US Permits Nvidia H200 Chip Exports to China Under Revenue-Sharing Mandate

On Monday, December 8, 2025, US President Donald Trump announced a significant adjustment to semiconductor export policy, authorizing Nvidia to ship its advanced H200 artificial intelligence chips to approved customers in China and other international markets. This decision partially reverses stringent controls previously enacted under the Biden administration, which sought to limit China's access to cutting-edge technology citing national security concerns.

The new arrangement, communicated by President Trump via Truth Social, mandates that the United States will receive a 25% share of the revenue generated from these H200 chip sales. This policy shift is presented as a strategic compromise balancing economic interests with national security objectives. President Trump confirmed informing Chinese President Xi Jinping, who reportedly offered a positive response to the proposal. The agreement explicitly excludes Nvidia's most advanced processor lines, the Blackwell and forthcoming Rubin architectures, which will remain exclusively available to US customers.

The Department of Commerce is now responsible for finalizing the technical guidelines for these sales to ensure they align with a strong national security posture. This revenue-sharing mechanism represents an escalation from the 15% cut that Nvidia and Advanced Micro Devices (AMD) had previously agreed to in August 2025 for the less powerful H20 chip, a prior deal that encountered legal scrutiny regarding export tax prohibitions. Nvidia stated that offering the H200 to vetted commercial customers strikes a thoughtful balance that benefits the United States, supporting high-paying jobs and domestic manufacturing.

The H200 chip, based on the Hopper architecture, features 141 gigabytes of HBM3e memory and a memory bandwidth of 4.8 terabytes per second, making it nearly six times more powerful than the H20 version. These enhanced memory capabilities are particularly advantageous for accelerating generative AI and large language models, positioning the H200 as the second most advanced product available behind the Blackwell series. The policy shift appears to move away from the previous strategy of requiring US chipmakers to develop 'degraded' products for China, aiming instead to secure revenue from the highly capable H200 while preserving the technological lead with the newest chips.

Analysts suggest that fully realized H200 sales to China could contribute between US$10 billion and US$15 billion to Nvidia's second half of 2025 results, with an estimated annual addressable opportunity reaching approximately US$50 billion. Furthermore, the framework is expected to be extended to competitors, with Intel and AMD anticipated to be subject to the same 25% revenue cut authorization for their comparable high-end chip exports. This development signals a notable pivot in US technology containment strategy toward conditional market engagement.