The most consequential competition in technology is measured in nanometres. Semiconductors — the chips inside phones, cars, data centres and weapons systems — have become the chokepoint of the world economy, and governments are spending hundreds of billions to secure their supply. At TechEpaper, our technology coverage returns to the chip race repeatedly because so much else in tech now depends on it: the AI boom runs on advanced processors that only a handful of factories on Earth can make.
The concentration is extraordinary. One company in Taiwan fabricates the large majority of the world’s most advanced logic chips; one Dutch firm makes the lithography machines those factories require; and a small set of suppliers provides the specialist chemicals, wafers and design software around them. That chain is a triumph of global specialisation — and, policymakers concluded, a strategic vulnerability.
The result is an industrial policy race not seen in decades, with the United States, Europe, Japan, South Korea, India and China all subsidising factories and restricting exports of the most sensitive technologies.
Why advanced chips are so hard to make
A leading-edge chip contains tens of billions of transistors, patterned by extreme ultraviolet light in factories that cost upwards of twenty billion dollars and must run, vibration-free and dust-free, around the clock. Yields — the share of chips that work — determine profitability, and the expertise behind them lives in teams of engineers accumulated over decades. This is why money alone does not buy a chip industry quickly: new fabs take years to build and longer to master.
Programmes such as the US CHIPS and Science Act, administered with research support from agencies including NIST’s CHIPS programme, try to rebuild domestic capacity across the whole chain — fabrication, packaging and materials. Advanced packaging itself has become a bottleneck and a battleground, as AI accelerators demand novel ways of stacking memory and processors together. Trade coverage by Reuters’ chip industry reporting tracks the export controls tightening around these technologies, which now shape where factories are built as much as costs do.
AI demand and the new bottlenecks
Artificial intelligence has rewritten the industry’s demand map. Data centre accelerators command enormous prices and long waiting lists; the constraint has shifted at times from the processors themselves to high-bandwidth memory and packaging capacity. Cloud giants have responded by designing their own silicon, energy suppliers have become kingmakers — a fab and a data centre both need abundant power — and investors have re-rated chip firms among the most valuable companies in the world.
At the other end of the market, mature chips for cars and industry caused shortages of their own during the pandemic, teaching manufacturers that old technology can stop a production line as completely as new. Resilience planning now covers both ends of the spectrum, and stockpiling, dual sourcing and regional capacity have replaced pure just-in-time efficiency as the industry’s operating philosophy.
What the race means beyond the industry
For consumers, the chip race appears as prices and availability: electronics, vehicles and AI services all carry its costs and constraints. For workers in regions landing new fabs, it means construction booms and technical jobs — alongside hard questions about water use, energy demand and whether subsidies deliver value. For governments, chips have become what oil was to an earlier era: a strategic input whose supply cannot be left entirely to the market or to rivals.
The race will not produce a single winner quickly. Expertise this deep diffuses slowly, and the likely outcome is a more regionalised industry — less efficient, more expensive and more secure by design. Whether that trade is worth it is the central industrial question of the decade, and the answer will be written, fittingly, in silicon.
