Beijing Freezes All Nvidia H200 Purchases, Costs $30B

A US policy move meant to restrict China's access to advanced AI chips has turned against Nvidia. Beijing has refused to approve any purchases of Nvidia's H200 chips since the Trump administration imposed a 25% tariff on semiconductor exports to China, wiping roughly $30 billion from Nvidia's market value.

The H200 is Nvidia's high-performance AI training chip, a step below the flagship H100 that was already restricted under earlier export controls. The H200 occupied a middle ground: powerful enough to matter for large-scale AI work, but not explicitly banned from sale to China under previous rules. That window appears to have closed.

Beijing's response is a form of economic counter-pressure. Rather than retaliating with a direct export ban of its own, China is using its domestic approval process to freeze purchases. No Chinese buyer, state-owned or private, has received clearance to complete an H200 deal. The effect is a de facto ban without a formal one, and it gives Beijing flexibility to reverse the stance if trade negotiations shift.

Why the tariff backfired

The original logic behind the 25% tariff was to make US chip exports more expensive and less attractive, slowing China's AI build-out. But the mechanism cut both ways. Higher costs on the US side gave Chinese regulators a political justification to block purchases entirely, framing it as a reciprocal trade measure rather than a technology restriction. The result: Nvidia loses revenue, China's AI sector faces delays, but Beijing controls the narrative and the timeline.

Nvidia's China business has been under pressure for over two years. The Biden administration's export control rules in 2022 and 2023 already barred sales of the A100 and H100 chips. Nvidia responded by engineering downgraded versions, including the A800 and H800, specifically tuned to stay within export thresholds. The H200 was the next attempt to maintain a foothold in one of the world's largest AI markets. That strategy now looks stalled.

The $30 billion figure reflects the market's read on how significant China revenues are to Nvidia's growth story. China had historically accounted for a meaningful share of Nvidia's data center sales, and investors price future growth partly on the assumption that some level of access would continue. A full freeze on H200 approvals removes that assumption.

What changes next

The immediate question is whether this is a negotiating position or a durable policy. China has used regulatory approvals as leverage in trade disputes before, and a broader US-China trade framework could reopen the channel. But if tariffs remain or escalate, Beijing has little incentive to approve purchases that hand Washington a win.

For Nvidia, the pressure is to either push for a policy carve-out through diplomatic channels or accelerate sales elsewhere, particularly in markets across Southeast Asia, the Middle East, and Europe where demand for AI infrastructure is growing. CEO Jensen Huang has previously argued that restricting Nvidia's China sales pushes Beijing toward domestic alternatives like Huawei's Ascend chips, which could eventually reduce China's dependence on US suppliers entirely.

That argument now has more urgency. Every quarter Nvidia spends locked out of the Chinese market gives Huawei and other domestic chip designers more time and more incentive to close the performance gap. The longer the freeze holds, the more irreversible the competitive shift becomes.

Investors will be watching Nvidia's next earnings call closely for any revision to revenue guidance tied to China, and any signal from trade talks in Washington about whether semiconductor tariffs are on the table for negotiation.