Posts
For most of American semiconductor history, owning your own fabrication plant wasn't a business decision — it was a matter of pride. When Taiwan launched a new company in 1987 built entirely around the opposite idea — a "pure-play foundry" that would manufacture chips for other companies' designs but never sell a chip of its own — the American industry's reaction was closer to mockery than concern. AMD's own CEO at the time is remembered for a line that aged badly: real men have fabs. Within two decades, AMD itself would spin off its manufacturing arm entirely and become a fabless company, following the exact model it once dismissed.
That reversal, repeated across the industry through the 1990s and 2000s, is why American semiconductor fabrication capacity fell from roughly 40 percent of the world's total in 1990 to around 12 percent by 2020. It wasn't offshored in the way a factory gets physically relocated. It was out-competed by a model nobody domestic wanted to build in time — and once the fabless approach proved cheaper, capital simply stopped flowing toward owning a fab at all.
Real Progress, Real Limits
Unlike ships or shells, this is a chapter where meaningful correction is already underway, and it deserves to be stated plainly rather than folded into a story that only goes one direction. TSMC — the same Taiwanese company that started this whole shift in 1987 — has committed more than $165 billion to a cluster of fabs in Phoenix, Arizona, the largest foreign direct investment project in American history. The first fab is already in volume production, making 4-nanometer chips for customers including Apple and Nvidia, the first time TSMC has produced its most cutting-edge silicon anywhere outside Taiwan.
The honest caveat sits right next to the achievement. Arizona's current output represents roughly 2 percent of TSMC's total global wafer capacity. Industry forecasts suggest total American advanced semiconductor capacity will roughly triple by 2028 — real, measurable progress — but for most of what that capacity will actually produce, it remains, as one procurement analysis put it, a story that pays off between 2027 and 2029, not one that's paying off today.
The Company, Not Just the Capital
Here's where this chapter rhymes with the last two. Fixing the shell shortage required importing Turkish production tooling, because the domestic know-how to build a modern forging line fast had atrophied to nothing. Fixing the shipyard bottleneck required a presidential memorandum inviting foreign builders to do the work American yards no longer could. And fixing the chip shortage required something even more direct: not tooling, not a policy exception, but the actual company. The only way to get leading-edge fabrication onto American soil at meaningful scale, in any reasonable timeframe, was to pay the Taiwanese firm that had spent thirty-seven years building the expertise nobody in America had bothered to keep.
Institutional knowledge cannot be purchased and installed on the same timeline as capital. It has to be imported wholesale, as a company, because it was never rebuilt at home in the first place.
Money alone didn't do this. The CHIPS Act put tens of billions of dollars behind the effort, and it helped — but capital can be wired anywhere in an afternoon. The Reagan-era version of this story would have called it a triumph of free trade — everyone doing what they do best. Thirty years later, sitting inside a supply chain crunch and a live strategic rivalry, it reads differently: proof that insulation from competition doesn't just cost market share. It costs the knowledge of how to compete at all, and that particular loss doesn't come back just because someone finally decided to write the check.
Next: the mineral that isn't rare, mined mostly somewhere else, refined almost nowhere here.
To Whom It May Concern —
This piece was produced through a collaboration between a human author and an AI system (Claude, made by Anthropic). The research, structure, and editorial judgment are a joint effort; errors, once found, are corrected openly rather than quietly revised away. We think that collaboration is worth being honest about, so we are.
No comments:
Post a Comment