Stakes and timeline
America’s chip problem is not abstract. It is about scale, time, and people. The 2027 risk that conflict disrupts Taiwan’s foundries is real, and our economy and military lean on those factories for the most advanced logic. Taiwan and South Korea still host almost all leading-edge capacity. If those lights blink, everything from data centers to defense programs feels it. That is why this must be a long game, not a one-off splash of funding.
What Washington has done
Both administrations moved the ball. The Biden team stood up CHIPS and began writing big checks. TSMC Arizona secured up to $6.6B. Samsung Texas won about $6.4B. Micron lined up roughly $6.1B. Intel received up to $8.5B in grants and up to $11B in loans, plus a separate national-security Secure Enclave award of up to $3B to make trusted chips for the government. The Trump administration then took an extraordinary step in Aug 2025. The United States bought a 9.9% equity stake in Intel. You do that when a capability is so strategic that market risk should not decide if it lives or dies. We have used versions of this play before, from the GM and AIG rescues to World War II’s Defense Plant Corporation, which financed and owned critical factories and leased them to industry.
The denominator and why one-offs are not enough
Global fab capacity is roughly 33.6M 200-mm-equivalent wafers per month. Taiwan alone runs near 5.8M wpm. TSMC plans $38B to $42B of capex this year. One firm’s annual spending in Taiwan sits in the same neighborhood as several years of U.S. grants. That is why single appropriations will not close the gap. We need steady, multi-year flows that match the pace of the competition and keep projects on schedule through tool installs, yield learning, and full production.
The quiet workhorses at home
Two U.S. manufacturers deserve more airtime because they shoulder essential work even if they are not chasing the tiniest transistors. GlobalFoundries won up to $1.5B to expand in New York and modernize Vermont. It makes the “essential chips” that keep autos, radios, power electronics, aerospace, and defense systems humming. onsemi acquired the East Fishkill 300-mm site and is turning it into a powerhouse for power and sensing devices, including silicon carbide for EVs and energy infrastructure. They are not always in the headlines, but they carry a big load for supply resilience and defense.
Defense demand in plain English
Defense demand is small in volume and large in consequence. Government end use is about 1% of global chip demand. As of 2021, only about 2% of the devices inside U.S. military systems were built at domestic “trusted” foundries, with the rest sourced from commercial supply chains. Apply those guideposts to today’s wafer base and you get a clear picture. Less than 1% of 33.6M wpm implies on the order of a few hundred thousand wafers per month globally support defense needs. Roughly 2% of that would be produced at trusted U.S. sites, which works out to about 7,000 wpm domestically, with the remainder purchased from commercial lines, often abroad. For the most advanced compute, only a few allied countries manufacture at that level today, led by Taiwan and South Korea.
Talent is the hinge
The fastest way to close risk is people. We have done this before. In the 1940s, refugee scientists like Leó Szilárd, Hans Bethe, and John von Neumann helped America build a nuclear deterrent and modern computing. After the war, Wernher von Braun’s team powered Saturn V. Talent changed outcomes. It can again if we recruit the world’s best device physicists, process engineers, and materials scientists into U.S. fabs and labs. That means targeted visas that move in weeks, not years, for chip specialists and their families. The industry will need about 115,000 more workers by 2030 and risks a shortfall of roughly 67,000 without action. Pair fast visas with compensation that signals urgency. In AI, nine-figure packages have been offered to top researchers and a large strategic stake in Scale AI has been reported at roughly $14B to $15B. We do not need to mirror every number, but we should be competitive for the best lithography, power, and materials talent on earth.
Stay candid about the calendar
TSMC’s U.S. build is moving, but most of its cutting-edge output will remain in Taiwan for years. Intel’s Ohio timeline has slipped toward the decade’s end. Samsung’s Texas ramp is significant but still coming online. That is why policy should be a steady march. Keep CHIPS awards tied to milestones. Use equity and guaranteed offtake where market risk is too high. Recruit global talent at speed and scale. Back the full stack, from onsemi’s power devices and GlobalFoundries’ RF and automotive chips to the leading-edge logic that powers AI.
Bottom line
Wafers and chips are just the units. What matters is capacity that runs in the United States, year after year, with people who know how to push it forward. Biden put scaffolding in place. Trump added ownership and urgency with a 9.9% Intel stake. Taiwan’s factories will keep investing every year. So must we. If we pair long-horizon dollars with smart visas, competitive offers for top scientists, and reliable demand for both leading edge and essential edge, America can build enough wafers, enough chips, and enough resilience to keep the arsenal stocked when it counts.
Leave a Reply