Chip War
Chip War — Chris Miller
The Central Thesis
Semiconductors are the foundation of modern economic and military power. Everything from smartphones and data centers to AI and precision-guided missiles depends on chips. Yet the ability to design and manufacture the most advanced semiconductors is concentrated in a handful of companies, mostly in the U.S. and allied countries.
The central paradox: Decades of globalization and specialization have made computing extraordinarily cheap, but created critical choke points. The U.S. controls key design software, intellectual property, and manufacturing equipment, while Taiwan produces most of the world’s most advanced processors. China is heavily dependent on this system and is investing enormous resources to escape that dependence.
The book traces how semiconductor leadership moved across countries, why manufacturing expertise is so difficult to replicate, and how control of computing technology became central to the U.S.–China rivalry.
1. The Birth of Silicon Valley: From Invention to Mass Production (1940s–1960s)
The semiconductor industry emerged from American military research, university physics laboratories, and a new entrepreneurial culture in California.
-
1947 — The transistor: William Shockley, John Bardeen, and Walter Brattain at Bell Labs develop the transistor, replacing large, unreliable vacuum tubes with small electrical switches. Shockley moves to California and founds Shockley Semiconductor in 1955, but his terrible management drives eight engineers to leave and establish Fairchild Semiconductor in 1957.
-
1958–1959 — The integrated circuit: Jack Kilby at Texas Instruments in Dallas and Bob Noyce at Fairchild independently develop methods for integrating multiple electronic components onto a single semiconductor. Noyce’s design, based on Jean Hoerni’s planar process, is better suited to mass production. Gordon Moore and Andy Grove are among the engineers who help turn these inventions into an industry.
-
Manufacturing is as important as invention: Early chips are expensive, unreliable, and difficult to produce. Jay Lathrop’s photolithography techniques allow microscopic patterns to be printed using light. Morris Chang, a young manufacturing engineer at Texas Instruments, raises the yield of a transistor production line from nearly zero to 25% through systematic experimentation with manufacturing conditions. The decisive challenge is making millions of reliable chips, not just inventing one.
-
NASA and the Pentagon create the initial market: The Apollo guidance computer and Minuteman II nuclear missile require compact, lightweight, reliable chips. Military and space programs can pay prices civilian customers cannot afford, funding the industry’s early development.
-
Bob Noyce creates the civilian market: Noyce aggressively cuts Fairchild’s chip prices to encourage mass adoption in commercial computers. By 1968, civilian computer manufacturers buy as many integrated circuits as the military. The industry’s growth increasingly depends on large commercial markets rather than defense contracts.
-
Moore’s Law (1965): Gordon Moore predicts exponential growth in the number of components on a chip. This produces dramatic reductions in the cost of computation, enabling mainframes, calculators, personal computers, smartphones, and eventually AI. The trend is sustained by constant advances in manufacturing, not by an automatic law of physics.
Key insight: Silicon Valley’s advantage was not just scientific talent. It combined government demand, university research, manufacturing expertise, venture capital, entrepreneurial competition, and access to a massive civilian market.
2. The USSR Falls Behind Despite Excellent Scientists (1960s–1980s)
The Soviet Union recognizes the importance of semiconductors and invests heavily in research facilities, notably the semiconductor center at Zelenograd. It has talented physicists and extensive espionage networks but never develops a competitive chip industry.
-
Copying instead of innovating: Soviet planners increasingly prioritize reproducing American chips rather than designing independent technologies. By the time a Western design is copied, American manufacturers are often producing a more advanced generation.
-
Manufacturing limitations: Producing a chip requires mastering precise chemical processes, clean manufacturing environments, equipment, and quality control. Obtaining a design does not reproduce the know-how needed to manufacture it reliably at scale.
-
Central planning and market size: Soviet factories operate within a rigid bureaucracy, with weaker incentives to reduce costs, improve quality, or develop products for customers. American companies benefit from much larger civilian markets across the U.S. and its allies, spreading R&D costs over massive production volumes.
-
Espionage cannot close the gap: The KGB steals designs and manufacturing information, but technological progress is too fast and industrial expertise too difficult to reproduce. By the 1980s, Soviet microprocessors typically lag leading American designs by several years.
Key insight: Technology is not simply a collection of scientific discoveries or blueprints. Competitive advantage lies in an industrial ecosystem capable of continuously improving designs and manufacturing processes.
3. Japan’s Rise and America’s First Semiconductor Crisis (1960s–1980s)
Japan becomes the first major challenger to American semiconductor dominance, initially through consumer electronics and later through superior memory-chip manufacturing.
-
Sony and consumer electronics: Akio Morita and Masaru Ibuka use transistor technology to build portable radios and later products like the Walkman. Japanese firms demonstrate that they can manufacture American-origin technologies cheaply and reliably, then create successful products of their own.
-
Cold War alignment: Washington supports Japan’s economic recovery and integration into the American-led trading system, viewing a strong Japan as strategically important against communism. Japanese firms gain access to American technologies and markets.
-
Manufacturing excellence: Japanese companies develop superior quality control, production processes, and manufacturing efficiency. Combined with government industrial policy and access to inexpensive bank financing, this enables them to challenge American companies in electronics, cars, and semiconductors.
-
The DRAM war: By the 1980s, companies including NEC, Toshiba, Hitachi, and Fujitsu are aggressively competing in memory chips. DRAM rewards high yields, large production volumes, and continuous capital investment. Japanese firms outperform many American producers on cost and reliability, forcing them out of the market.
-
American national security concerns: Chips have become essential to military satellites, radar, surveillance, and precision-guided weapons. Losing domestic semiconductor production threatens the U.S. military’s technological advantage. American chipmakers lobby Washington to respond to Japanese competition.
-
Washington’s response: The Reagan administration pressures Japan over alleged dumping and market access, leading to the 1986 U.S.–Japan Semiconductor Agreement. In 1987, the industry and Pentagon establish Sematech, led by Bob Noyce, to improve American manufacturing capabilities and coordination with equipment suppliers. Trade restrictions raise memory prices but do not rescue most American DRAM producers.
Key insight: Japan demonstrates that manufacturing quality and industrial investment can displace technological pioneers. But government protection and trade restrictions cannot substitute for companies developing competitive products and production processes.
4. Intel’s Reinvention and the PC Revolution (1970s–1990s)
Bob Noyce and Gordon Moore leave Fairchild to found Intel in 1968. Intel initially focuses on memory chips before becoming the dominant microprocessor company.
-
The microprocessor: Intel introduces the 4004 in 1971, integrating a computer’s processing functions onto a chip. Intel’s subsequent x86 architecture becomes central to personal computing.
-
The IBM PC (1981): IBM chooses an Intel microprocessor and Microsoft’s operating system for its personal computer. IBM originally dominates the system, but value and market power increasingly shift toward Intel’s processors and Microsoft’s software. PC makers can compete in assembly while relying on these shared technologies.
-
Intel’s DRAM crisis: By the mid-1980s, Japanese competitors have made Intel’s memory business uncompetitive. Andy Grove and Gordon Moore reluctantly decide to exit DRAM and concentrate on microprocessors, despite memory having been Intel’s founding business.
-
Grove’s turnaround: Intel makes painful workforce and capacity reductions, improves manufacturing discipline, and focuses on relentlessly advancing x86 processors. Its success depends on both chip architecture and the ability to manufacture increasingly sophisticated processors efficiently.
-
Wintel dominance: The combination of Windows software and Intel chips becomes the standard PC architecture. Better processors enable more capable software, which increases demand for computing power. Intel achieves extraordinary profitability and market dominance.
-
Japan misses the PC transition: Japanese semiconductor firms remain strong in memory but do not establish comparable control over the PC processor and software ecosystem. Japan’s financial bubble bursts in the early 1990s, weakening an economy already suffering from structural problems and excessive dependence on bank financing.
Key insight: Intel survives Japan’s challenge not by winning back the DRAM market, but by abandoning it and dominating a different, faster-growing market. Its processor and manufacturing capabilities become more valuable than its original memory business.
5. South Korea Takes Over Memory Manufacturing (1980s–1990s)
South Korea uses aggressive industrial investment, American technology, and favorable geopolitical conditions to become a leading producer of memory chips.
-
Samsung’s bet: Samsung commits enormous capital to DRAM production despite the technical challenges and cyclical nature of memory prices. Government support and financing help the company invest through industry downturns.
-
American cooperation: U.S. semiconductor firms view Korean companies as useful competitors to Japan. Samsung obtains technology from American firms, including Micron, while Intel and other companies cooperate with Korean memory producers.
-
The U.S.–Japan trade conflict helps Korea: Restrictions on Japanese DRAM exports create opportunities for Korean suppliers. Lower labor costs and substantial capital investment help Korean firms gain market share.
-
Japanese decline: As Korean companies improve their manufacturing processes and expand production, Japan loses its dominant position in memory. By the late 1990s, South Korea has become the world’s leading DRAM-producing country.
Key insight: Semiconductor leadership shifts through a combination of technological learning, scale, capital investment, and geopolitical relationships. America’s attempts to weaken Japanese dominance inadvertently help create another powerful Asian semiconductor industry.
6. Taiwan and TSMC Reinvent Chip Manufacturing (1970s–2000s)
Taiwan initially develops semiconductor capabilities through American investment, technology transfer, and government support. Its transformation comes when Morris Chang creates a new business model for the industry.
-
Taiwan’s industrial strategy: Officials including K. T. Li invest in electronics research, technology transfer, and Hsinchu Science Park. Taiwan’s Industrial Technology Research Institute acquires manufacturing technology from RCA and trains local engineers. The island builds capabilities beyond low-cost assembly.
-
Morris Chang returns: After decades at Texas Instruments, where he became a senior manufacturing executive but was passed over for the top position, Chang moves to Taiwan. In 1987, he founds Taiwan Semiconductor Manufacturing Company (TSMC), with support from Taiwan’s government and Philips.
-
The foundry model: Traditional semiconductor companies design and manufacture their own chips. Chang proposes that TSMC manufacture chips designed by other firms without developing competing designs. Customers can focus on product differentiation without needing to invest billions in fabrication facilities.
-
The fabless revolution: Nvidia, Qualcomm, and other chip designers benefit from outsourcing manufacturing. AMD eventually separates its manufacturing business. New companies can compete in complex chip design without owning fabs, greatly reducing the capital required to enter the industry.
-
TSMC’s scale advantage: By aggregating demand from many designers, TSMC can finance increasingly expensive equipment and achieve high manufacturing volumes. More production generates greater process expertise and better yields, reinforcing its technological lead.
-
Advanced manufacturing concentrates: As leading-edge fabs become more expensive, fewer firms can compete. TSMC and Samsung emerge as the primary manufacturers of the most advanced logic chips, while Intel struggles to maintain its previous lead.
Key insight: Morris Chang’s business-model innovation transforms the entire semiconductor industry. Separating design from fabrication makes chip design more accessible, while concentrating advanced manufacturing in a tiny number of firms.
7. ASML and the Lithography Choke Point (1980s–2010s)
Manufacturing advanced chips requires extraordinarily precise tools. Lithography—the process of printing microscopic patterns onto silicon wafers—becomes one of the industry’s most difficult and strategically important technologies.
-
The lithography competition: American equipment companies initially lead, but Nikon and Canon become dominant competitors. ASML, founded in the Netherlands in 1984, gradually gains market share through partnerships, advanced engineering, and a strategy of integrating components from specialized international suppliers.
-
Extreme ultraviolet lithography (EUV): As transistors shrink, manufacturers need shorter-wavelength light to print smaller features. EUV uses light with a wavelength of 13.5 nanometers, requiring exceptionally sophisticated light sources, mirrors, vacuum systems, and precision mechanics.
-
Decades of investment: Commercializing EUV requires American national laboratory research, billions of dollars of R&D, and suppliers across multiple countries. Intel, Samsung, and TSMC invest directly in ASML to support development.
-
ASML becomes indispensable: ASML is the only company able to supply commercial EUV machines. Its equipment relies on specialized components, including light sources from its U.S. subsidiary Cymer and optics from Germany’s Zeiss.
-
A broader system of dependencies: Chip design relies heavily on American electronic design automation software from Cadence, Synopsys, and Mentor/Siemens. Manufacturing requires specialized machines from companies including Applied Materials, Lam Research, KLA, ASML, and Japanese suppliers.
Key insight: Owning a critical part of the manufacturing process can provide enormous strategic leverage. Even companies that can design advanced chips and finance fabs cannot independently recreate the tools required to manufacture them.
8. Intel Misses Mobile While Apple and TSMC Win (2000s–2020)
After Andy Grove’s leadership, Intel’s dominance in PC and server chips makes it highly profitable but increasingly resistant to major changes in technology and business models.
-
Protecting the x86 business: Intel earns large margins from PC and server processors. Management emphasizes financial performance and struggles to respond to mobile computing, which requires different processor architectures and power-consumption tradeoffs.
-
The iPhone mistake: Intel declines an opportunity to manufacture chips for Apple’s early iPhones. CEO Paul Otellini later acknowledges that he underestimated the opportunity. Smartphones instead develop around Arm-based architectures.
-
Apple designs its own processors: Apple develops chips optimized for its hardware and software, relying on TSMC to manufacture them. The relationship combines Apple’s enormous product demand with TSMC’s manufacturing capabilities.
-
Intel’s manufacturing delays: Intel struggles to introduce new fabrication processes on schedule, while TSMC and Samsung continue investing in advanced manufacturing. By 2020, TSMC is producing Apple’s sophisticated A14 processor with 11.8 billion transistors.
-
The changing value chain: Apple’s most important processor is designed in California, manufactured in Taiwan, and assembled into a phone in China. Chips account for a large share of the iPhone’s component costs, even though final assembly attracts most of the attention.
Key insight: Intel’s dominant position in PCs does not guarantee leadership in the next computing platform. TSMC benefits from manufacturing for multiple successful chip designers rather than betting on one processor architecture or end market.
9. China’s Semiconductor Dependence and the Chip Choke (2000s–2022)
China becomes the world’s largest electronics assembly hub but remains dependent on foreign technology for many of the chips and production tools its economy needs.
-
The scale of dependence: China imports more semiconductors by value than oil, including chips used in smartphones, computers, telecommunications equipment, industrial electronics, and military systems. Although China manufactures many less advanced chips, it lacks a fully independent leading-edge supply chain.
-
Made in China 2025: Beijing identifies semiconductor self-sufficiency as a national priority and commits enormous state funding. Chinese firms invest in design, fabrication, memory, and equipment, while pursuing foreign partnerships, acquisitions, and technology transfers.
-
The limits of copying: China faces many of the same challenges that plagued the Soviet industry: chipmaking requires specialized equipment, materials, software, process knowledge, and decades of accumulated manufacturing expertise. Funding alone cannot immediately reproduce the capabilities of TSMC, ASML, or leading American suppliers.
-
Fujian Jinhua (2018): A Chinese DRAM manufacturer attempting to build domestic memory capacity becomes embroiled in a dispute involving Micron trade secrets. U.S. restrictions on equipment exports severely disrupt its operations, demonstrating the leverage created by American manufacturing technology.
-
Huawei and HiSilicon: Huawei becomes a major smartphone and telecommunications company. Its HiSilicon division designs advanced processors, but relies on Taiwan’s TSMC to manufacture them using equipment and technology with important American inputs.
-
The U.S. export restrictions (2019–2020): The Trump administration places Huawei on the Entity List and later extends restrictions to foreign-produced chips made using specified U.S. technology. This prevents TSMC from continuing to manufacture Huawei’s advanced processors and deals a severe blow to its smartphone business.
-
Weaponized interdependence: Although advanced chip manufacturing has moved abroad, the U.S. retains influence because essential tools and software remain under American control. It can exploit specific choke points without directly controlling foreign factories.
-
China’s dilemma: Restrictions slow access to leading-edge technology while increasing China’s determination to become independent. But rebuilding an entire domestic semiconductor ecosystem is enormously expensive, technically difficult, and slow—especially while global leaders continue improving.
Key insight: America’s semiconductor power does not depend solely on where chips are physically made. Control of indispensable upstream technologies gives it the ability to restrict access to advanced computing across the global supply chain.
10. Taiwan: The Semiconductor Industry’s Greatest Vulnerability
The concentration of advanced manufacturing in Taiwan creates a geopolitical risk far larger than the island’s size would suggest.
-
Advanced logic is highly concentrated: TSMC produces almost all of the most advanced processors designed by companies such as Apple, while Samsung is the other major leading-edge foundry. The book estimates that Taiwan supplies roughly 37% of the world’s new computing power.
-
The silicon shield: Taiwan’s importance to global technology may discourage China from attacking because a conflict would devastate industries and economies around the world—including China’s own electronics sector.
-
But dependence also creates vulnerability: An invasion is not necessary to disrupt chip production. A naval blockade, loss of electricity, interruptions to material supplies, or physical damage to fabs could cripple global production.
-
Fabs cannot be quickly replaced: Leading-edge factories require billions of dollars, scarce machines, highly trained engineers, and years of process development. Replicating Taiwan’s capacity elsewhere would be difficult even with enormous investment.
-
The 2020–2021 shortages: COVID disruptions, shifting demand, a Japanese fab fire, Texas winter storms, and Malaysian lockdowns interrupt chip supplies. Automakers and other manufacturers shut factories because they cannot obtain even relatively simple chips.
-
Attempts to diversify: The U.S., Japan, and Europe pursue domestic manufacturing investment to reduce dependence on East Asia. TSMC plans fabs in Arizona and Japan, Samsung expands U.S. investment, and Intel under Pat Gelsinger launches an ambitious effort to regain manufacturing leadership and establish a foundry business.
-
The economic tradeoff: Global specialization made semiconductor production more efficient and technologically advanced. Recreating redundant capacity for national security is expensive and difficult, and may slow the benefits of global integration.
Key insight: The semiconductor supply chain was built for efficiency and technological progress, not resilience against geopolitical conflict. Its greatest concentration of advanced manufacturing lies at the center of one of the world’s most dangerous territorial disputes.
11. The Future of Computing and Semiconductor Competition
Maintaining Moore’s Law becomes increasingly difficult as transistors shrink and manufacturing costs rise.
-
The rising cost of progress: Advanced fabs require ever more sophisticated lithography, materials, transistor structures, and process control. Fewer companies can afford to compete at the leading edge.
-
New transistor technologies: FinFETs and other three-dimensional transistor designs enable continued improvements despite the limits of traditional structures.
-
Specialized computing: GPUs and other processors optimized for specific applications become increasingly important, especially for AI. Nvidia’s success demonstrates how new processor architectures can create enormous value beyond general-purpose CPUs.
-
Intel’s attempted comeback: Pat Gelsinger’s strategy, as described in the book, involves reclaiming manufacturing leadership, obtaining next-generation ASML equipment, and expanding Intel’s foundry business to manufacture chips for external customers.
-
No guaranteed technological outcome: The book does not conclude that Moore’s Law has ended or that any country’s dominance is permanent. Continued advances depend on engineers overcoming increasingly difficult technical and economic constraints.
Key insight: As advanced chipmaking becomes harder and more expensive, the capabilities of the few companies that can operate at the frontier become even more strategically important.
The Most Important Takeaways
-
Manufacturing is as important as invention. Bell Labs invented the transistor, but the semiconductor revolution required companies that could manufacture increasingly sophisticated chips reliably, cheaply, and at scale. Process expertise is extremely difficult to copy.
-
The U.S. built an innovation ecosystem, not just a chip industry. Government research, defense demand, universities, venture capital, entrepreneurs, and enormous commercial markets reinforced one another. The Soviet Union could copy designs but not reproduce this system.
-
Semiconductor leadership constantly shifts. Japan overtook American memory producers through manufacturing excellence; Intel recovered by dominating microprocessors; Korea displaced Japan in memory; Taiwan became the center of advanced foundry manufacturing.
-
Business-model innovation can be as consequential as technical innovation. TSMC’s separation of design and manufacturing enabled the fabless industry and transformed how advanced chips are developed.
-
The most powerful positions are often upstream choke points. ASML’s lithography machines, American EDA software, and specialized manufacturing equipment are indispensable even when the final chips are produced elsewhere.
-
Commercial computing underpins military superiority. American precision weapons and military systems increasingly depend on advances generated by civilian semiconductor markets.
-
China cannot simply spend its way to semiconductor independence. The challenge is recreating a network of specialized companies, equipment, software, materials, and manufacturing expertise that took decades to develop.
-
America retains substantial technological leverage despite offshoring fabs. Restrictions on Huawei demonstrate how control of critical tools and intellectual property can be used to cut companies off from advanced chips.
-
Taiwan is the central geopolitical vulnerability. The most sophisticated processors are disproportionately manufactured on an island at the center of U.S.–China tensions. Disruption would affect the entire global economy.
-
Globalization made chips cheaper but the supply chain more fragile. International specialization accelerated innovation and concentrated technical expertise, creating dependencies that countries now struggle to unwind.