Background of US-China AI Chip Tensions

The global race for AI dominance has intensified US-China tech rivalry, with semiconductors like Nvidia’s GPUs at the center.

Under previous Biden-era export controls enforced through the US Department of Commerce’s Bureau of Industry and Security, sales of advanced AI chips to China were heavily restricted to curb Beijing’s military and surveillance capabilities.

Nvidia responded by developing compliant versions like the H20 chip, but these were less powerful and came with revenue-sharing deals. The H200, launched in 2023 as an upgrade to the H100, offers significantly higher performance—nearly six times that of the H20 for AI training tasks—making it a prized asset.

On December 8, 2025, President Donald Trump announced via Truth Social that the US would permit Nvidia to export H200 chips to “approved customers” in China and other countries, marking a reversal of stricter Biden policies.

Exports require vetting by the Commerce Department for national security. Sales of Nvidia’s more advanced Blackwell (B200) and upcoming Rubin chips remain banned. The US government will receive a 25% surcharge on H200 sales up from 15% on prior chips, framed by Trump as a win for jobs, manufacturing, and taxpayers.

Trump stated he informed Chinese President Xi Jinping, who “responded favorably.” This follows Nvidia CEO Jensen Huang’s recent Capitol Hill meetings with Trump. Similar terms will apply to competitors like AMD and Intel, potentially opening a $10-20 billion market in China for US firms.

Nvidia hailed the move as a “thoughtful balance” that supports American competitiveness without fully ceding ground. Shares rose ~2% initially on the news but pared gains later.

On December 9, 2025, reports emerged that Beijing is preparing to impose its own restrictions on H200 imports, effectively curbing widespread access even after Trump’s approval. According to sources familiar with the matter.

Chinese regulators are exploring “limited access” mechanisms, requiring buyers like data centers, AI firms to obtain government approval before purchasing. This could prioritize state-owned enterprises or align with national security reviews.

The move aligns with China’s “Made in China 2025” strategy and recent bans on foreign chips in new state-funded data centers. Beijing aims to triple domestic AI chip production by 2026, reducing reliance on US tech amid retaliatory tariffs and export curbs.

Customs has also tightened semiconductor import checks, delaying shipments. Analysts note Beijing’s concerns over potential US “backdoors” in chips, alongside a desire to boost local players like Huawei whose Ascend chips are gaining traction.

This dual-layer restriction—US export controls plus Chinese import limits—could blunt the deal’s impact. As Swissquote Bank’s Ipek Ozkardeskaya noted, it may not significantly boost Nvidia’s China revenue unless extended to newer lines like Blackwell.

For Nvidia ($NVDA): China accounts for ~13% of Nvidia’s revenue, but restrictions have forced pivots to compliant chips. The H200 approval could add $5-10 billion annually if unhindered, but Chinese limits might cap it at vetted buyers. Stock dipped 0.4% post-FT report.

AMD and Intel stand to gain similarly, but critics like Council on Foreign Relations’ Chris McGuire warn it risks eroding US AI leadership by accelerating China’s domestic tech. Coinciding with the news, US authorities busted a $160M ring smuggling H100/H200 GPUs to China, seizing $50M+ in hardware—highlighting enforcement gaps.

This tit-for-tat could spur EU and Asian allies to tighten their own controls, while boosting Huawei’s market share in China. Trump’s move eases one barrier but meets another from Beijing, underscoring the fragile US-China tech detente.

Watch for formal Chinese guidelines in the coming weeks, which could further shape AI supply chains.