A few chip firms now hold more leverage than oil

1

Engineers who walked out

Miller holds that Silicon Valley’s edge came from a culture that let engineers quit and found rivals, so talent circulated instead of staying locked in one lab.

Miller opens with the 1957 departure of eight engineers from William Shockley’s lab to create Fairchild Semiconductor. That quit became a template: ideas moved with people across firms rather than staying inside one company or a government lab. Robert Noyce and Gordon Moore later left Fairchild again to start Intel in 1968. Each move spread process know-how further; rivals hired one another’s engineers instead of relying on secrecy alone. The transistor had appeared at Bell Labs in 1947 inside a regulated monopoly. Fairchild’s planar process and the integrated circuit—developed by Noyce and independently by Jack Kilby at Texas Instruments in 1958-59—made complex circuits manufacturable at scale. Moore’s 1965 claim that density would roughly double every two years hardened into an industry commitment. Firms spent billions expecting rivals to hit the same pace, so falling one generation behind could end a player. Leadership, Miller argues, rested on capital intensity and organizational churn as much as clever circuits.

In practice: The traitorous eight left Shockley for Fairchild; Noyce and Moore then left Fairchild to found Intel, spreading expertise across the valley.

2

Missiles bought the first chips

Cold War military demand, not consumer markets, absorbed early fixed costs and made semiconductors commercially viable.

In the first decade chips cost too much for ordinary buyers. The Apollo program and the Minuteman II guidance system were the customers who mattered. NASA and the Air Force paid premium prices for integrated circuits when almost no one else could. That volume let Fairchild and Texas Instruments ride costs down the learning curve. By the mid-1960s cheaper parts opened civilian markets—first calculators, then computers. Miller’s claim is that state demand did not so much distort a market as create one before commerce could sustain itself. Missile accuracy mapped directly onto transistor counts in a nose cone, locking in the idea that chip leadership equals military power. By the 1970s consumer electronics outgrew defense demand, and private global firms took the frontier—setting up the later shock when Japan surged in memory.

In practice: Apollo and Minuteman II paid premium IC prices that funded the learning curve before calculators and computers could.

3

Japan takes memory

Japanese discipline and patient capital seized the DRAM market in the 1980s, proving manufacturing could beat US design in commodity chips.

By the early 1980s NEC, Toshiba, and Hitachi held most of the global DRAM market and pushed aside American pioneers including Intel. Miller points to lower capital costs, bank ties that tolerated thin margins for years, and a defect-reduction culture. Hewlett-Packard’s own quality tests found Japanese chips failed less often than American ones. MITI coordinated some joint research, yet Miller stresses firm-level discipline mattered more than central planning and that popular accounts often overstate the ministry. Alarm in the US ran hot. Intel exited DRAM entirely in 1985 and bet on microprocessors—a pivot that later paid off when PCs exploded. Washington answered with trade pressure and the 1986 US-Japan Semiconductor Agreement, setting minimum prices on Japanese chips in the US and prying open Japan’s market. The chapter shows the first great disruption came from an ally, not a peer rival, so dominance was never guaranteed.

In practice: HP testing showed Japanese DRAMs failed less often; Intel quit the memory business it had invented in 1985.

4

Design leaves the factory

American firms went fabless to escape soaring plant costs, a rational move that seeded the overseas manufacturing dependency now treated as strategic risk.

Fab costs roughly doubled every few years as features shrank. Firms had to keep building ever-dearer plants or specialize in design and outsource production. Many chose the latter. Qualcomm and later Nvidia designed chips but never ran fabs, freeing capital for research and focusing on their strongest work. Actual fabrication—the hardest, most capital-heavy step—migrated to firms willing to own the plants. Offshoring here was not mainly cheap labor; wafer fabs are automated and capital-intensive. The driver was capital efficiency and specialization. Sematech, launched in 1987 with chipmakers and the federal government, pooled research to keep domestic manufacturing alive and helped equipment supply chains. It could not reverse the economics that made outsourcing attractive. Concentration of advanced capacity overseas became the vulnerability the rest of the book tracks.

In practice: Qualcomm and Nvidia designed without owning fabs; Sematech tried to slow the hollowing-out but could not rewrite the cost curve.

5

Chang’s foundry model

Morris Chang’s pure-play foundry at TSMC did not just build a company; it reorganized the industry around one geographic manufacturing choke.

Chang, a longtime Texas Instruments manager, returned to Taiwan and founded TSMC in 1987 with government backing. The structural bet: manufacture only, on contract, for other firms’ designs. That removed the conflict that made fabless customers fear using an integrated rival’s plant. TSMC would never sell a competing chip against its own clients. The model fit the US fabless wave already underway. By the 2000s TSMC built for Apple, Nvidia, and Qualcomm. Miller stresses this was not chiefly cheap labor. Taiwan accumulated engineers, equipment suppliers, and process skill over decades that others could not copy quickly. By the 2020s TSMC made the large majority of the world’s most advanced logic chips—those below 10-nanometer nodes—needed for smartphones, AI, and advanced weapons. Globalization here consolidated risk onto one contested island rather than spreading it.

In practice: TSMC’s contract model won Apple, Nvidia, and Qualcomm as customers and came to dominate logic production below 10 nanometers.

6

The uncopied EUV machine

ASML’s sole mastery of extreme-ultraviolet lithography gives allied export licenses a lever stronger than tariffs on finished goods.

Etching the tiniest patterns requires EUV light. Only ASML in the Netherlands builds those machines. Dominance grew from Dutch optics work, German lenses from Zeiss, American research linked partly to Sematech-era programs, and a decision to pursue EUV when many doubted the physics. One machine costs over 150 million dollars, takes months to install, and needs parts from dozens of specialized suppliers with no ready substitutes. The barrier is technical difficulty—light source, atomically precise mirrors, vacuum engineering—not simple collusion. China has not replicated it despite years of state effort. Because ASML exports need Dutch licenses shaped by US pressure, Washington and allies hold a choke over advanced Chinese fabs. Withholding tools hits deeper than duties on end products. Miller treats this monopoly as proof that one allied-controlled step can outweigh another country’s broad industrial bulk.

In practice: When the Dutch blocked EUV sales to China in 2019 under US pressure, the move did more to slow Chinese advanced chipmaking than years of tariffs.

7

Chips decide the battlefield

The 1991 Gulf War showed precision munitions built on semiconductors multiply force, turning chip leadership into a direct military requirement.

Precision-guided bombs, satellite navigation, and stealth depended on processing power that barely existed at scale a decade earlier. American forces hit targets with far fewer munitions than older wars demanded. Every planner watching—including in Beijing and Moscow—saw the disparity. Miller notes the shift in Chinese military doctrine that followed: future wars would turn on information technology and precision weapons, not raw troop counts. China began a decades-long drive for indigenous semiconductor capacity as national security, not just commercial policy. Made in China 2025 and heavy subsidies to firms such as SMIC sit in that line. Miller links the pattern back to the Cold War birth of the industry: chips prove decisive in war, states scramble for supply, and the scramble rewrites civilian incentives for a generation. Treating chips as pure commodities, he argues, has always been a category error.

In practice: After watching the Gulf War gap, Chinese strategists recast information technology and precision weapons as the decisive factors and pushed domestic chip capacity.

8

Controls to freeze the gap

US export rules on advanced chips and equipment try to hold China a fixed distance behind the frontier, trading near-term market loss for strategic delay.

By the late 2010s Huawei and other Chinese firms looked globally competitive in telecom gear and phones. Washington feared chip self-sufficiency and an eventual pass of US and allied capability. In 2019 and more sharply in October 2022, Commerce barred sales of advanced chips, design software, and equipment above set thresholds to Chinese firms. It pressed the Netherlands and Japan to align. The logic exploits earlier chokes: subsidies cannot replace ASML’s EUV tools or software from firms such as Synopsys and Cadence. Moore’s Law means a few lost years compound into a wider gap. The aim is delay at the leading nodes, not total shutdown of China’s industry. Miller is candid about costs. Controls spur Chinese domestic alternatives, risk rival technical standards, and hurt American and allied vendors who lose a huge customer base. The open mobility that once built the US lead gives way, in his telling, to deliberate restriction.

In practice: October 2022 rules extended bans on advanced chips, design software, and manufacturing equipment, with allied pressure to match.

9

Taiwan as world fulcrum

Roughly ninety percent of the most advanced logic chips are made in Taiwan, a silicon shield Miller treats as fragile deterrent and tempting target alike.

Beijing claims the island and has not ruled out force. Some voices Miller cites, including in Taiwan’s policy world, call the concentration a shield: wrecking or seizing TSMC would shock the global economy, including China’s electronics sector, and could force heavy external intervention. Miller stays skeptical. Advanced fabs need continuous ultrapure materials, specialty gases, and skilled staff. A blockade could achieve aims without missiles hitting plants. TSMC leadership has quietly noted the factories cannot be airlifted or rebuilt elsewhere on short notice. Responses have begun: TSMC building in Arizona, Samsung expanding in Texas, Intel pushing leading-edge work with backing from the 2022 CHIPS and Science Act’s roughly 52 billion dollars in subsidies. Miller stresses these efforts need years to reach Taiwan’s process maturity. For the foreseeable future the world still depends on one geopolitically exposed island. That dependence is the open question the industry’s history leaves unresolved.

In practice: TSMC’s Arizona plans, Samsung in Texas, and Intel’s CHIPS Act-backed push still lag Taiwan’s existing leading-edge maturity.