NextFin News - In a revealing testimony before the House Foreign Affairs Committee on Wednesday, February 25, 2026, a senior U.S. official confirmed that Nvidia has not yet received authorization to export its advanced H200 artificial intelligence GPUs to China. David Peters, Assistant Secretary for Export Enforcement at the Department of Commerce, stated during the hearing in Washington D.C. that "not a single unit" of the H200 has been approved for sale to Chinese entities. This revelation comes nearly three months after U.S. President Trump signaled a significant easing of export restrictions, a move intended to allow American semiconductor firms to reclaim lost ground in the world’s second-largest economy.
According to Tom's Hardware, the lack of approved licenses persists despite Nvidia’s aggressive preparation for a market reentry. Following the White House’s policy shift in late 2025, Nvidia reportedly prepared a massive shipment of approximately 82,000 AI-focused GPUs, anticipating a swift resumption of trade. However, the disconnect between executive-level policy announcements and the granular licensing process managed by the Department of Commerce’s Bureau of Industry and Security (BIS) has created a strategic vacuum. Representative Sydney Kamlager-Dove, who questioned Peters during the hearing, highlighted the ambiguity surrounding the current regulatory environment, which has left billions of dollars in hardware sitting in logistical purgatory.
The current situation represents a complex paradox in U.S. trade policy under U.S. President Trump. While the administration has publicly advocated for "economic pragmatism"—allowing U.S. firms to sell high-margin hardware to China to maintain American technological leadership—the underlying national security apparatus remains deeply skeptical. The H200, which offers a significant performance leap over the previously restricted H20 and H100 models, is a critical component for training large language models (LLMs). The delay in licensing suggests that while the political will to trade has shifted, the technical criteria for ensuring these chips do not enhance Chinese military capabilities have become more stringent and difficult to satisfy.
From a market perspective, the stakes for Nvidia are monumental. Before the initial wave of export bans, Nvidia controlled nearly 95% of the Chinese AI chip market. That dominance has since evaporated, falling toward zero as Beijing accelerated its "de-Americanization" of the domestic supply chain. In the absence of Nvidia’s flagship products, Chinese tech giants such as Alibaba and Tencent have pivoted toward domestic alternatives. Huawei’s Ascend series, particularly the 910B and its successors, has seen a surge in adoption. Every month that the H200 remains blocked by U.S. regulators is a month that Huawei and other local players like Biren Technology use to entrench their software ecosystems and refine their hardware-software integration.
The financial implications for Nvidia are equally stark. The 82,000 units currently in limbo represent a significant portion of the company’s high-end inventory. If these units cannot be cleared for the Chinese market, Nvidia may be forced to reallocate them to other regions, potentially at lower margins or into a market that is already reaching a short-term saturation point for H200-class hardware. Furthermore, the hearing addressed the rise of a "gray market" for AI chips. Peters noted that the Department of Commerce is intensifying efforts to combat the smuggling of GPUs into China, an unintended consequence of the very restrictions that the administration is now theoretically trying to ease.
Looking forward, the trajectory of U.S.-China semiconductor trade appears increasingly volatile. The "green light" from the White House has clearly not translated into a "fast track" at the Department of Commerce. This suggests a two-tier regulatory reality where U.S. President Trump sets the broad geopolitical tone, but career officials and security hawks within the BIS maintain a veto through administrative delay. For Nvidia and its investors, the primary risk is no longer just the policy itself, but the execution of that policy. If the licensing bottleneck is not resolved by the second half of 2026, the window for the H200 to serve as a "market re-entry" tool may close entirely, as Chinese customers move toward the next generation of domestic silicon, rendering the American technological advantage a moot point in the region.
Explore more exclusive insights at nextfin.ai.
