Nvidia Hit With $5.5 Billion Charge As US Ramps Up Restrictions On Chip Exports To China

Chipmaker Nvidia on Tuesday said that it might have to incur a $5.5 billion charge following new restrictions imposed by the US government on the export of its H20 artificial intelligence chips to China.

The H20, designed specifically to comply with earlier American export rules, has become central to Nvidia’s strategy in China—a key market for the company, reported Reuters.

But recent regulatory changes have now put a halt to those plans, raising fresh concerns about Nvidia’s ability to maintain momentum in the global AI chip race.

The H20 was developed as a workaround to earlier US export restrictions, offering reduced training capabilities while remaining strong in inference tasks—where AI models deliver results to users. Inference, now the fastest-growing segment in the AI chip market, has made the H20 highly sought after by Chinese tech giants like Tencent, Alibaba, and ByteDance. According to a Reuters report from February, these companies  significantly ramped up orders of the chip, driven in part by demand from startup DeepSeek.

Also Read : IndusInd Bank Says Mistakes In Derivatives Portfolio To Impact Bank's Net Worth By Rs 1,979 Crore

US Officials Tighten The Reins On Advanced Chips

On April 9, US authorities informed Nvidia that the H20 would now require an export license to be shipped to China. Just days later, on April 14, Nvidia was told the new rules would remain in effect indefinitely. The decision is part of Washington’s broader push to prevent advanced chips from being used in Chinese supercomputers. “The Commerce Department is committed to acting on the President's directive to safeguard our national and economic security,” a department spokesperson said.

While the H20 is not Nvidia’s most powerful chip, its high-speed memory and connectivity features raise red flags among US officials concerned about its potential use in supercomputing. Restrictions on such chips have been in place since 2022. The Institute for Progress, a Washington-based think tank, supported the latest move, stating, “At least one of the buyers, Tencent, has already installed H20s in a facility used to train a large model, very likely in breach of existing controls restricting the usage of chips in supercomputers exceeding certain thresholds. DeepSeek’s supercomputer used to train their V3 model is also likely in breach of the same restrictions.”

The $5.5 billion charge includes costs related to inventory, purchase commitments, and associated reserves for H20 chips. Nvidia declined to elaborate beyond its regulatory filing.

Just a day earlier, the company  announced a plan to build AI servers worth up to $500 billion in the US over the next four years, in collaboration with partners like TSMC. The move aligns with the Trump administration’s push to boost domestic manufacturing of advanced tech infrastructure.

business