Nvidia CEO Jensen Huang reportedly secured a deal with the Trump administration to keep the company’s H20 AI chips free from export restrictions, at least for the time being. The H20 is Nvidia’s most advanced artificial intelligence chip still permitted for export from the United States to China. Reports indicate that Huang’s commitment to significant future investments in U.S.-based AI data centers was the key factor convincing the administration to avoid imposing anticipated export controls.
The discussions apparently took place at Trump’s Mar-a-Lago resort during a dinner meeting last week, according to media reports. Nvidia has not offered any official statement on the matter.
Industry experts and analysts had been concerned about the potential inclusion of the H20 chips under U.S. export restrictions due to their reported use by Chinese company DeepSeek. DeepSeek’s recent R1 open AI model has attracted attention because of its robust performance, rivaling AI tools developed by leading American firms such as OpenAI.
Lawmakers from both political parties had previously urged the Trump administration to swiftly place controls on the H20 chip exports. Early indications suggested the administration was seriously considering implementing these limitations before reversing its approach following the meeting with Nvidia’s leadership.
The administration’s decision has raised questions about policy consistency, particularly given its broader stance toward asserting American technological dominance. This development comes despite retaining the stringent AI chip export rules established by former President Biden earlier this year. Those guidelines restrict AI chip exports globally, especially targeting sensitive technologies bound for China and Russia, receiving fierce criticism from Nvidia and other industry leaders as “unprecedented and misguided.”
Major tech companies, knowing the current administration’s “America-first” policy, have significantly ramped up their domestic investments in AI infrastructure in recent months. For instance, OpenAI recently partnered with SoftBank and Oracle to form the $500 billion “Stargate Project,” aiming at building substantial U.S. data center capacity. Similarly, Microsoft has committed $80 billion toward new AI-focused data center construction, pledging half specifically to U.S. projects.
In line with this shift, President Trump has aggressively pursued agreements enforcing domestic production, even reportedly threatening Taiwanese manufacturer TSMC with severe tariffs if it declined to establish chip fabrication facilities within the United States.
