The Headline Number Is Real. The China Access Is Not.

Nvidia CEO Jensen Huang landed in Taipei on Saturday and told reporters that his $200 billion CPU market forecast includes China. When asked directly, he said: "I would think so."

According to Reuters, the U.S. government has cleared approximately 10 Chinese companies to purchase Nvidia's H200 chip — the company's second-most powerful AI chip. Not a single H200 has been delivered. Zero.

Two Walls Standing Between Nvidia and China Revenue

Huang was blunt about the situation at Taipei's Songshan airport: "H200 has been licensed to ship to China. It would be terrific to be able to serve that market. The Chinese market is very important. It's very large, of course."

U.S. export controls have blocked Nvidia's top-tier chips — including the flagship Blackwell architecture — from entering China. The H200 is a second-tier option, already behind the cutting edge.

China hasn't approved the sales either. According to CNBC, Chinese officials have NOT cleared the H200 purchases. Beijing is deliberately fostering its own domestic chip suppliers like Huawei and CXMT. This isn't just Washington blocking sales — Beijing is also saying no.

President Trump traveled to Beijing earlier this month and met with President Xi Jinping. Huang was reportedly part of that U.S. delegation. The talks produced no breakthrough on H200 sales, according to Reuters.

The CEO of the world's most valuable company personally accompanied the U.S. president to Beijing, and nothing came of it.

What Huang Is Actually Selling Investors

On Wednesday's earnings call, Huang told investors that Nvidia's new "Vera" central processors open up a brand new $200 billion CPU market — entirely separate from the GPU market that made Nvidia a multi-trillion-dollar company.

GPUs train large AI models. CPUs run agentic AI — systems that operate autonomously, make decisions, and execute tasks in the real world. Agentic AI is the next wave, and it needs a different kind of chip.

Nvidia's Vera CPU, part of the broader Vera Rubin platform, is the company's bet on that market. Huang told reporters in Taipei he is ramping production of the Vera Rubin platform now.

He also confirmed Nvidia will meet with TSMC — the Taiwan Semiconductor Manufacturing Company — during this visit. TSMC makes the advanced semiconductors that power Nvidia's entire product line. That relationship is the backbone of the AI chip industry.

The Taiwan Play

Huang arrived in Taipei ahead of next month's Computex trade show. AMD, Nvidia's chief competitor, announced Thursday it would invest more than $10 billion in Taiwan's AI sector to expand chip assembly capacity, according to Reuters.

When asked whether Nvidia planned similar Taiwan investments, Huang didn't announce a number — but he was pointed: "We haven't announced anything in the past, but we've invested in and supported our partners here far more than that."

Translation: Nvidia's existing commitment to Taiwan's supply chain likely dwarfs AMD's $10 billion pledge. He just didn't name the figure.

What the Numbers Don't Show

Most headlines treat "Nvidia sees China in its $200 billion CPU forecast" as near-term revenue.

It's not. The gap between Huang's market vision and what's actually happening on the ground is massive. H200 licenses exist on paper. No chips are moving. Beijing is blocking sales just as much as Washington's export controls are complicating them. Trump's Beijing summit didn't crack the door open. China is actively funding its own chip alternatives to reduce dependence on American technology entirely.

Nvidia counting China in a $200 billion TAM estimate signals long-term ambition to investors. But treating it as near-term revenue is a serious mistake.

The China Problem Is Structural, Not Temporary

This isn't a paperwork delay. China is making a deliberate industrial policy choice to build semiconductor independence. Huawei's Ascend chips, while behind Nvidia's best offerings, are improving. Beijing has financial and strategic incentives to keep buying domestic, regardless of what licenses Washington issues.

Nvidia needs China. China is the world's largest single AI market. Losing sustained access to it is a real long-term threat to the company's growth trajectory — no matter how impressive the Vera Rubin platform looks on a spec sheet.

Huang is in Taiwan meeting with TSMC and preparing to launch next-generation chips. The technology is extraordinary. The geopolitics are a wall.

Until one H200 actually ships to a Chinese data center, the $200 billion China opportunity remains a forecast.