The U.S. Department of Commerce formally opened a path for limited exports of Nvidia’s H200 artificial intelligence chips to China in January 2026, replacing a blanket restriction with a case-by-case licensing system. The policy also covers AMD’s MI325X processors and requires buyers to meet strict security and compliance checks. Yet weeks after the rule took effect, not a single H200 chip had reached Chinese soil, exposing a gap between Washington’s regulatory green light and the real-world ability to complete a sale.
How the BIS Licensing Rule Works
The Bureau of Industry and Security announced the revised policy on January 13, 2026, and the formal rule appeared in the Federal Register two days later. Rather than applying an automatic denial to export applications for advanced AI semiconductors destined for China, BIS now evaluates each license request individually. The rule specifically names the Nvidia H200 and AMD MI325X, two of the most capable commercially available AI training accelerators, as products eligible for this new review track, a change the agency outlined in a January statement from the Commerce Department.
Approval is not guaranteed. Each application must satisfy security requirements that include purchaser compliance procedures and, according to Associated Press reporting, third-party review of end users before chips can ship. Nvidia praised the framework, stating in a corporate release that it applies to “approved commercial customers, vetted by the Department of Commerce.” The U.S. government also collects 25% of the sales price as an import fee, a revenue-sharing mechanism that Brookings researchers have flagged as unusual and potentially problematic from a legal standpoint. In practical terms, the rule converts an outright ban into a conditional gateway, one designed to let Washington capture economic value from chip sales while retaining the power to block any individual transaction.
From White House Clearance to Formal Rule
The January 2026 regulation did not appear in a vacuum. It followed a presidential clearance announcement on December 8, 2025, when the White House confirmed that sales of more advanced Nvidia AI chips to China would be permitted under certain conditions. That announcement, described by one Washington Post account, drew immediate pushback from national security hawks who argued that any loosening of chip controls would hand Beijing tools to accelerate military AI programs. The White House framed the move as a calibrated adjustment rather than a wholesale retreat, emphasizing that the most advanced processors would remain off-limits.
The internal debate was sharp. Reporting from Politico journalists described objections from China hawks inside the administration who viewed the decision as a concession that weakened the broader technology containment strategy. Supporters of the policy countered that keeping Nvidia locked out of the Chinese market simply redirected revenue to foreign competitors while doing little to slow China’s own chip development. The administration framed the approach around selling only to “approved customers,” according to a separate technology-focused report, and excluded Nvidia’s most advanced chip families from the new licensing track. The December announcement served as the political predicate; the BIS rule published in January translated that political decision into enforceable regulatory language.
Zero Chips Delivered Despite U.S. Approval
The most striking outcome so far is that the policy has produced no actual trade. As of February 24, 2026, Nvidia had sold zero H200 chips to China, according to a U.S. export official quoted by Reuters. Nvidia did receive a U.S. license for a small quantity of H200 exports to China around February 25, 2026, but even that approval had not translated into deliveries at the time it was reported. On paper, the United States has shifted from a categorical ban to a controlled permission regime; in practice, the flow of chips remains at zero.
One reason for the stall sits on the other side of the Pacific. Chinese customs authorities have reportedly moved to block entry of H200 chips that the U.S. government had cleared for export, according to Reuters-sourced coverage in a UK newspaper. That means Beijing is exercising its own import controls as a countermeasure, effectively neutralizing Washington’s conditional opening. For Chinese buyers, the risk is twofold: they must navigate Washington’s licensing hurdles and Beijing’s political calculus, with either side able to halt a shipment. The result is a policy that exists on paper but has yet to move a single processor across the border.
Why the Standoff Matters for AI Competition
Most coverage of the H200 licensing shift has focused on whether the U.S. is “loosening” or “tightening” controls. That framing misses the more consequential dynamic: the policy may be accelerating the very outcome it was designed to prevent. By signaling that advanced chip access will be rationed, conditional, and subject to political volatility on both sides, Washington and Beijing are giving Chinese firms a stronger incentive to develop domestic alternatives and to re-architect AI systems around hardware they can reliably obtain. Even if some H200 shipments eventually clear customs, the episode underlines that foreign accelerators are an inherently fragile foundation for long-term industrial or military planning.
In the near term, the standoff also complicates global supply chains. Multinational cloud providers and hardware vendors must now navigate a patchwork of export licenses, import blocks, and shifting political red lines when deciding where to deploy cutting-edge AI infrastructure. The unusual 25% U.S. import fee highlighted by Brookings adds another layer of friction, effectively functioning as a tax on cross-border AI capacity. For governments and companies outside the U.S.–China axis, the episode is a reminder that access to top-tier AI chips is increasingly mediated by great-power rivalry rather than market demand alone.
What Comes Next for Export Controls
The H200 saga offers a preview of how future export control fights may unfold. Rather than a one-way restriction imposed by Washington, advanced technology trade is becoming a two-sided negotiation in which Beijing also wields customs and regulatory tools. If Chinese authorities continue to block imports of U.S.-approved chips, they can undercut any attempt by Washington to use export licensing as a calibrated pressure valve, leaving both sides worse off economically while pushing innovation into more politically insulated spheres. At the same time, U.S. policymakers may conclude that if China is blocking imports anyway, there is little reason to tolerate domestic criticism over even a conditional opening.
For now, the most concrete impact is uncertainty. Nvidia, AMD, and their customers cannot reliably plan around a licensing regime that is technically open but practically inert. Chinese AI developers face a similar fog, unsure whether to invest in systems built around H200-class hardware that may never arrive. International observers, including media outlets that routinely ask readers to support deeper coverage of such issues, such as the subscriber appeals tied to some UK reporting, are left parsing a policy that has generated headlines but not hardware flows. Unless either Washington or Beijing chooses to de-escalate, the current equilibrium (licenses granted, chips undelivered) may become a template for how advanced AI technology is contested rather than traded.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
