Why US Nvidia’s chip sales to China is far more complicated than it looks

Dubai: US President Donald Trump’s decision to allow Nvidia to sell its H200 artificial-intelligence chips to “approved customers” in China is rippling across global tech and geopolitical circles.

On the surface, the move appears to reopen a major market for Nvidia and soften Washington’s hard line on China. In reality, the implications are far murkier — raising fresh questions about national security, China’s appetite for US hardware, and the future trajectory of the US-China technology rivalry.

The approval, first reported by Axios, gives Nvidia access to a market it had been largely blocked from under the Biden administration’s export controls. The H200 is far more capable than the earlier H20 chip designed to comply with those rules — and one that Chinese buyers mostly rejected.

Washington will reportedly take a 25% share of sales, though the mechanics remain unclear. Trump framed the shift as undoing what he called Biden’s “terrible idea” of forcing US companies to build deliberately weakened chips for China.

Yet the move comes at an awkward moment. China is less reliant on Nvidia than it was just two years ago, thanks to a fast-growing domestic semiconductor programme and a sizable stockpile of restricted US chips that entered the country through unofficial channels. That combination makes it uncertain whether Chinese firms will queue up for the newly approved H200 — even if Washington is now willing to allow limited sales.

Here are the six factors shaping the real impact of the policy.

The H200 is significantly stronger than the H20 that China rejected, offering real performance gains for companies training AI systems. But it remains well behind Nvidia’s top-end Blackwell and upcoming Rubin chips, which remain restricted and are central to training frontier-level AI models. Analysts say the H200 is roughly 18 months behind Nvidia’s cutting-edge technology — a gap that limits its usefulness for firms aiming to develop large, complex systems. Nvidia CEO Jensen Huang has openly acknowledged uncertainty about demand, telling Axios he “cannot say” how Chinese buyers will respond.

The Financial Times recently reported that more than $1 billion worth of “banned” Nvidia chips entered China despite export controls. On the same day Trump announced the new approval, US prosecutors charged two businessmen with smuggling more than $160 million in H100 and H200 units into China. This parallel supply weakens the urgency for Chinese firms to buy older or downgraded models through official channels. For US officials, the case has reignited concerns about losing what they describe as America’s “technological edge” in AI.

Chinese regulators are reportedly considering rules requiring companies to justify why domestic chips cannot meet their needs before buying US hardware. This fits Beijing’s push for semiconductor self-sufficiency and its desire to avoid building critical industries on technology vulnerable to future bans. If China enforces strict approval processes, demand for Nvidia’s H200 may stay modest.

The decision has drawn sharp criticism from Democratic lawmakers, who argue it risks strengthening China’s military and cyber capabilities. Senator Elizabeth Warren linked the move to what she described as a “backroom meeting” with Nvidia’s CEO, referencing a previous donation to a White House project. Congress members are calling for oversight hearings.

Nvidia counters that the policy supports American manufacturing while allowing the Commerce Department to vet sensitive sales. The debate captures a deeper divide over whether the US should try to fully block China’s access to advanced AI chips or allow controlled sales that protect US tech firms and limit China to older-generation hardware.

Several researchers argue the policy does little to change the overall AI race. China’s domestic chip push is accelerating, it already has large quantities of Nvidia GPUs, and the H200 is not competitive with the frontier-grade chips powering the most advanced models today. Alex Stapp of the Institute for Progress called the decision a “massive own goal,” noting the H200 is much more capable than the previously allowed H20. But others say that even with that performance jump, the chip is still behind the curve. Chinese analysts say the move only reinforces the need for long-term self-reliance.

Despite its limits, the shift is significant. It marks the first major reversal of Biden-era AI chip controls, signals a willingness to trade restricted access for revenue, and adds uncertainty to an already tense US-China tech relationship.

Even if uptake is low, the approval highlights how AI hardware policy is evolving — and how quickly a single decision can reshape geopolitical dynamics.

That is why global investors, diplomats and national-security officials are watching closely. The H200 itself may not reshape the AI landscape, but the political signal behind its approval is far larger than the chip.