldquo;time to market\u0026rdquo; or cost figure does not represent all medicines. 2. Commercial Incentives and Public Health Needs: Severe Mismatch in Public Health Needs Commercial incentives do not always match public-health needs: Expected revenue influences private investment, but clinical feasibility, public funding, and regulation also shape which therapies advance:\nNew Antibiotic Crisis: Given the small dosage, short duration of use, and ease of developing drug resistance, pharmaceutical companies cannot recoup their R\u0026amp;D costs, leading to the bankruptcy of many biotech companies developing antibiotics. The world is facing the threat of untreatable superbugs. Rare and Tropical Diseases Forgotten: Diseases with extremely small patient populations or very low affordability are unlikely to attract commercial R\u0026amp;D funding. 3. Distorted Intermediary Rent-Seeking and High Drug Prices Pricing and intermediary incentives: U.S. pharmacy benefit managers (PBMs) negotiate formularies and rebates for payers. Whether rebate arrangements raise list prices or out-of-pocket costs depends on plan design and contract terms; it cannot be assumed that intermediaries capture the majority of every price increase.\n4. Geopolitics and Global Supply Chain Vulnerability API and intermediate supply: Some important supply chains are geographically concentrated. Their resilience depends on product-specific manufacturing capacity, sourcing and inventories; a blanket country share or impact estimate requires a defined dataset.\nSource notes and primary materials FDA: Frances Oldham Kelsey and the thalidomide application — supports the opening history; it does not establish an exact modern industry-wide R\u0026amp;D cost. Kefauver–Harris Drug Amendments, Public Law 87-781 (1962) — the original statute, supplied with the research materials; U.S. law, not a global law. Pfizer: company history — company account of penicillin production. Cencora: AmerisourceBergen becomes Cencora — verifies the distributor\u0026rsquo;s present name. The accompanying Pharmaceutical Manufacturing working note is a reading list, not an independently verified market dataset. Its links have been matched to the relevant claims above; unverified figures have been removed rather than attributed to the note.\nView or download the supplied original Kefauver–Harris Drug Amendments (1962)\nOriginal PDF · U.S. Government Publishing Office, 17 pages\nOpen PDF in a new tab Download original PDFPublisher's copy If the preview is unavailable in your browser, use “Open PDF in a new tab” above.\nLinks contained in the Pharmaceutical Manufacturing research note These are the note\u0026rsquo;s reading leads, not independent verification of every claim or an endorsement of paid databases. Two links that pointed through Google searches are shown as direct destinations; duplicate links appear once.\nRegulation and trials:\nDrugs@FDA — direct destination of the note\u0026rsquo;s search link. China\u0026rsquo;s CDE. European Medicines Agency. ClinicalTrials.gov. Chinese Clinical Trial Registry. EU Clinical Trials Register. Scientific and commercial research:\nDrugBank. IUPHAR/BPS Guide to Pharmacology. PubChem. DXY Insight. FiercePharma. FierceBiotech. Endpoints News. BioWorld — direct destination of the note\u0026rsquo;s search link. Historical reading:\nNLM biography of Frances Oldham Kelsey. Bayer\u0026rsquo;s historical article and Bayer homepage. Pfizer company history. ","date":"2026-08-21T00:00:00Z","image":"/post/pharmaceutical-industry/cover.jpg","permalink":"/post/pharmaceutical-industry/","title":"The Pharmaceutical Industry: History, Structure, and Challenges"},{"content":"What DEX publishes, how the research is produced, and how to suggest corrections or collaborations. About DEX DEX is an independent, student-led business research and publishing project. I study business and accounting and use this site to turn long-form research into practical, readable industry intelligence.\nThe goal is not to predict which company or asset will win. It is to build a durable body of work that explains how industries developed, how their value chains and business models work, where credible growth signals exist, and which risks could change the outcome.\nWhat This Site Publishes Industry reports covering history, market structure, value chains, economics, competition, regulation, and risk. Structured datasets and tools that make assumptions visible and let readers inspect or test the numbers. Video explainers that translate the core findings into a faster format. Each format supports the same research system: one research project can become an article, a dataset, a visual explanation, and a video without changing the underlying evidence.\nResearch Methodology 1. Define the question and scope Every report begins with a specific question, a geographic and time scope, and a set of definitions. When different sources use different market definitions, the report treats them as different estimates rather than combining them into false precision.\n2. Prioritize stronger sources Sources are generally considered in this order:\nGovernment agencies, regulators, statistical offices, and company filings. International institutions, peer-reviewed research, and established industry bodies. Reputable research firms, major financial institutions, and well-sourced journalism. Secondary summaries used only as discovery aids or context. Published research should link material claims to specific sources, including the source\u0026rsquo;s date, geography, market definition, and unit. A source list alone does not establish which source supports a particular estimate. The 50-industry screening dataset retains its original market sizes, growth figures and projections as unverified research leads. Separate historical baselines have been source-checked for semiconductors and cybersecurity; broader prescription-medicine context accompanies the biopharmaceutical row. Each added record includes its own year, unit, geography, market definition, publication date and review date. A partial or context status applies only to that separate record: it does not verify the original screening estimate or forecast. Original-source fields remain empty until the corresponding original claim is verified. A contextual reading link is also not evidence for the row\u0026rsquo;s numbers.\n3. Separate evidence from interpretation Reported figures, editorial classifications, scenarios, and DEX interpretations are not the same thing. Tools on this site label calculated outputs as scenarios. Categories and maturity labels are editorial judgments intended to make comparison easier; they are not official classifications.\n4. State uncertainty and limitations Forecasts are sensitive to definitions, base years, exchange rates, regulation, and adoption assumptions. Missing data is shown as missing rather than silently replaced. A dataset update date records when the file was edited, not when each underlying source was published or checked.\n5. Review, update, and correct As reports are revised, key numbers and links should be checked against cited material; where that review has not happened, the limitation is stated alongside the data. Substantive corrections or new data may be reflected in the report\u0026rsquo;s last-updated date. If you find an error, please send the exact page, claim, and supporting source to dex222444@gmail.com.\nUse of AI AI tools may assist with discovery, outlining, translation, formatting, code, and production. They are not treated as evidence. Factual claims should be checked against the cited public sources, and final editorial responsibility remains with DEX.\nIndependence and Disclaimer Unless a page clearly states otherwise, the content is independently produced and no company paid for inclusion. Future sponsored or commissioned work will be labeled. Nothing on this site is investment, legal, accounting, or medical advice.\nContact and Collaboration Corrections, source suggestions, dataset feedback, sponsorship inquiries, and research collaborations are welcome.\nEmail: dex222444@gmail.com Proton Mail: qizhangdong325@proton.me YouTube: @ZeRuiDong GitHub: BAOZ121 Website: thedexs.com Support the Research If this work is useful, you can support it by sharing a report, subscribing on YouTube, starring the website repository, or using GitHub Sponsors. These actions help fund more research and better datasets.\n","date":"2026-08-12T00:00:00Z","permalink":"/about/","title":"About \u0026 Methodology"},{"content":"A deep dive into the 60-year evolution of cybersecurity, upstream-midstream-downstream value chains, the four major market camps, and core industry bottlenecks. Part 1: Story An often-repeated account describes attackers using a connected aquarium sensor as an entry point to a casino network. It illustrates how an overlooked device can expand a network\u0026rsquo;s attack surface.\nThe public retellings do not supply enough independently verifiable information to confirm the casino, date, defenses, or amount of data taken. Treat it as an illustrative anecdote, not a documented case study or a quantitative measure of cyber risk.\nThis story brings us to an important subject—\nWelcome, I\u0026rsquo;m Dex. Welcome to my industry report. Before we dive in, let\u0026rsquo;s take a look at a brief history of the industry.\nPart 2: Industry History 1970s: ARPANET and Creeper Early networked experiments such as Creeper and Reaper are part of the history of self-propagating programs and countermeasures. Assigning a single unqualified \u0026ldquo;first worm\u0026rdquo; or \u0026ldquo;first antivirus\u0026rdquo; to either program obscures differences in definitions and surviving records.\n1980s: Birth of Commercial Antivirus Software Commercial antivirus products emerged in the 1980s, and their precise chronology depends on how a \u0026ldquo;first\u0026rdquo; product is defined. The transition from standalone personal computers to connected business networks expanded the range of threats and defenses.\nKey Turning Point: Mid-1990s During the 1990s, more connected personal computers and business networks created additional opportunities for malicious code, denial-of-service attacks, and intrusions. This is a directional overview rather than a complete chronology of named incidents.\n2000s (2000–2009): Commercial and Organized Cybercrime The 2000s marked a transitional period for cybersecurity threats, shifting from mere pranks to serious, organized, and commercially driven criminal activity. Driven by core threat data and landmark incidents, people began to realize the vulnerabilities inherent in the early digital age during this explosion of cybersecurity incidents:\nFast-spreading worms (2000–2004): Incidents such as ILOVEYOU and SQL Slammer showed how email and software vulnerabilities could cause rapid, widespread disruption. Exact global infection and loss estimates vary by source and method.\nRise of Commercial Cybercrime (Mid-to-Late 2000s): Hacker motivations shifted from technical boasting to economic gain.\nBotnets and data breaches: Compromised computers were increasingly used for spam and fraud, while payment-card incidents highlighted the costs of weak data protection. Incident totals and exposed-record counts require case-specific primary reports. Distributed denial of service: High-profile incidents exposed the operational costs of making online services unavailable; dollar-loss estimates are not directly comparable between incidents.\nEspionage and advanced intrusions: Public disclosures such as Operation Aurora increased attention to persistent, targeted threats; cyber espionage itself predated the incident.\n2010 to Present: Cloud-Native \u0026amp; AI-Driven Era Between 2010 and 2019, the global cybersecurity landscape evolved from simple virus defense to geopolitical cyber warfare, massive data breaches, and ransomware ecosystems (e.g., Stuxnet, Sony Pictures hack, WannaCry).\nSince 2020, the industry has undergone profound transformation characterized by supply chain attacks, open-source vulnerabilities, critical infrastructure ransomware, and AI-driven threats. The industry spans nearly six decades of history.\nPart 3: Industry Value Chain INDUSTRY MAP\nThe cybersecurity value chain Explore the foundations, products and services that connect security suppliers to customers.\nFit to view Expand all Collapse Full screen Read the full text outline CybersecurityUpstream · FoundationsCloud infrastructureAWS Microsoft Azure Alibaba Cloud Core componentsCryptographic libraries Specialized chips Threat intelligenceIndicators and telemetry Threat-data feeds Midstream · ProductsEndpoint \u0026amp; workloadsCrowdStrike Network securityPalo Alto Networks Fortinet Identity \u0026amp; accessOkta CyberArk Security analyticsSplunk / Cisco Downstream · DeliveryIntegration \u0026amp; resaleSystem integrators Value-added resellers Managed securityMSSPs Customer security teams Incident responseMandiant Consulting and response teams DEX editorial map based on the accompanying report. Examples are illustrative, not exhaustive or ranked. Companies can operate across several stages; connections show categories, not verified supplier contracts.\nSources: Cybersecurity industry breakdown and source notes. Reviewed 2026-09-29.\nLet\u0026rsquo;s briefly summarize the structure of the cybersecurity industry.\nUpstream: Foundational Infrastructure \u0026amp; Threat Intelligence The upstream sector serves as the cornerstone of the entire security industry, supplying midstream vendors with computing power, fundamental components, and critical threat intelligence:\nCloud Infrastructure \u0026amp; Computing Power: AWS, Microsoft Azure, and Alibaba Cloud are examples of infrastructure on which security services may run. Foundational Core Components: Deep-tech companies mastering cryptographic algorithm libraries and high-precision processing chips (FPGAs, ASICs). Threat Intelligence Providers: Acting as the \u0026ldquo;radar\u0026rdquo; of the industry. They gather Indicators of Compromise (IOCs) globally and package data feeds to power midstream security engines. Midstream: Core Products \u0026amp; Solutions Midstream vendors directly face hacker attacks and provide defensive tools to clients. Based on modern enterprise IT architecture, midstream is categorized into four major segments:\nEndpoint \u0026amp; Workload Security: Antivirus, endpoint detection and response (EDR), and workload protection address different risks; CrowdStrike is one example of an EDR vendor. Network \u0026amp; Perimeter Security: Firewalls, secure access service edge (SASE), and segmentation coexist; Palo Alto Networks and Fortinet are examples of suppliers. Identity \u0026amp; Access Management (IAM): Identity is an important control alongside devices and networks, not the sole perimeter; Okta and CyberArk are examples of suppliers. Security Operations \u0026amp; Data Analytics: Systems such as Splunk (acquired by Cisco) aggregate and investigate security events. Monitoring and analytics products differ in scope and are not interchangeable. Downstream: Channels \u0026amp; Security Services Because security products are complex, a massive downstream service market has emerged:\nSystem Integrators \u0026amp; VARs: Service providers assisting enterprises with procurement, installation, and basic hardware configuration. Managed Security Service Providers (MSSP): Addressing the global shortage of security engineers by directly managing enterprise security operations 24/7 on a subscription basis. High-End Consulting \u0026amp; Incident Response: Teams like the Big Four or Mandiant providing penetration testing and emergency rescue during ransomware attacks. Summary: Simply put, upstream provides materials and infrastructure; midstream builds weapons and trains troops; downstream handles tactical deployment and command.\nPart 4: Industry Market Landscape MARKET SHARE · Full year 2024\nModern endpoint security revenue share IDC modern endpoint security segment; not the whole cybersecurity market.\nWorldwide · Share of modern endpoint security revenue (%)\nFull screen View the data table Modern endpoint security revenue share · Full year 2024 · Share of modern endpoint security revenue (%) CompanyShare Microsoft28.6%Other71.4% Download data (CSV) IDC estimates reproduced on a vendor\u0026#39;s official website. Other is the residual share of all remaining suppliers. Revenue share does not measure customer counts or product effectiveness. Historical 2024 snapshot.\nSource: IDC, reproduced by Microsoft — Microsoft ranked number one in modern endpoint security market share third year in a row (2025-08-27). Reviewed 2026-09-29.\nThere is no single comparable \u0026ldquo;cybersecurity market\u0026rdquo; figure without specifying geography, year, whether services and cloud infrastructure are included, and the research method. The supplied 2022 Menlo Ventures map identifies product categories and companies; it does not substantiate this article\u0026rsquo;s earlier $250B–$300B size, $500B forecast, CAGR, or vendor-share estimates. Those numbers have been removed pending a traceable dataset.\nThe global market is divided into four major camps:\n1. Cross-Domain Tech Giants Key Players: Microsoft (Defender / Sentinel), Google (Mandiant) Competitive Moat: Leveraging software ecosystems and distribution to integrate security products. 2. Pure-Play Security \u0026ldquo;Big Three\u0026rdquo; Key Players: Palo Alto Networks, CrowdStrike, Fortinet Product focus: Palo Alto Networks sells network and cloud security; CrowdStrike emphasizes endpoint and cloud protection; Fortinet sells network-security appliances and software. These are illustrative positions, not audited share rankings. 3. Traditional IT \u0026amp; Hardware Giants Key Players: Cisco, IBM, Trend Micro Product focus: Enterprise networking and IT software, with acquisitions used to expand security portfolios. 4. Niche Specialists Key Players: Zscaler (Zero Trust / SASE), Cloudflare (Edge Protection), Okta (Identity) Competitive Moat: Dominating specific technical niches to attract top-tier enterprise clients. Two Trends Shifting Market Dynamics Vendor Consolidation: Some buyers prefer fewer integrations and vendors; the outcome depends on their existing architecture and procurement needs. Cloud and AI: Cloud-delivered tools and automated detection are growing areas of investment, while hardware controls still serve important use cases. Part 5: Industry Challenges \u0026amp; Bottlenecks Despite intense competition, the industry faces fundamental challenges:\n1. Asymmetric Warfare Defenders must protect every single endpoint and password, whereas attackers need only find one weak link using AI tools. Defenders remain in a reactive cycle while AI drastically lowers attack costs and sky-rockets defense expenses.\n2. Compliance-Driven \u0026ldquo;Shelfware\u0026rdquo; Many non-critical enterprises buy security tools primarily to pass audits rather than stop hackers, creating a market flooded with \u0026ldquo;shelfware\u0026rdquo; installed for inspection and then ignored.\n3. Tool Fragmentation \u0026amp; Alert Fatigue Large organizations can struggle with overlapping tools and alert volumes. A universal average number of tools or false-positive rate would need a defined sample and measurement method.\nValue Chain Bottlenecks Upstream: Shared software components can create widespread exposure, as CISA\u0026rsquo;s Log4j advisories illustrate. Midstream: Ongoing research, complex integrations and operational resistance can slow adoption of zero-trust approaches. Downstream: Labor-intensive services face staffing and incident-response challenges; margins vary across businesses. This competition appears to be a death spiral with no end in sight; as for how the cybersecurity industry will evolve—whether a super-giant akin to Google will emerge, or if the advent of AI will trigger a commercial tsunami—only time will tell.\nThat concludes my industry report. If you found it interesting, please like the video and subscribe to my channel. I’m Dex—see you next time.\nSource notes and primary materials NIST SP 800-207: Zero Trust Architecture (2020) — original standard supplied with the working materials. It defines an architectural approach, not market size or company share. Menlo Ventures: Cybersecurity Market Map (2022) — the supplied category/company map, a 2022 snapshot rather than a 2026 revenue dataset or endorsement of the named vendors. CISA: Apache Log4j vulnerability advisory — primary security guidance for the software-supply-chain example. The accompanying Network Security document is a research reading list, not primary verification for the anonymous casino account or the removed market figures. The incident specifics remain unverified in public primary records.\nView or download the supplied original NIST SP 800-207: Zero Trust Architecture (2020)\nOriginal PDF · National Institute of Standards and Technology, 59 pages\nOpen PDF in a new tab Download original PDFPublisher's copy If the preview is unavailable in your browser, use “Open PDF in a new tab” above.\nThe Menlo Ventures Cybersecurity Market Map PDF is available directly from its publisher; it is not hosted here because permission to redistribute that copyrighted PDF has not been established.\nLinks contained in the Network Security research note These are the supplied note\u0026rsquo;s research and video links. They have not all been independently verified and should not be read as endorsement or primary evidence. Links without Word hyperlink formatting have also been included.\nIncident and industry background:\nThe Hacker News: aquarium thermometer incident. Entrepreneur: casino thermometer account. Privacy International: aquarium thermometer account. Cyber Magazine: history of cybersecurity. History of Information: first computer virus. Wikipedia: Creeper and Reaper. KMC Controls: Creeper and Reaper. Atari Magazine: virus explainer. Atari Mania: ST Virus Killer. Carifred: UVK. Wikipedia: ESET NOD32. Internet Archive: Malware Museum. Wikipedia: G Data CyberDefense. Wikipedia: security-hacking incidents. Purdue TAP: hackers of the 2000s. Cofense: history of phishing. Wikipedia: computer virus and worm timeline. CISA: Log4j guidance. Wikipedia: Sony Pictures hack. Wikipedia: WannaCry attack. Market and technical references:\nMordor Intelligence: cybersecurity market. Menlo Ventures: market map PDF. Cloudflare: next-generation firewalls. Cybersecurity Ventures. U.S. Securities and Exchange Commission. IBM: a decade of global cyberattacks. CSO: Target breach timeline search. NIST SP 800-207 PDF. Video references from the note (third-party material, not licensed for reuse here):\nVideo 1. Video 2. Video 3. Video 4. Video 5. ","date":"2026-08-11T00:00:00Z","image":"/post/cybersecurity-industry-report/cover.jpg","permalink":"/post/cybersecurity-industry-report/","title":"Cybersecurity: An Industry Map From Network Defenses to Zero Trust"},{"content":"A simplified VR/XR industry map, historical context, and what company filings can and cannot establish. What does it mean for a company to spend heavily on an emerging computing platform? Meta\u0026rsquo;s Reality Labs segment reported an operating loss of $19.193 billion for full-year 2025; that is a segment accounting result, not the net loss of Meta Platforms as a whole. Meta acquired Oculus in 2014 and continues to invest in virtual reality, augmented reality, and related devices.\nVR, as the name suggests, is Virtual Reality.\nHistorical context In 1935, more than 90 years ago, Stanley G. Weinbaum\u0026rsquo;s story Pygmalion\u0026rsquo;s Spectacles imagined an immersive experience through glasses; that literary vision preceded the hardware.\nIn the 1960s, Morton Heilig\u0026rsquo;s Telesphere Mask was an early head-mounted stereoscopic display concept. Its historical significance does not make it mechanically equivalent to a modern tracked headset.\nIn his 1965 paper The Ultimate Display, computer graphics pioneer Ivan Sutherland described an ambitious vision for interactive computer displays. A later tracked head-mounted display was presented in 1968; the 1965 paper is historical context, not evidence for the later device\u0026rsquo;s mechanical design.\nJaron Lanier and VPL Research helped popularize the term \u0026ldquo;virtual reality\u0026rdquo; and develop early commercial systems and data gloves. Claims about a single inventor or first company are more contested than this simplified history can establish.\nOver several decades, VR has experienced cycles of enthusiasm and retrenchment. Meta and Apple have invested in different visions of spatial computing, but whether headsets become a mass-market computing platform remains an open question.\nNow that the story is finished, let\u0026rsquo;s look at today in 2026. What does the development of the VR industry look like? Let\u0026rsquo;s simply dissect this industry.\nIndustry chain INDUSTRY MAP\nThe VR / XR value chain Explore the components, headsets and applications behind immersive computing.\nFit to view Expand all Collapse Full screen Read the full text outline VR / XRUpstream · ComponentsComputeQualcomm XR platforms Apple silicon Displays \u0026amp; opticsMicro-OLED displays Pancake optics Tracking \u0026amp; powerCameras and motion sensors Eye / hand tracking Batteries Midstream · HeadsetsDevice brandsMeta Quest Apple Vision Pro Sony PlayStation VR Pico HTC Manufacturing \u0026amp; supplyGoertek Luxshare Precision Downstream · ExperiencesPlatforms \u0026amp; storesSteamVR Meta Quest Store App Store Consumer usesGaming Social experiences Video and entertainment Enterprise usesSimulation and training Design and digital twins Medical education DEX editorial map based on the accompanying report. Examples are illustrative, not exhaustive or ranked. Companies can operate across several stages; connections show categories, not verified supplier contracts.\nSources: Report sources and original materials. Reviewed 2026-09-29.\nFirst is the upstream of this industry:\nChips: Qualcomm (XR2 series), Apple (M-series + R-series chips).\nDisplay/Optics: Sony (Micro-OLED), BOE, Sunny Optical (Pancake lenses).\nSensors \u0026amp; Batteries: Suppliers of various sensors tracking eye movements and gestures.\nThen let\u0026rsquo;s look at the midstream of this industry (hardware terminal manufacturing): Who makes the headsets we buy?\nBrands: Meta (Quest series, affordable mass adoption route), Apple (flagship Vision Pro, high-end spatial computing route), Sony (PSVR products, focusing on console gaming ecosystem), Pico (under ByteDance), HTC.\nContract manufacturers and component suppliers include Goertek and Luxshare Precision. Their consolidated financial reports disclose company-wide results, but do not by themselves reveal or predict the unit sales of an individual headset brand.\nFinally, the downstream of this industry (content and application scenarios): What are its application scenarios?\nC-end ecosystem: SteamVR, Meta Quest Store, App Store. Core applications are gaming, social (VRChat), and video/film.\nB-end applications: Originally used for Air Force flight training and space navigation, later expanded to other fields, such as medical simulation training, industrial digital twins, automobile design, etc.\nCompetition and adoption MARKET SHARE · Full year 2024\nGlobal AR/VR headset shipment share AR/VR headset market tracked by IDC; consumer and commercial shipments. This is broader than VR alone.\nWorldwide · Share of headset unit shipments (%)\nFull screen View the data table Global AR/VR headset shipment share · Full year 2024 · Share of headset unit shipments (%) CompanyShare Meta74.6%Apple5.2%Sony4.3%ByteDance (Pico)4.1%XREAL3.3%Other8.5% Download data (CSV) Other is calculated as 100% minus the five published shares. Shipments are neither installed base nor retail sell-through or revenue. This historical 2024 snapshot is not a 2026 estimate. Do not mix it with Counterpoint\u0026#39;s narrower VR-only estimate.\nSource: IDC — Growth Expected to Pause for AR/VR Headsets, according to IDC (2025-03-25). Reviewed 2026-09-29.\nXR (Extended Reality) is often used as an umbrella term for VR, AR and MR. Categories overlap; market-share claims require carefully defined devices and time periods.\nMeta sells the Quest series and Apple sells Vision Pro, but a 2026 worldwide market-share comparison requires a specified device category, period, geography and shipment source. Meta\u0026rsquo;s segment reports include multiple products and do not split headset and smart-glasses sales sufficiently to establish the relative revenue of those categories here.\nApple positions Vision Pro as a premium spatial-computing device. Its market share and competitive effects should not be inferred from a list price alone.\nNew platforms and lighter glasses may compete for attention, but future product adoption cannot be inferred from announcements alone.\nComfort, fashion, battery life, and the application ecosystem may affect adoption. An earlier attributed quotation about these factors has been removed because its original publication could not be verified.\nTechnical and commercial constraints In the current industry context, the VR/XR industry is plagued by four major challenges:\nFirst, image clarity and display cost. Visible pixel structure depends on more than a single screen-PPI threshold, and Micro-OLED is one of several display approaches. High-resolution optics and displays can add cost, but without a bill of materials it is not possible to attribute a specific portion of a headset\u0026rsquo;s price to one supplier.\nSecond, vergence-accommodation conflict (VAC). In natural viewing, the eyes converge on an object and focus at its distance; a conventional stereoscopic headset can present a different simulated depth while the display\u0026rsquo;s optical focal plane remains fixed.\nThis mismatch can contribute to visual discomfort, although motion, latency, individual sensitivity and other factors also matter. There is no universal severe-nausea rate for all users after a fixed wearing time; device and study conditions matter.\nThird, the current standalone VR headset is a heavily integrated electronic product. Chips, cooling fans, complex Pancake lens sets, and even batteries are all stacked around the user\u0026rsquo;s eye sockets.\nWeight distribution, thermal design and fit can limit comfortable use. Rendering, heat, battery capacity and form factor require tradeoffs, but wear time and battery life vary by headset and workload. There is no single physical \u0026ldquo;impossible trinity\u0026rdquo; or universal one-hour comfort limit.\nFourth, upstream supply chain and scarce materials.\nSupply chains for chips, displays and optics can be exposed to bottlenecks, but the supplied company reports do not establish a direct causal chain from particular minerals to XR display costs. Meta\u0026rsquo;s $19.193 billion 2025 Reality Labs operating loss is verified in its results; it alone cannot prove why individual suppliers or competitors changed product plans in 2026.\nWhere the VR/XR industry will ultimately go is a question that only time can answer.\nThat\u0026rsquo;s the simplified map for this video. It is a starting point for checking historical sources, company filings and device-specific data—not a forecast of a winner. I\u0026rsquo;m Dex, see you next time.\nSource notes and primary materials Meta: fourth-quarter and full-year 2025 results — source for Reality Labs\u0026rsquo; 2025 segment operating loss, not a headset unit count or company-wide net loss. Ivan Sutherland, The Ultimate Display (1965) — the supplied historical paper about an interactive display vision; not a 2026 product or market forecast. Goertek investor relations: 2024 annual report — company-wide filing supplied for manufacturing context; no brand-specific headset shipment claim is inferred. Luxshare Precision investor relations: 2025 annual report and Q1 2026 report — company-wide financial filings supplied for manufacturing context; neither independently measures the entire VR market. The supplied China Mobile Research Institute VR/AR Product Development Status and Trend report dates from November 2022. An original publisher-hosted URL was not verified, so it is recorded here bibliographically rather than linked to an unlicensed copy or used to justify a 2026 market-share claim. The Goertek and Luxshare reports likewise cannot substitute for a defined current shipment survey.\nSupplied original documents The following documents can be opened from their publishers. They are not copied to this website while redistribution rights remain unverified:\nSutherland, The Ultimate Display (1965): University of Utah PDF. Goertek, 2024 Annual Report: publisher-hosted PDF. Luxshare Precision, 2025 Annual Report (English): PDF on the company\u0026rsquo;s disclosure platform. Luxshare Precision, 2026 Q1 Report (Chinese): PDF on the company\u0026rsquo;s disclosure platform. The supplied 2022 China Mobile Research Institute VR/AR report is not offered here as a download because its publisher-hosted original and redistribution terms have not been confirmed.\n","date":"2026-08-11T00:00:00Z","image":"/post/vr-industry-report-2026/cover.jpg","permalink":"/post/vr-industry-report-2026/","title":"Meta Reality Labs' $19.2B 2025 Loss: A VR/XR Industry Map"},{"content":" ","date":"2026-08-10T00:00:00Z","permalink":"/archives/","title":"Archives"}]