Nvidia has secured a crucial win in its ongoing negotiations with the U.S. government over AI chip exports to China. The Trump administration has decided to allow the tech giant to resume sales of its H20 chips to Chinese clients, following a reversal of earlier restrictions. According to Commerce Secretary Howard Lutnick, the green light for sales was granted on the condition that Nvidia would not export its most advanced chips, ensuring that national security interests remain protected while preserving a strategic business edge.
The chip in question, the H20, is considered Nvidia’s “fourth best” AI processor—less powerful than its leading Blackwell-generation chips such as the B100, B200, and the ultra-performance GB200. Lutnick explained that the U.S. remains committed to restricting access to its cutting-edge hardware but is willing to allow access to downgraded versions in order to maintain technological influence. By continuing to supply Chinese firms with mid-tier hardware, the U.S. intends to keep those companies reliant on the American technology stack.
The shift in policy reportedly followed a meeting last week between President Donald Trump and Nvidia CEO Jensen Huang in Washington, where the strategic benefits of continued exports were discussed. The administration's new stance appears to be tied to broader negotiations, including a rare-earth magnet deal, which may have factored into the licensing decision. The rare-earth materials are critical components in the manufacture of electronics, and the U.S. aims to secure a stronger position in the supply chain while avoiding China’s dominance in rare-earth processing.
Howard Lutnick emphasized that allowing China access to older chip models like the H20 ensures continued dependence on U.S. semiconductor design and manufacturing. The administration’s approach, he stated, is to remain one step ahead in innovation so that competitors cannot entirely replicate or replace American offerings. By strategically choosing what can be exported, U.S. companies like Nvidia can both profit and maintain influence without compromising national security.
Jensen Huang, for his part, has consistently argued that blocking all chip exports would only drive China to accelerate domestic innovation and investment in alternative AI infrastructure. He has pointed out that some Chinese firms, such as Huawei, are already making strides in building homegrown chips. Continuing to supply the H20 gives U.S. developers a foothold in the market and slows China’s independence trajectory.
The H20 was originally launched in 2022 as a response to export controls established by the Biden administration. It was designed specifically for markets with tighter regulatory limits, and it includes fewer GPU cores and reduced memory bandwidth. Despite these limitations, the H20 remains useful for many Chinese AI projects, such as DeepSeek R1, and has proven effective in less compute-intensive applications.
Nvidia claimed that its inability to sell the H20 earlier this year could have cost the company up to $8 billion in potential revenue for the quarter. The H20, while inferior to Nvidia’s newer Blackwell models, remains a sought-after product in AI circles where top-end performance is not a strict requirement.
The Trump administration’s export policy now appears to hinge on selective engagement—encouraging Chinese companies to remain within the U.S.-controlled ecosystem, rather than fully blocking access and risking the growth of a separate and competing Chinese chip infrastructure.
As the AI hardware arms race intensifies, Nvidia continues to innovate at a rapid pace. Its latest Blackwell generation chips, including the GB200, B100, and B200, are already being deployed in high-performance computing environments across the globe. The company has also begun rolling out the Blackwell Ultra series to data centers, with full-scale deployment expected over the coming year. Looking further ahead, Nvidia plans to launch a new chip architecture, codenamed “Vera Rubin,” by 2027.
This move by the Trump administration may offer short-term relief for Nvidia, but it also reflects a broader strategic calculus about how best to manage technological competition with China. Rather than a complete decoupling, the U.S. seems to be opting for a layered control system that retains leverage without sacrificing market share. Whether this strategy proves sustainable in the face of rapid Chinese development remains to be seen.
Meanwhile, the tech industry, investors, and geopolitical analysts will be watching closely to gauge the long-term impact of these export decisions on U.S.-China relations and the global AI ecosystem. The delicate balance between economic opportunity and national security continues to shape the semiconductor landscape in complex and evolving ways.
