The Fab visualizations
First edition · May 2026

The Fab

A history of the semiconductor industry, from Bell Labs to the AI compute war. 71 chapters, 9 parts. Foreword.

Part I

Cold War Chips

The origins: how the transistor was invented and how military demand turned it into an industry.

  1. 01 From Steel to Silicon Opens with Akio Morita (future Sony founder) struggling with vacuum tubes in postwar Japan. Establishes why the world needed something smaller, faster, more reliable. → Sets up why miniaturization mattered.
  2. 02 The Switch Bell Labs invents the transistor (1947–48). Shockley, Bardeen, and Brattain. → The foundational physics moment that everything else builds on.
  3. 03 Noyce, Kilby, and the Integrated Circuit Parallel invention of the chip at TI (Kilby) and Fairchild (Noyce). → The two founding traditions of the industry — military and commercial.
  4. 04 Liftoff Sputnik, Apollo, and the explosion of military/space demand. → How government money paid the upfront cost of building the industry.
  5. 05 Mortars and Mass Production Photolithography emerges (Jay Lathrop at TI). → The manufacturing process that still defines chipmaking.
  6. 06 "I… WANT… TO… GET… RICH" Noyce pushes Fairchild toward commercial markets. → The birth of Silicon Valley capitalism and the venture model.
Part II

The Circuitry of the American World

How the US exported chips globally and built the first supply chains in Asia.

  1. 07 Soviet Silicon Valley Zelenograd — the USSR's planned chip city — and why it failed. → Why command economies couldn't do chips.
  2. 08 "Copy It" Soviet espionage and reverse-engineering strategy. → Why copying always loses to innovation when iteration speed matters.
  3. 09 The Transistor Salesman Sony, Akio Morita, and Japan's rise. → How Japan absorbed US tech and turned it into consumer products.
  4. 10 "Transistor Girls" Female assembly-line labor in Asia. → The labor model that made Asian manufacturing economically viable.
  5. 11 Precision Strike Vietnam-era guided weapons. → Chips as a military force multiplier — the start of "chips = power."
  6. 12 Supply Chain Statecraft Texas Instruments offshores assembly to Taiwan. → The original sin of offshoring; the first link in today's supply chain.
Part III

Leadership Lost?

The 1980s panic when Japan nearly killed the US chip industry. The most direct historical analogue to today's US-China dynamic.

  1. 13 Intel's Revolutionaries Noyce, Moore, Grove found Intel; the microprocessor is born. → The pivot that defined American computing dominance.
  2. 14 The Pentagon's Offset Strategy DoD's tech-over-numbers doctrine to counter Soviet conventional superiority. → Why the Pentagon bet so heavily on chips — context for Chapter 48.
  3. 15 "That Competition Is Tough" Japanese DRAM makers crush US memory chip producers. → How fast a US lead can evaporate.
  4. 16 "At War with Japan" Trade tensions escalate. → Industrial policy becomes geopolitics.
  5. 17 "Shipping Junk" Collapse of GCA, the leading US lithography company. → How a single chokepoint can be lost — and never recovered.
  6. 18 The Crude Oil of the 1980s US industry seeks government intervention. → Chips reframed as a strategic resource, not just a product.
  7. 19 Death Spiral Existential fears for the US chip industry. → Conditions that produced the SEMATECH consortium and policy response.
  8. 20 The Japan That Can Say No Rising Japanese nationalism and economic confidence. → The peak of Japan's challenge before its decline.
Part IV

America Resurgent

How the US clawed its lead back through Micron, Intel's reinvention, Korea's rise, and Cold War victory.

  1. 21 The Potato Chip King Micron's improbable survival in DRAM. → How one Idaho company saved US memory.
  2. 22 Disrupting Intel Andy Grove pivots Intel from memory to microprocessors. → A textbook strategic reinvention.
  3. 23 "My Enemy's Enemy" — The Rise of Korea Samsung becomes a chip power with US support. → How the US deliberately cultivated Korea as a Japan counterweight.
  4. 24 "This Is the Future" Carver Mead and Lynn Conway revolutionize chip design with VLSI. → The conceptual split between design and fabrication — the seed of fabless.
  5. 25 The KGB's Directorate T Soviet espionage operations and KGB agent Vladimir Vetrov. → Espionage couldn't substitute for an innovation ecosystem.
  6. 26 "Weapons of Mass Destruction" — The Impact of the Offset Precision-guided munitions reshape military balance. → The Soviets realize they've already lost.
  7. 27 War Hero Gulf War (1991) showcases chip-enabled weapons. → The world watches the future of warfare on CNN.
  8. 28 "The Cold War Is Over and You Have Won" Japan's chip industry declines. → Why the first chip war ended in US victory.
Part V

Integrated Circuits, Integrated World?

The globalization era. TSMC's founding and China's first attempts.

  1. 29 "We Want a Semiconductor Industry in Taiwan" Morris Chang founds TSMC (1987). → The single most consequential business decision in modern tech.
  2. 30 "All People Must Make Semiconductors" China's failed early attempts under Mao and after. → Why China kept losing for decades despite massive effort.
  3. 31 "Sharing God's Love with the Chinese" SMIC's founding by Richard Chang. → China's first real foundry — and its limits.
  4. 32 Lithography Wars Nikon vs Canon vs ASML. → How ASML won the most critical tool monopoly.
  5. 33 The Innovator's Dilemma Intel begins to slip. → Where Intel's current troubles trace their roots.
  6. 34 Running Faster? Strain on Moore's Law. → Why each new node now costs exponentially more.
Part VI

Offshoring Innovation?

The fabless model, Apple Silicon, EUV, and the cementing of TSMC's monopoly.

  1. 35 "Real Men Have Fabs" Jerry Sanders' famous line about owning manufacturing. → The debate over whether to keep fabs at home.
  2. 36 The Fabless Revolution Nvidia, Qualcomm, AMD as pure designers. → Critical context for understanding Nvidia's role today.
  3. 37 TSMC's Grand Alliance How TSMC out-coordinated everyone through ecosystem orchestration. → Why TSMC is structurally unreplaceable, not just technically.
  4. 38 Apple Silicon Apple becomes a chip designer. → Vertical integration's return; how custom silicon became a moat.
  5. 39 EUV Extreme ultraviolet lithography invented after a decades-long bet. → The single biggest tech bet of the century.
  6. 40 "There Is No Plan B" ASML's monopoly on EUV machines. → The ultimate chokepoint — the technical reason export controls work.
  7. 41 How Intel Forgot Innovation Intel's manufacturing stumbles in the 2010s. → How the former US champion ceded leading-edge manufacturing to TSMC.
Part VII

China's Challenge

China's catch-up strategy: industrial policy, IP transfer, Huawei, and the military offset.

  1. 42 Made in China 2025 Beijing's industrial policy. → What China is actually trying to do, and why Xi made it personal.
  2. 43 "Call Forth the Assault" China commits to technological self-sufficiency. → Davos rhetoric vs the domestic reality.
  3. 44 Technology Transfer Joint ventures, forced licensing, and personnel poaching. → How tech transfer actually happens — directly echoes the dynamics in Apple in China.
  4. 45 "Mergers Are Bound to Happen" Chinese acquisition attempts in the global chip industry. → How Beijing tried to buy capability — and how the US started blocking it.
  5. 46 The Rise of Huawei From telecom equipment to chip design ambitions via HiSilicon. → Why Huawei became Public Enemy #1 in Washington.
  6. 47 The 5G Future Why 5G mattered geopolitically. → The first wireless standards battle the US tried to win on national-security grounds.
  7. 48 The Next Offset The Pentagon revisits its offset doctrine for the AI/autonomous era. Covers DoD thinking, DARPA, and the realization that future military advantage depends on chip and AI superiority. → The conceptual hinge where chips become explicitly about AI dominance. Most directly relevant chapter for an AI/ML practitioner — it's the bridge between semiconductor policy and AI strategy.
Part VIII

The Chip Choke

The current era: weaponized chokepoints, export controls, and the Taiwan question.

  1. 49 "Everything We're Competing On" Trump-era tech war begins in earnest. → The shift from industrial policy to weaponizing chokepoints.
  2. 50 Fujian Jinhua The Micron IP-theft case study. → How one DOJ enforcement action changed the rules of the game.
  3. 51 The Assault on Huawei US cuts Huawei off from TSMC and EDA tools. → The single most important precedent for current AI chip controls.
  4. 52 China's Sputnik Moment? Beijing's massive response: subsidies, talent recruitment, indigenization push. → Why sanctions may have accelerated rather than slowed China's effort.
  5. 53 Shortages and Supply Chains The 2020–21 chip shortage. → How fragile the global system actually is when one node hiccups.
  6. 54 The Taiwan Dilemma TSMC concentration risk and the cross-strait scenario. → The trillion-dollar geopolitical question that hangs over everything.
Part IX

The AI Chip War

The 2022–2026 period in which AI compute demand reorganized the entire industry: ChatGPT broke open the demand side; Oct 7 2022 weaponized the supply side; Nvidia became the most valuable company on earth; HBM made Korea indispensable; hyperscaler ASICs went from exotic to existential; and DeepSeek, Qwen, Kimi, Huawei Ascend, and AMD's MI line all began testing the assumption that frontier AI required leading-edge Western chips. The book's open ending, brought through to "as of writing" (May 2026).

  1. 55 The ChatGPT Detonation OpenAI launches ChatGPT (Nov 30, 2022); GPT-3.5/4 era; scaling laws as industry doctrine; the transformer-era inflection that made AI compute-bound for the first time. → Why chips became the binding constraint of intelligence itself.
  2. 56 The October Seventh Earthquake Oct 7, 2022 BIS export controls: 14nm logic, 18nm DRAM, 128-layer NAND, AI performance thresholds, the Foreign Direct Product Rule extension, the U.S. persons rule that triggered an exodus of Americans from Chinese chip jobs. Gregory Allen's "unprecedented act of economic statecraft." → The day the chip war became open economic warfare.
  3. 57 Jensen's Decade Nvidia from cryptocurrency-cycle hangover (early 2023) to over $3 trillion (mid-2024) to top-of-the-S&P (2025) to fiscal 2026 data-center revenue near $200B. The CUDA moat built since 2006. H100, H200, B100/B200, GB200 NVL72, Rubin. The "AI factory" doctrine. Jensen Huang's leather jacket, Computex keynotes, geopolitical theater. → How a fabless GPU designer became the hinge of the global economy.
  4. 58 HBM and the Memory Wall High-bandwidth memory as AI-era kingmaker. SK Hynix's surprise leadership at HBM3 and HBM3E. Samsung's stumble through 2024, late-2025 Nvidia qualification, and 2025–26 share recovery. Micron's HBM3E leapfrog. The Q3 2025 ranking shift. Yield secrets, advanced packaging, the through-silicon-via process. → How Korea suddenly became as indispensable as Taiwan.
  5. 59 The Hyperscaler Silicon Pivot Google TPU comes of age (TPU v4 → v5p → Trillium / v6). AWS Trainium and Inferentia. Microsoft Maia. Meta MTIA. Apple's Project ACDC and the Baltra chip. Broadcom and Marvell as the picks-and-shovels designers. Custom silicon as a margin and dependency strategy. → Custom silicon goes from exotic to essential at hyperscaler scale.
  6. 60 Mate 60 and the SMIC Surprise Aug 29, 2023 Huawei Mate 60 Pro launch timed to Raimondo's Beijing visit. TechInsights teardown of the Kirin 9000s. SMIC 7nm via DUV multi-patterning, deep technical analysis of yields and cost. The 2024 Pura 70 and Mate 70 follow-ups. → How China climbed a node without EUV — and where the climb stalls.
  7. 61 The October 2023 Reload Oct 17, 2023 BIS expansion: closing the H800 / A800 loophole, performance-density metrics, advanced computing rules, gaming chip dual-use concerns. Nvidia's H20 China-specific chips. The ratchet pattern: every Beijing workaround answered, every six to twelve months. → Why export controls became a rolling negotiation rather than a single act.
  8. 62 CHIPS Act, Promises and Friction The $52B program execution 2022–2025. TSMC Arizona Phase 1 delays, the labor-and-workmanship dispute, the 2024 award negotiation. Intel's $7.86B finalized award and the August 2025 federal equity stake. Samsung Texas. Micron New York. Raimondo as architect. The roughly $500B in induced private capex. → America's biggest industrial-policy bet in fifty years, and its mid-game scoreboard.
  9. 63 DeepSeek Day Jan 27, 2025: DeepSeek-R1 / V3 release; Nvidia loses roughly $600B in market cap in a session. The accusations: smuggled H100s, training on banned hardware. The rebuttals: efficient training, MoE architecture, MLA, FP8 PTX-level kernels, real software discipline. The Jevons-Paradox debate. The export-control thesis put on trial in real time. → The day the chip war's strategic premise was challenged in public.
  10. 64 The Open-Source Tide The Chinese open-source ecosystem behind DeepSeek: Alibaba Qwen2.5/3 and QwQ; Moonshot Kimi K2 and the long-context push; Zhipu/Z.ai GLM-4.5; MiniMax-01 and M1; ByteDance Doubao and Seed; Baidu ERNIE 4.5/5.0; Tencent Hunyuan; 01.AI's Yi and Kai-Fu Lee's pivot away from pre-training; Stepfun, Baichuan. The Hugging Face data: ~40% of new derivatives running on Chinese weights by mid-2025; the Beijing–Hangzhou–Shanghai–Shenzhen geography; Xi's Feb 17, 2025 private-enterprise symposium with Liang Wenfeng in the front row. → Why DeepSeek was a node in an ecosystem, not an anomaly — and why open weights became China's most successful export.
  11. 65 The Trump II Theater Trump returns to the White House January 2025. "Taiwan stole our chips" rhetoric. The AI Diffusion Rule (issued Jan 13, 2025; rescinded May 2025) and the three-tier global system. Howard Lutnick at Commerce. Section 232 chip-import investigations and the January 2026 narrow 25% tariff. The Aug 2025 federal equity stake in Intel. Tariff negotiations with Taiwan and Korea. The Feb 2026 Supreme Court IEEPA ruling. → How the chip war's American politics fractured along industrial-policy and tariff lines, and why the second front became domestic.
  12. 66 The Energy Wall AI data center power demand outstrips grid availability. Microsoft–Constellation reopening Three Mile Island (Sep 2024). xAI's Memphis substation and unpermitted gas-turbine fight. Texas ERCOT and PJM strain. The new bottleneck that isn't fab capacity. Nuclear renaissance bets, SMR contracts, hyperscaler PPA scarcity. Meta Hyperion Louisiana. → The binding constraint of AI scale shifts from wafers to gigawatts.
  13. 67 AMD's Awakening AMD's Instinct line goes from afterthought to credible alternative: MI300X (Dec 2023), MI325X, MI355X, the MI400 roadmap. ROCm versus CUDA and the SemiAnalysis "CUDA Moat Still Alive" debate. The hyperscaler design wins (Microsoft Azure ND MI300X v5, Meta running Llama 3.1 405B on MI300X exclusively, Oracle BM.GPU.MI300X.8 superclusters). The Zyphra ZAYA models. The October 2025 OpenAI MI450 warrant deal. Lisa Su's long quiet rebuild. → The first crack in the assumption that "AI training" means "Nvidia."
  14. 68 The ASIC Inflection Custom silicon takes meaningful share of frontier workloads. Trainium2 ramp at AWS through 2024–2025; Trainium3 in 2025–2026; Project Rainier in New Carlisle, Indiana scaling toward a million Trainium2 chips. Google TPU v6 Trillium and the v7 Ironwood successor; Anthropic's million-Ironwood commitment. OpenAI's Stargate footprint and the Broadcom XPU co-design. Microsoft's CoreWeave dependence. Apple's first reported TPU training disclosure for Apple Intelligence. → Custom silicon stops being side-business and starts being load-bearing.
  15. 69 The Five Labs and Their Compute The big three frontier labs and the two smaller-but-powerful brothers, told through their chip choices. OpenAI: Microsoft-Azure dependence, Stargate diversification, the Broadcom custom-silicon project, the AMD MI450 warrant deal, AWS $38B partnership. Anthropic: Google TPU primary, AWS Trainium2/3, the Colossus rental of May 2026. Google DeepMind: TPU vertical integration as the structural advantage. Meta: MTIA inference + Nvidia GPU training clusters; Llama as open-source play; Hyperion. xAI: Colossus 1 → 2; the Feb 2026 SpaceX fold-back; cofounder exodus through March 2026. Apple Project ACDC / Baltra and Tesla Dojo's August 2025 cancellation as side-strokes. Cerebras / Groq / Tenstorrent / SambaNova as the inference-specialist alternatives. → How the labs' silicon dependencies define their competitive paths.
  16. 70 Neoclouds and Circular Money The financialization of compute: CoreWeave (founded 2017 as a crypto miner; pivot to GPU-as-a-service; ~60% Microsoft revenue concentration; March 2025 IPO; market cap above $70B at peak); Lambda Labs and the Series E ramps; Crusoe and the Stargate Abilene build; Nebius (the Yandex Europe spinout); Together / Vultr / Voltage Park / Fluidstack as the second tier. The circular financing concerns: Nvidia invests in CoreWeave; CoreWeave buys Nvidia GPUs; CoreWeave leases them to Microsoft and OpenAI; Microsoft funds OpenAI; OpenAI books capacity at CoreWeave; the same dollars round-trip through the same balance sheets. The 2000 telecom-build parallel. Bain's "$800B question." Jim Chanos, Michael Burry, the bear case. → Why the chip war's commercial layer became its own systemic risk story.
  17. 71 DeepSeek-V4 and the Huawei Inflection DeepSeek-V4 in April 2026: the lab discloses Huawei Ascend (910C and CloudMatrix-class systems) for inference at scale and validates V4 on both Nvidia and Ascend NPU platforms. Huawei's CloudMatrix 384 and the Ascend 910C → 950 ramp through 2025–2026. The Eric Xu Connect 2025 roadmap (950, 960, 970). What this means for the export-control thesis: Beijing has demonstrably sustained frontier-class AI without EUV-leading-edge access. What it does NOT mean: that Taiwan's role is replaced or that the gap has closed. The book's honest finale, set in May 2026: what's settled, what isn't, and the question the reader is now responsible for thinking about.