Nvidia has pulled the plug on manufacturing its H200 AI chips for the Chinese market, redirecting all of that production capacity at contract chipmaker TSMC to its next-generation Vera Rubin hardware, according to a report by the Financial Times on Thursday.

The End of H200 for China

The H200 — Nvidia's second-most advanced AI chip — was supposed to be the company's offering to Chinese customers under U.S. export control rules. But the reality has been far messier than the policy suggested.

In January, the Trump administration formally approved China-bound sales of Nvidia's H200 chips. On paper, that was a green light. In practice, not a single H200 chip has actually been delivered to a Chinese customer. A U.S. Commerce Department official confirmed last month that shipments remained stalled due to guardrails built into the approval process.

Last week, Nvidia said it had received licenses to ship "small amounts" of H200s to China. But the decision to halt production entirely tells a different story: Nvidia doesn't expect any meaningful H200 sales to China in the near term.

Vera Rubin Takes Priority

The Vera Rubin architecture, announced at CES 2026 in January, represents Nvidia's next leap in AI chip design. Already in full production at TSMC on a 3-nanometer process, Vera Rubin is where Nvidia sees its future — and where it wants every available slice of TSMC's manufacturing capacity directed.

TSMC declined to comment on the report. Nvidia did not immediately respond to a request for comment.

The reallocation makes strategic sense. With global demand for AI compute continuing to surge — driven by the training and inference needs of frontier models — Nvidia's most rational move is to dedicate its constrained manufacturing capacity to the highest-margin, highest-demand products. And right now, that's Vera Rubin, not export-controlled H200s stuck in regulatory limbo.

The Broader Chip War

This development is the latest chapter in the escalating U.S.-China semiconductor conflict. The pattern has become familiar:

• The U.S. restricts exports of advanced chips to China
• Nvidia designs compliance-friendly variants (A800, H800, H200)
• Additional regulations and guardrails make those variants impractical
• Nvidia eventually gives up on the China market for that generation

Meanwhile, Chinese companies have been accelerating their own chip development efforts. The new five-year plan released today explicitly claims breakthroughs in "independent R&D of chips," and companies like Huawei continue to push their Ascend AI accelerator lineup.

What It Means for the AI Industry

For the global AI infrastructure buildout, the implications are significant. More TSMC capacity going to Vera Rubin means faster availability of next-gen chips for U.S., European, and allied-market customers — the hyperscalers, cloud providers, and sovereign AI projects that are spending hundreds of billions on data center construction.

For China, it means the window for accessing cutting-edge Nvidia hardware through legitimate channels has effectively closed for this generation. The question becomes whether domestic alternatives — DeepSeek's software efficiency gains, Huawei's Ascend chips, and other homegrown solutions — can fill the gap.

Nvidia's stock has been resilient through these geopolitical tensions, with investors largely pricing in the loss of China revenue in exchange for dominant positioning in the rest of the world's AI infrastructure boom. This latest move further cements that strategic direction.