OpenAI CEO Sam Altman prefers having a dominant TSMC over Intel's foundry ambitions on AI chip production

OpenAI CEO Sam Altman has a blunt message for the chip industry: he’d rather see TSMC expand capacity than depend on Intel’s foundry push. Altman’s remark, “I would like TSMC to just build more capacity,” was highlighted in a recent interview by Stratechery, underscoring how buyer urgency is reshaping supply-chain priorities.

The reasoning is straightforward: TSMC currently leads at the advanced process nodes that high-end AI accelerators need; its fabs and yield maturity make it the quickest route to more chips. Intel has been pitching onshoring and its own foundry roadmap (including the 18A node), but scaling competitive capacity and reaching reliable yields takes time, time many AI firms don’t have as model sizes and compute demands balloon. Altman’s public nudge is effectively a market signal: when the largest buyers ask for more wafers, suppliers and policymakers tend to listen.

There’s a geopolitical wrinkle, too.TSMC’s capacity is concentrated largely in Taiwan, and while that concentration gives buyers fast access to cutting-edge silicon, it raises strategic and resilience concerns for governments and enterprises that prefer diversified supply chains.

Sam Altman’s comment can therefore be read two ways: a pragmatic short-term push for immediate capacity, and an implicit admission that diversification (including onshore options) is a longer-term project. Reuters coverage of Altman’s regional discussions framed this as part of a wider industry scramble for chip capacity and investment.

Why this matters for products: training and serving modern large models requires vast quantities of accelerators built on the latest nodes. If foundry capacity lags, companies face delayed rollouts, higher cloud costs, and throttled innovation cycles. OpenAI pressing publicly for TSMC to expand is less an endorsement of monopoly than a practical plea for throughput, and it’ll likely prompt more private deals, investor interest, and government attention to wafer supply.

Intel remains a crucial part of the long game: its foundry ambitions aim at resilience, localization, and alternative capacity. But catching TSMC at the bleeding edge is a heavy lift. Expect a mixed industry response where TSMC helps shoulder immediate high-performance demand while Intel and other players scale up as part of a broader diversification strategy. In short, Altman’s comment captures a defining tension in today’s AI era: insatiable compute demand versus the slow, capital-heavy reality of building chips.