Semiconductors have become the central nervous system not only of the digital economy but of industrial policy. The CHIPS Act, the EU Chips Act and parallel programs in Japan, Korea, India and the Gulf together represent the largest sectoral industrial intervention since the post-war reconstruction. The aggregate fiscal commitment now exceeds USD 600 billion. The decisions companies make in the next three years will shape competitive positions for the rest of the decade.
From component to instrument of state
What used to be a deeply globalised, ruthlessly cost-optimised supply chain has become an arena where industrial policy, export controls and supply security regularly override unit cost. Companies that still treat the semiconductor stack as a procurement category will be outmanoeuvred by those that treat it as a strategic capability with geopolitical exposure.
Implications for chipmakers
Capacity decisions must now incorporate political optionality alongside cost. Leading-edge fabs increasingly resemble joint ventures between technology, finance and statecraft. The economics of any new fab are shaped not only by demand and yield but by which government's subsidies are unlocked, which export-control regimes apply, and which long-term off-take agreements anchor utilisation. Build decisions that ignore those dimensions are not actually decisions; they are bets.
Implications for buyers
Hyperscalers, automakers and industrial manufacturers are signing decade-long capacity reservations to underwrite supply security. The legacy spot-purchase procurement model is dying for any node that matters. The buyers we admire have organised cross-functional teams — technology, treasury, government affairs, legal — that operate as a single decision unit on supply commitments.
The new geography of fabs
The next decade will see the most ambitious wave of fab construction since the 1980s, but the geographic footprint will be very different. New leading-edge capacity is being built in Arizona, Ohio, Dresden, Tsukuba and southern Korea; mature-node capacity is expanding in mainland China, India and southeast Asia. Each region carries different policy entitlements, different talent constraints, and different costs of capital. The companies that match the right node to the right region will outperform.
Talent is the binding constraint
The bottleneck for new fabs is increasingly not silicon, but skilled engineers and qualified technicians. The cost of senior process engineers in the US has more than doubled since 2020. Companies are responding with multi-year academic partnerships, internal apprenticeship programs, and aggressive international relocation. Without solving the talent equation, the fiscal subsidies do not produce the capacity they were intended to underwrite.
Export controls as a permanent feature
Export controls have moved from episodic instruments to a structural feature of the industry. The control surface now includes specific lithography tools, foundational AI chips, EDA software, advanced packaging equipment and specific design IP. Compliance has become a strategic capability, not a back-office function. Companies are investing in licence-tracking systems, scenario planning for control updates, and government-affairs functions with operational authority.
AI demand is rewriting the demand curve
Generative AI is producing the first sustained, structurally elevated demand cycle in semiconductors since the smartphone era. The implications travel far beyond the AI accelerator itself: high-bandwidth memory, advanced packaging, networking, optical interconnects and specialised power delivery are all running ahead of historical demand baselines. Suppliers in adjacent categories are receiving multi-year capacity commitments they had stopped expecting.
Mature nodes are not the boring story
The strategic conversation is dominated by leading-edge nodes, but mature nodes — 28nm and above — represent the majority of automotive, industrial and consumer demand. Capacity additions in this segment are concentrated in regions that present long-term geopolitical questions for buyers. Industrial customers are quietly diversifying their mature-node sourcing, which is changing the supplier landscape in ways that will be visible in the second half of the decade.
Capital intensity is rising
Industry capex is at all-time highs and rising. Self-funding the leading edge is becoming impossible without subsidy support, partner co-investment and strategic customer prepayments. The financial structures of the next generation of fabs will look more like infrastructure assets than traditional manufacturing assets, with implications for cost of capital, return profile and the appetite of public-market investors.
What boards should ask
For boards of chipmakers and major buyers alike, three questions deserve attention. Are we building a capability to navigate export controls, or are we still treating compliance as a legal cost centre? Have we underwritten the talent plan that supports the capex plan? And do we have a coherent view of how AI demand changes our position over a five-year horizon, or are we extrapolating last year?
A CEO action list
Build a single integrated demand model that combines AI, automotive, industrial and consumer segments — most companies still run them separately and miss the interactions. Treat government affairs as an operating capability with executive presence. Stress-test the next two capex decisions against three plausible export-control scenarios, not one. And, for buyers, design supply contracts that share risk and reward with the manufacturer, because the spot market for any node that matters is going to be small for the foreseeable future.