U.S. Clears Nvidia’s H200 Sales to China, Reopening a High‑Stakes Channel for AI Hardware

U.S. Clears Nvidia’s H200 Sales to China, Reopening a High‑Stakes Channel for AI Hardware

Key Takeaways

• The U.S. will allow Nvidia, Intel, AMD, and others to sell certain AI chips to approved customers in China, ending an effective ban.
• Nvidia’s H200 is included, but its Blackwell-generation chips remain off‑limits.
• The move faces bipartisan skepticism in Congress, even as Nvidia and other chipmakers push for access to China’s commercial market.

The White House is preparing to reopen a controlled path for U.S. AI semiconductors into China, with President Donald Trump stating he informed Chinese President Xi Jinping that Nvidia and other American chipmakers will be allowed to ship certain products to approved Chinese customers. It is a narrow opening—limited to Nvidia’s H200 and comparable chips from Intel and AMD—but the business implications are significant enough that the industry reacted immediately. Nvidia shares jumped nearly 3% in after-hours trading within minutes of the announcement.

The H200 is one generation behind Nvidia’s Blackwell architecture, which remains excluded from the plan. That is a small detail, but it says a lot about how carefully the administration is calibrating this policy. Washington is letting older—but still highly capable—AI accelerators return to the Chinese market while keeping the most advanced hardware locked down.

Nvidia already has more than $500 billion in orders for its top-tier chips stretching through this year and next. That backlog doesn’t even account for potential Chinese demand, which could expand the pipeline substantially if buyers decide the H200 is worth adopting. And that raises an immediate question: will Beijing actually allow these chips in? The last time the U.S. permitted sales of a specifically tailored model—the H20—China’s government signaled disinterest, pressuring domestic firms to buy local alternatives instead.

Still, Trump noted that Xi “responded positively” to the latest proposal, adding yet another layer of geopolitical texture. Whether that enthusiasm holds as details emerge is another matter. The Department of Commerce is still finalizing the rules, which are expected to include a 25% U.S. government fee on approved chip sales, an increase from the previously discussed 15%.

Nvidia publicly backed the decision, arguing it strikes a balance that supports U.S. manufacturing and high-paying jobs. The company’s leadership, especially CEO Jensen Huang, has been lobbying for months to regain access to at least part of the Chinese AI market. Huang has been unusually vocal about the strategic value of keeping China inside the U.S. technology ecosystem. In a recent interview, he compared it to how the American dollar serves as the foundation for global economic systems—framed around the idea that the U.S. “tech stack” should underpin global industry, including China’s.

And yet, Huang’s argument hasn’t convinced everyone on Capitol Hill. A bipartisan group of lawmakers has warned the administration against loosening AI chip restrictions, citing national security risks. Some Republicans met with Huang recently, but Sen. John Kennedy outright skipped the meeting, stating he didn’t view the Nvidia CEO as an objective source on China sales. His critique wasn’t subtle: Huang has a financial incentive, Kennedy noted, to expand shipments to Chinese buyers.

Other lawmakers took a more measured tone. Sen. Mike Rounds described the conversation as healthy and necessary. Meanwhile, Sen. Pete Ricketts—who, along with Sen. Chris Coons, recently introduced a “Safe Chips Act”—argued that keeping advanced AI hardware out of Beijing’s hands is “essential to our national security.” These concerns aren’t new, but the renewed friction comes at a moment when U.S.–China relations are seeing a tactical thaw. China is increasing soybean purchases and has signaled it will allow rare earth exports to American buyers. Those gestures add a pragmatic backdrop to the chip decision, even if they don’t erase the strategic tensions.

Intel declined to comment on the development, and AMD didn’t respond to a request for comment. That silence isn’t surprising; both companies have been walking a tightrope for years, trying to maintain their global revenue base while navigating a shifting regulatory environment. For context, the U.S. has been imposing export controls on semiconductors and manufacturing equipment since 2022, creating a complex compliance landscape that shifts frequently.

The irony, of course, is that American chip companies have long depended on customers in China, even as Washington has tightened restrictions. The announcement doesn’t unwind that dynamic, but it creates a small, structured channel that could restore some of the lost revenue while still keeping strategic technologies off the table. Whether that line holds depends on how Commerce defines “approved commercial customers”—a phrase that could mean very different things depending on the final rules.

It is also worth remembering that China has been working aggressively on its own AI chip alternatives. The performance gap between domestic Chinese accelerators and U.S. offerings is still significant, according to reporting from Bloomberg, but it is narrowing. Beijing may see limited-value chips as distractions rather than enablers. Or, it may recognize that buying vetted U.S. hardware, even a generation behind, keeps Chinese commercial AI systems tied to the global mainstream—something many of its largest companies quietly prefer.

For B2B technology leaders, the uncertainty is the point. AI hardware roadmaps, cloud service build-outs, and long-term data center commitments tend to span three, five, even seven years. When a single policy change can swing billions in market opportunity, it forces CTOs and infrastructure teams to revisit assumptions they thought were locked down. The U.S. has now created a precedent for partial access, but not necessarily a stable one.

As for Nvidia, the company now faces the same challenge it had before this announcement—just with a larger potential customer base. It must keep up with overwhelming global demand while also navigating a geopolitical environment that can shift with a single press release. That dual pressure is becoming the new normal for advanced chipmakers. They are no longer just selling hardware; they are operating in the middle of foreign policy.