The artificial intelligence revolution has made Nvidia (NVDA) one of the most valuable firms in the world.
But for its chief executive, Jensen Huang, the next phase of growth may hinge on something outside the company’s control: whether Washington and Beijing can agree on the rules of the AI chip trade.
Reuters reported that the U.S. has approved 10 Chinese companies to buy Nvidia’s H200 artificial intelligence chips. The approved purchasers include Alibaba (BABA), Tencent, ByteDance and JD.com, they said. Lenovo and Foxconn are licensed as distributors.
That sounds like a major breakthrough.
But there is a catch: No H200 deliveries have been made, leaving a potentially significant China revenue opportunity stuck between U.S. export rules and Beijing’s push to reduce dependence on foreign technology.
The announcement comes as Huang accompanied President Donald Trump on a trip to Beijing, sparking optimism that Nvidia could finally unlock a blocked China semiconductor contract.
“Any deal that allows Nvidia to sell more chips to China means fewer Nvidia chips for U.S. firms, and a smaller U.S. lead in AI over China,” Chris McGuire, senior fellow for China and emerging technologies at the Council on Foreign Relations, told Reuters.
Nvidia’s China opportunity is too big to ignore
Before the U.S. tightened its export limits, Nvidia held almost 95% of the advanced chip market in China. China was formerly responsible for 13% of Nvidia’s revenue and Huang has estimated the country’s AI business might be worth $50 billion this year.
That’s why the H200 clearance is important.
Even with no significant income from data centers in China, Nvidia has managed to grow at a spectacular speed. The company forecasted $78 billion in fiscal first-quarter sales and noted that forecast anticipates no data center compute revenue from China.
That makes China less about survival and more about upside.
In other words, any real H200 exports to China may be incremental upside rather than revenue that Wall Street is already pricing in.
Related: Nvidia CEO shocks AI community over the one thing he didn’t do
The H200 is not a low-end workaround either.
The Nvidia H200 has 141 gigabytes of GPU memory and 4.8 terabytes per second of memory bandwidth, making it a formidable chip for memory-intensive AI tasks, Nvidia said.
H200 processors could help Chinese cloud businesses and AI developers bridge performance gaps. For Nvidia, that might open up a market that formerly defined its global supremacy.
Key numbers behind Nvidia’s China chip fight
- 95%: Nvidia’s reported share of China’s advanced chip market before tighter U.S. restrictions.
- 13%: China’s former share of Nvidia revenue.
- $50 billion: Huang’s reported estimate for China’s AI market this year.
- Around 10: Chinese firms reportedly approved to buy H200 chips.
- 75,000: Maximum H200 chips each approved customer can reportedly purchase.
- $78 billion: Nvidia’s fiscal first-quarter revenue guidance.
- 141GB: H200 GPU memory.
- 4.8TB/s: H200 memory bandwidth.
Photo by Bloomberg on Getty Images
Nvidia still faces a China roadblock
The biggest problem for Nvidia is that approval does not equal sales.
Reuters claimed that Chinese purchasers had retreated afterBeijing’s guidance to strengthen surveillance of foreign technology dependencies.
That hesitancy fits into China’s bigger agenda.
Beijing encourages local firms to buy more Chinese AI chips including technology from Huawei. The aim is not only commercial. It’s a tactical maneuver. The more China depends on Nvidia, the more vulnerable its AI sector becomes to any future U.S. export restrictions.
The U.S. side is tricky too.
More AI:
- Micron sits at the center of a red-hot chip rally
- IBM CEO sends blunt message on AI and quantum computing
- Anthropic CEO makes shocking admission about AI
Reuters claimed that US guidelines require Chinese bidders to exhibit adequate security processes and to verify that the chips will not be used for military reasons. The Trump administration also secured a deal that would allow the U.S. to take 25% of the revenue from approved semiconductor sales, with chips needing to pass through U.S. territory en route to China.
That structure creates political risk in both countries.
Sales to Nvidia might help China catch up to the U.S. in AI, hardliners in Washington say. Beijing officials seem unwilling to depend on U.S.-approved chips that transit U.S. territory before delivery.
This puts Nvidia in a weird situation.
The product is at the company. It has the customer pull. It is U.S.-cleared. But it still doesn’t have finished shipping.
For investors, this means that the H200 narrative is a China bet, not a fixed revenue stream.
U.S. cloud giants’ huge AI infrastructure spending continues to drive the core business of Nvidia. But if Huang can turn approvals into actual deliveries, China can add another layer of upside to an already huge growth narrative.
The danger is that the window will close before Nvidia can exploit it.
China is pushing ahead with its drive for homegrown processors, and every month of delay offers Huawei and other local vendors more runway to gain credibility with Chinese AI clients.
This is why Huang’s journey to Beijing matters.
Nvidia needs more than Washington’s blessing. Beijing should be confident.
Until then, Nvidia’s China success is just that: a breakthrough on paper, not yet in revenue.
Related: Nvidia’s AI dominance faces uncomfortable new reality