The US government has mandated that Nvidia obtain a license to export its H20 chips to China and several other countries. As a result of the new restrictions, the company will incur expenses of approximately $5.5 billion for the financial quarter.
In 2022, the US restricted the export of Nvidia’s AI chips. The company developed the H20 specifically for the Chinese market, considering export restrictions. It is comparable to the H100 and H200 but features lower connection speed and bandwidth. The processor is based on the previous generation Hopper architecture.
DeepSeek, which stirred the world with affordable AI models, used the H20 in its research.
Nvidia argues that further control over its chips will hinder competition and could even undermine US competitiveness in technology.
The government has informed the company that licensing requirements for the H20 will be in effect “for an indefinite period.”
In April, media reported that the White House abandoned plans to impose restrictions on the import of the H20 shortly after the company’s CEO, Jensen Huang, attended a paid dinner with Donald Trump. In response, Nvidia promised to invest in developing data centers within the US.
Later, the chipmaker announced the creation of AI infrastructure in the country worth $500 billion.
In March, Nvidia unveiled new processors for creating and running AI models — Blackwell Ultra and Vera Rubin.