Nvidia CEO Jensen Huang seems to have made an agreement with the Trump administration To bypass export limitations on the firm's H2O AI processors.
The H2O, the highest-performing Nvidia-developed AI chip approved for export from the U.S. to China, was apparently exempt due to Huang’s commitment to establish new AI data centers within the United States. According to NPR Huang put forth the suggestion during an evening meal at Trump’s Mar-a-Largo property earlier last week.
Nvidia declined to comment.
Many within the semiconductor sector worried that H20s, which are altered to offer reduced performance compared to other Nvidia processors, might face limitations as they were apparently among the chips under scrutiny. A China-based company named DeepSeek utilized their resources to develop the R1 open AI model. Released in January, R1 garnered attention due to its impressive performance compared to models from U.S.-based AI laboratories such as OpenAI.
Senators from both parties have requested limitations on the H20 Even the Trump administration reportedly had plans to implement H2O export controls before changing direction, as reported by NPR.
It shouldn’t be entirely shocking that Trump reportedly consented to suspend certain semiconductor limitations in return for NVIDIA’s pledge to invest in American AI infrastructure. However, permitting NVIDIA to keep sending H100 chips to China seems to contradict the administration’s aim of ensuring U.S. supremacy in artificial intelligence.
Adding to the confusion is the Trump administration’s choice to maintain intact the AI chip export control measures Rules implemented by former President Joe Biden in January imposed chip export limitations on almost every nation outside the U.S., including American allies, with stricter constraints specifically targeting China and Russia.
Nvidia has described these guidelines as "unprecedented and misguided," stating that they could potentially hinder worldwide innovation.
Several AI firms apart from Nvidia have adopted Trump’s "America-first" strategy for AI initiatives to gain favor with his administration. OpenAI collaborated with SoftBank and Oracle on a $500 billion U.S.-based data center project known as the Stargate Project in January. Microsoft pledged $80 billion To construct AI data centers during its 2025 fiscal year, half of which will be allocated to the U.S.
Trump has coerced specific partners to achieve his preferred result. reportedly informed Taiwan Semiconductor Manufacturing Company (TSMC) that they might face taxes as high as 100% if the company decided not to construct additional chip fabrication facilities in the United States.