NextFin News - In a move that reshapes the geopolitical landscape of the global semiconductor industry, U.S. President Trump has formally authorized the re-entry of Nvidia’s high-performance H200 AI chips into the Chinese market. The decision, finalized in early February 2026, comes after months of intense lobbying by Nvidia CEO Jensen Huang and a series of high-stakes negotiations between Silicon Valley and the White House. Under the new policy, Nvidia, along with competitors AMD and Intel, is permitted to sell advanced AI processors to Chinese firms, provided they pay a 25% 'revenue-sharing' fee directly to the U.S. Treasury. This transactional shift in export control policy marks a departure from the previous administration’s rigid bans, aiming to monetize American technological dominance while maintaining a degree of oversight through mandatory third-party testing of chips routed through U.S. soil.
According to The Diplomat, the return of the H200—a chip significantly more powerful than the previously 'sanitized' H20 variants—is being met with cautious optimism by Chinese tech giants like Alibaba and Tencent, who have struggled to train large language models under previous hardware constraints. However, the 'revenue-for-access' model has triggered a fierce domestic debate in Washington. U.S. President Trump’s administration argues that the 25% tax ensures the American public benefits from the sale of sensitive technology, but national security hawks in Congress, led by Representative Brian Mast, have introduced the 'AI Overwatch Act' to challenge the executive branch's authority. They argue that the policy treats advanced silicon as a mere commodity rather than a strategic weapon, potentially allowing China to narrow the AI gap.
The impact of this policy shift is already being felt across the AI supply chain. By allowing Nvidia to resume high-end shipments, the U.S. is effectively re-inserting its hardware into the heart of China’s AI infrastructure. Yet, the 25% surcharge acts as a double-edged sword. For Nvidia, it represents a significant margin squeeze; for Chinese buyers, it is a 'tax on progress' that makes American hardware increasingly expensive. This price pressure is playing directly into the hands of Beijing’s long-term strategy. Despite the availability of the H200, the Chinese government has not wavered in its support for domestic champions. According to Lawfare, Beijing continues to issue internal directives urging state-linked enterprises and private tech leaders to prioritize Huawei’s Ascend 910C series, which reportedly nears parity with Nvidia’s older H100 architecture.
From an analytical perspective, the return of Nvidia to China is less a sign of thawing relations and more a tactical maneuver in a broader economic war. The Trump administration’s approach utilizes the 'Section 232' framework of the Trade Expansion Act to treat chip exports as a matter of national security that can be 'adjusted' through tariffs. However, this creates a paradoxical incentive structure. While U.S. President Trump seeks to extract revenue from Nvidia’s $3 trillion market cap, the high cost of these licensed chips is subsidizing the R&D of Chinese rivals. Huawei, in particular, has benefited from the 'forced' localization of the past two years, developing a robust software ecosystem, CANN, that rivals Nvidia’s CUDA. By the time the H200s arrive in volume, many Chinese developers may have already migrated their workloads to domestic stacks.
Furthermore, the legal stability of this export regime is in question. Legal analysts suggest that the 25% fee may violate the Export Control Reform Act (ECRA), which prohibits the Bureau of Industry and Security from charging fees for export licenses. If the 'AI Overwatch Act' gains traction, or if chipmakers launch a constitutional challenge based on the Export Clause, the flow of H200s could be halted as quickly as it began. This regulatory uncertainty is prompting Chinese firms to continue 'dual-tracking'—buying Nvidia chips for immediate needs while aggressively investing in Huawei and Cambricon for long-term security.
Looking ahead, the 'Nvidia-Huawei' rivalry in China is set to enter a new phase of 'co-opetition.' Nvidia will likely dominate the high-end training market in the short term due to the H200’s superior efficiency, but Huawei is poised to capture the massive inference market as Chinese enterprises seek lower-cost, politically 'safe' alternatives. The 2026 semiconductor outlook suggests that while Nvidia’s revenue from China will rebound from its 2025 lows, its market share will never return to pre-2023 levels. The ultimate winner of this policy shift may not be the U.S. Treasury or Silicon Valley, but rather the Chinese domestic industry, which now has a clear price ceiling and a renewed sense of urgency to achieve total technological independence.
Explore more exclusive insights at nextfin.ai.