The AI supply chain has a capital. Nvidia just made it official
From US$10 billion to US$150 billion in under five years, Nvidia's Taiwan spending reflects how completely AI has reordered the economics of the global semiconductor supply chain, and how deeply the industry has staked itself to a single geography.
Five years ago, Nvidia was spending US$10 to US$15 billion a year in Taiwan. On Wednesday, Jensen Huang announced that figure is heading to US$150 billion annually. The jump tells you everything about what AI has done to the global semiconductor supply chain, and who sits at its center.
The announcement came at Nvidia's Constellation All-Employee Celebration in Taipei, where Huang also confirmed the company is building a Taiwan headquarters, expected to break ground this year and be operational by 2030, employing 4,000 people. Nvidia, which became the first company to reach US$5 trillion in market value late last year, is now physically planting its flag in the country that makes its most critical products possible.
Why Taiwan, and why now
The answer is not complicated, but its implications are. Taiwan is home to TSMC, the world's largest contract chipmaker and Nvidia's most important manufacturing partner for the advanced semiconductors powering the AI infrastructure build-out. It is also home to Foxconn, Wistron, and Quanta Computer, the manufacturers assembling the AI servers that data centers globally are racing to deploy.
For Nvidia, proximity to this ecosystem is not a preference. It is an operational necessity.
AMD underscored the point last week, announcing more than US$10 billion in Taiwan AI sector investment to expand its capacity to build and assemble advanced chips. The pattern is consistent: every major player in the AI compute stack is moving closer to the island that manufactures it.
Nvidia's new headquarters accelerates that logic, tightening the feedback loop between chip design, production, packaging, and system assembly, the loop that has become the defining competitive advantage in AI infrastructure.
The geography problem
What Huang's announcement makes vivid is a concentration the industry has largely chosen not to examine too closely. The global AI build-out runs through a remarkably small number of companies in a remarkably small geography. TSMC's dominance in advanced logic. SK Hynix and Samsung's grip on high-bandwidth memory.
Taiwan's ODMs controlling AI server assembly. And now Nvidia, the company whose chips sit at the center of all of it, formally anchoring itself to the same island. That is an extraordinary degree of dependency for an industry that underpins everything from financial services to national defense. The bigger Nvidia gets, the more consequential Taiwan's position becomes, and the more exposed the entire AI supply chain is to whatever disrupts it.
It is also worth noting that Huang accompanied US President Donald Trump on his delegation to Beijing earlier this month for a summit with Chinese President Xi Jinping. The US$150 billion Taiwan commitment, announced days later, lands in that context whether or not it was intended to.
What the number doesn't say
One detail missing from Wednesday's announcement: Huang did not specify how many years the US$150 billion annual commitment covers. For a pledge of this magnitude, that is not a minor omission. It is the difference between a sustained strategic realignment and a headline figure attached to a building opening.
The trajectory from US$10 billion to US$150 billion is real and verifiable. Whether the ceiling holds is a different question. What is not in doubt is the direction. Taiwan has spent decades as the indispensable but often underleveraged backbone of the global technology industry. AI changed that calculus completely.
Nvidia's US$150 billion annual commitment is the clearest signal yet that the industry knows exactly where its center of gravity lies.