Nvidia gets green light to sell H20 AI chips in China again

Nvidia has received approval to resume shipments of its modified H20 AI chips to China, signaling a cautious win for the company in one of the world’s most critical AI hardware markets.

The chips, designed to comply with U.S. export controls, had previously been in limbo due to regulatory bottlenecks and geopolitical tensions.

This move offers Nvidia a much-needed pathway back into China’s AI supply chain, even as broader tech decoupling between the U.S. and China continues to accelerate.

As Nvidia re-enters the Chinese market with a tweaked AI chip, the global tech balance is shifting once again. But this isn’t just about one product; it’s about what comes next for U.S. policy, Chinese innovation, and Nvidia’s global strategy.

Read on to understand how this move could reshape the AI race in ways that matter far beyond silicon.

Why were the H20 Chips blocked in the first place?

The H20 chip was created in response to U.S. government restrictions initially expanded in mid‑2023, and by April 2025, the Trump administration imposed a formal ban on exports of high‑performance GPUs such as A100 and H100 to China.

In an effort to comply, Nvidia developed weaker variants such as the H20. According to a report by TopMostAds, Nvidia scaled back specific features of the H20 chip, including its interconnect bandwidth, to meet the export control thresholds set by the U.S. Bureau of Industry and Security, making it eligible for sale in China under current regulations.

Still, approval for the H20 was delayed earlier this year, as Chinese officials moved cautiously amid escalating tech tensions.

China remains a crucial market despite geopolitical tensions

Despite tightening export controls and rising political friction, China remains one of Nvidia’s most vital markets, especially in the AI and data center sectors, even though Nvidia doesn’t reveal country-level breakdowns.

Despite Nvidia not disclosing country-level revenue, CEO Jensen Huang’s stark warning that the U.S. export ban on H20 chips could cost the company up to $15 billion in lost sales underscores just how critical China remains to Nvidia’s business.

Calling the loss of access to China’s $50 billion AI chip market “deeply painful,” Huang made clear that geopolitical tensions aside, China continues to be one of Nvidia’s most indispensable markets.

Major Chinese tech giants like Alibaba, Baidu, Tencent, and ByteDance are all racing to expand their AI infrastructure to support large language models, generative AI tools, and cloud services.

Despite efforts to localize hardware supply chains, these firms still heavily rely on Nvidia’s chips due to their proven performance, mature developer ecosystem, and established CUDA software stack, which many domestic alternatives struggle to replicate.

Even under constraints, Nvidia’s modified H20 remains attractive to Chinese buyers. Its reliability and compatibility with existing data center environments make it a safer and more cost-effective choice than investing in untested domestic chips.

What makes the H20 different from Nvidia’s other AI chips

The H20 isn’t just a weaker chip; it’s a carefully redesigned product. Compared to the H100, it features.

• Significantly reduced NVLink bandwidth for slower chip-to-chip communication
  
  
• Lower floating-point performance for training large models
  
  
• A focus on AI inference tasks, which require less raw power than training
  

These adjustments were made to pass under the U.S. export radar, allowing Nvidia to maintain a presence in China without violating federal controls.

According to TechNode, which cites Nikkei Asia, Chinese tech giants Alibaba, Tencent, and ByteDance placed orders totaling $12 billion for Nvidia’s H20 chips, and several billion dollars’ worth were already shipped before the April export curbs

Curious how Nvidia pulled this off? Watch this quick breakdown by Industry Insights to understand why being able to sell AI chips in China matters:

Strategic Implications for Nvidia and the Global AI Race

Nvidia’s ability to reenter China has several strategic consequences.

First, it gives the company a short-term revenue cushion. With the global chip market experiencing a slight cooldown and competition rising from firms like AMD and Intel, the Chinese market provides an essential buffer.

Second, the approval prevents Chinese clients from fully pivoting to domestic options like Huawei’s Ascend series or Alibaba’s Hanguang processors. That’s crucial for Nvidia to avoid long-term displacement.

Finally, the move helps Nvidia walk a delicate line, staying on good terms with Washington regulators while still participating in one of the fastest-growing AI markets in the world.

What this means for U.S. chip export policy

Nvidia’s ability to sell the H20 in China again points to a broader issue with how U.S. chip export controls are structured. While the rules aim to block China’s access to cutting-edge AI hardware, companies like Nvidia can often redesign chips just enough to meet compliance, without losing their appeal in overseas markets.

This kind of workaround shows how hardware restrictions can be outpaced by fast-moving innovation. Instead of halting progress, the rules end up encouraging minor technical adjustments that preserve market access while weakening the intended pressure.