Beijing just dealt a seismic shift to the US-China tech cold war. While Nvidia CEO Jensen Huang toured Shanghai restaurants and fruit markets this week, Chinese regulators quietly approved the sale of over 400,000 H200 AI chips to the country's tech giants—a dramatic reversal of Biden-era export controls that had banned the sale of advanced semiconductors. The move marks the culmination of Huang's year-long lobbying campaign in Washington and signals that Trump's White House is betting on a radically different strategy: keeping China dependent on American chips rather than locking them out entirely.
Nvidia CEO Jensen Huang looked remarkably relaxed this week, biking through Shanghai and sampling beef hot pot in Shenzhen. He had every reason to be upbeat. While he was making the rounds, Beijing dropped a bombshell that could reshape the global AI chip market for years to come.
China approved the sale of hundreds of thousands of Nvidia's powerful H200 AI chips to major Chinese tech companies, according to reports from Reuters and the Wall Street Journal. The approvals cover more than 400,000 chips destined for ByteDance, Alibaba, and Tencent under conditional licenses granted during Huang's visit, with more approvals expected in coming weeks.
The timing isn't coincidental. This represents the payoff from Huang's aggressive lobbying campaign in Washington over the past year, and it marks a stunning reversal of American tech policy toward China. Under the Biden administration, the US sharply tightened export controls on high-end AI chips and explicitly barred models like the H200 from being sold to Chinese customers. The restrictions aimed to prevent Beijing from developing powerful AI systems with military applications or other sensitive uses.
But President Trump's White House has embraced a fundamentally different calculation, one championed by Huang and AI czar David Sacks. Their argument goes like this: allowing China controlled access to some American AI chips beats ceding the entire market to Chinese chipmakers like Huawei. The strategy promises economic benefits while theoretically keeping Chinese firms dependent on US technology.
In recent internal discussions, White House officials have pointed to another justification: the continued smuggling of advanced chips into China proves US restrictions have been ineffective anyway, according to two people familiar with the matter. Officials now argue that allowing limited, regulated sales gives authorities better visibility into where chips ultimately end up than an opaque gray market ever could.
The White House didn't immediately respond to requests for comment, and neither did Nvidia or the Chinese tech companies involved.
But this isn't just a win for Huang and Trump's team. Beijing appears to be playing a sophisticated double game. By allowing domestic companies to purchase H200 chips in limited quantities, China achieves two strategic goals simultaneously, says Samuel Bresnick, a research fellow at Georgetown's Center for Security and Emerging Technology.
China's tech champions get the compute power they desperately need to train frontier AI models competitive with OpenAI and other American labs. Yet by keeping tight control over who receives Nvidia hardware, Beijing ensures demand for Huawei chips remains robust and companies still have strong incentives to build out China's domestic semiconductor ecosystem.
"I see this as excellent evidence that this David Sacks idea of keeping China hooked on American technology is just not how this is going to go," Bresnick told Wired. "This is proof that China is totally uncomfortable with the idea of letting its own burgeoning chip industry be swamped by Nvidia."
The Georgetown researcher warns the real damage stems from Washington's policy whiplash. For years, policymakers have sent contradictory signals about what the US wants to accomplish with chip controls. "The worst possible thing we can do is just go back and forth," says Bresnick. "We have already given China the imperative to get their own chips going while also giving them access at the same time."
The H200 represents a significant upgrade over previous-generation chips that Chinese companies could access through workarounds. With 141GB of high-bandwidth memory and specialized architecture for training large language models, the H200 can dramatically accelerate AI development timelines. The 400,000-chip approval could support training runs for multiple frontier models simultaneously.
For Nvidia, the stakes couldn't be higher. China represented a massive market before Biden-era restrictions, and the company's lobbying efforts reflected just how much revenue was at risk. The reversal reopens a multibillion-dollar opportunity, though the conditional licensing arrangement means Beijing retains significant control over the tap.
The approval comes as global competition in AI chips intensifies. AMD has been making inroads with data center customers, while startups like Cerebras and Groq target specific AI workloads. But Nvidia's CUDA software ecosystem and years of optimization for AI training have created powerful network effects that make switching costly.
Industry observers note the policy shift also affects semiconductor equipment makers like ASML and Applied Materials, whose extreme ultraviolet lithography machines and other tools face their own export restrictions. The new approach to AI chips could signal broader changes in how the Trump administration approaches technology exports to China.
What happens next remains unclear. Beijing has approved the first batch of chips but maintains discretion over future allocations. Chinese companies will need to demonstrate compliance with usage restrictions and reporting requirements. And Washington could reverse course again if political winds shift or if intelligence agencies raise new concerns about technology transfer.
For now, though, Huang's bike ride through Shanghai looks like a victory lap. His bet that the economic and strategic logic of engagement would eventually win out over containment appears to be paying off—at least for this round of the US-China tech competition.
Beijing's approval of 400,000+ Nvidia H200 chips doesn't just reopen a multibillion-dollar market—it exposes the fundamental tension in US tech policy toward China. By giving Chinese companies access to advanced AI chips while simultaneously maintaining domestic chip development incentives, Beijing is threading a strategic needle that Washington's back-and-forth approach has made possible. The real question isn't whether this deal benefits Nvidia or helps Chinese AI labs catch up to American frontier models. It's whether the Trump administration's bet on managed engagement will prove more effective than Biden's containment strategy, or whether the policy whiplash has simply handed China the best of both worlds: access to cutting-edge US technology and the motivation to build self-sufficient alternatives. Georgetown's warning rings clear—the worst outcome is oscillating between strategies, and that's exactly what's happening.
