Nvidia is poised to regain a foothold in the world’s largest AI market after securing U.S. government approval to sell its high-end H200 chips to China.
The move marks a significant reversal from Washington’s earlier blanket restrictions that barred Nvidia from shipping advanced AI semiconductors to the country. Yet the question now is not whether Nvidia can sell to China, but whether Beijing will let its companies buy.
Under the new arrangement, Nvidia can ship the H200 to “approved customers,” but must hand over 25% of revenue from those sales to the U.S. government. The company had already run up against roadblocks earlier this year, which forced it to design the lower-performance H20 chip specifically to meet export rules.
Reports soon suggested Beijing discouraged local firms from purchasing the H20, and industry watchers have treated the H200 with similar skepticism. The Financial Times reported China would “limit access” to the chip, citing unnamed sources. Earlier this year, Beijing expressed security concerns over Nvidia’s H20, following Washington’s approval for the chip’s supply.
Looking at comments from Nvidia CEO Jensen Huang and tracking Beijing’s behavior over the past year provides hints about how this latest approval might play out.
A political green light from Washington does not guarantee enthusiasm in Beijing. Huang had earlier this month said he was still unsure whether Beijing would even allow Chinese firms to buy the company’s proposed H200 artificial intelligence chips, even if Washington greenlights sales.
China’s reasons to hold off on the H200
On paper, the H200 is one of Nvidia’s most advanced AI accelerators, widely used for training state-of-the-art models. But China has spent years trying to reduce its dependence on American technology, pouring resources into domestic semiconductor programs that can eventually rival Nvidia’s performance.
Neil Shah, partner at Counterpoint Research, told CNBC the U.S. approval may reopen access to American chips, but it does little to change China’s strategic direction.
“The strategic train has already left the station,” he said.
Huang himself has helped set expectations. In a May interview with Bloomberg, he described Huawei’s semiconductor lineup as “probably comparable” to the H200 — a notable endorsement for a company Washington has tried for years to cripple. Huawei’s Ascend chips, once seen as niche competitors, are now being deployed in massive clusters as the company seeks to close the gap in total compute capacity.
In June, Huang told CNBC that if Nvidia were cut off permanently from China, Huawei would eventually meet the country’s chip needs entirely.
China’s technology giants, Alibaba, Tencent, and Baidu, have also been using stockpiled Nvidia chips purchased before restrictions took effect. Combined with accelerating advances in local AI silicon, these companies have been able to train competitive models without depending on new U.S. hardware.
Shah warned that relying heavily on Nvidia carries political risks for Chinese companies. Being “locked in” to an American supply chain, he said, is a “liability with a hanging sword of political uncertainty.” Beijing’s national strategy is built on the opposite idea: reducing exposure to foreign pressure.
For that reason, domestic self-sufficiency remains the overriding priority.
Why China may still want the H200 — at least for now
Despite the political calculus, the H200 remains far more capable than local alternatives, and shortages across China’s semiconductor ecosystem could create immediate demand.
Trump said Chinese President Xi Jinping “responded positively” to the export approval. Meanwhile, Alibaba CEO Eddie Wu recently pointed to supply constraints across the entire chip supply chain — a shortage that stretches from training-grade GPUs to networking components.
Ben Barringer, head of technology research at Quilter Cheviot, said China’s tech firms are likely to buy the H200 simply because the alternative is slower progress.
“There will be demand for H200 as it is a better chip than H20 and there is a shortage of chips in China,” he told CNBC. “The big Chinese tech companies will want to use Nvidia and AMD if possible.”
China may be catching up, but it is still behind.
The country remains unable to manufacture top-tier semiconductors at scale, largely due to U.S. export restrictions that block access to the lithography tools required to produce cutting-edge chips. Even Huawei’s newest products fall short in power efficiency and peak performance when compared to Nvidia’s H200 and AMD’s MI300 series.
Shah said that ramping up domestic systems that can match Nvidia’s efficiency remains “elusive,” making it a costly and time-consuming alternative. “The gap between Nvidia, AMD and Huawei and others is still quite wide,” he said.
For now, the H200 offers an immediate boost that domestic silicon cannot match.
The long game still belongs to Beijing
Even if Chinese firms buy the H200 in the near term, analysts agree that Beijing will not deviate from its self-reliance strategy.
George Chen of The Asia Group said Jensen Huang has a “time window” to sell the H200 in China, but it will not last indefinitely.
“Xi will not be foolish that today Trump can sell H200, and then China just totally relies on U.S. chips,” he said.
China’s leading companies — Huawei, Alibaba, and Baidu — remain essential to the country’s long-term ambition: winning the global AI race without depending on the United States.
The approval gives Nvidia a renewed opportunity to generate revenue from the world’s second-largest AI market. But the combination of political uncertainty, rising domestic alternatives, and Beijing’s long-term industrial strategy means the H200 will be competing not just with local chips — but with the entire direction of China’s semiconductor future.
If Chinese firms buy the H200 now, it may be because they need it — not because Beijing plans to make room for U.S. chips in the years ahead.
