Nvidia is in talks with the U.S. government about shipping a new, more advanced chip to China, CEO Jensen Huang said on Friday.

Earlier this week, Reuters reported the U.S. tech giant is developing a new artificial intelligence chip for China, dubbed the B30A, that will be more powerful than the H20 — the only semiconductor Nvidia is allowed to sell in the country at present. The U.S. has grown concerned in the past few years that advanced American chips could be used in Chinese military applications.

A journalist asked Huang about the B30A during a trip to Taiwan.

“Offering a new product to China for the data center, AI data centers, the follow on to H20, that’s not our decision to make. It’s up to of course the United States government. And we are in dialogue with them. But it’s too soon to know,” Huang said in response.

Last month, Huang said he hopes that Nvidia can sell more advanced chips in China than the H20 during a visit to the country.

Nvidia’s position in China has become a headache for Huang. The company created a special, less-advanced chip for China called the H20, which this year the U.S. government restricted for export. In July, Nvidia said it had given permission to sell this chip again in China. Later, it was revealed that Nvidia will give 15% of its China chip sales to the U.S. government in exchange for export licenses.

Just as it appeared that Nvidia was back in China, it hit other roadblocks, with Chinese authorities raising concerns this month about potential security vulnerabilities in the company’s chips. Nvidia said its products do not have “kill switches and backdoors” built into them.

Several reports this month have suggested that the Chinese government has urged local companies not to use Nvidia chips.

Huang has argued that Nvidia should be allowed to sell its chips to China, so that the country’s AI is built on American technology and domestic tech giants like Huawei don’t fill the void.

That message appeared to get through to Washington. In July, when the H20 was approved for export again, U.S. Commerce Secretary Howard Lutnick told CNBC that the move was allowed because Nvidia would not be giving over its best technology.

“We don’t sell them our best stuff, not our second best stuff, not even our third best,” Lutnick said.

However, the Financial Times reported on Thursday that these comments were seen as “insulting” by Chinese officials and that local regulators are moving to dissuade domestic firms from buying the H20.

A report by the The Information on Friday said that Nvidia has asked some of its component suppliers to stop production related to the H20 graphics processing units.

The company’s shares were down 1.34% in premarket trading at 5:53 a.m. E.T.

– CNBC’s Dylan Butts contributed to this report.

Chinese artificial intelligence startup DeepSeek has hinted that China will soon have homegrown “next generation” chips to support its AI models, while announcing an update to one of its large language models. 

In a comment under a post on its official WeChat account, DeepSeek said the “UE8M0 FP8” precision format of its newly released model V3.1 is tailored for the next-generation domestically built chips that will be launched soon.

FP8, or 8-bit floating point, is a data processing format that can boost the computational efficiency for training and inference of large deep learning models.

DeepSeek’s mention of China’s coming next-generation chips may signal plans to work more closely with China’s emerging AI chip ecosystem in the face of Washington’s advanced semiconductor export restrictions and Beijing’s push for chip self-sufficiency.

The comments come about two weeks after Beijing reportedly urged Chinese AI developers to use domestic alternatives to Nvidia’s graphics processing units used in AI training. While analysts say China’s domestic AI chipmakers lag behind Nvidia in technological advancement and scale, players like Huawei have been making progress.

In its Thursday post, DeepSeek did not disclose the chips it used to train the V3.1, or what local chips the UE8M0 FP8 might be compatible with.

DeepSeek shook up the tech world earlier this year after it released its R1 reasoning model, which demonstrated capabilities comparable to those of Western competitors like OpenAI, despite U.S. export controls restricting it from using Nvidia’s most advanced AI training chips.

Prior to that, in December, the company released its V3 model, which it said had been trained on about 2,000 of Nvidia’s less advanced chips.

Following DeepSeek’s model breakthroughs, the U.S. further tightened export restrictions in April, effectively banning Nvidia’s H20 chips, which had been specially designed to meet prior export restrictions on China. 

Last month, officials from the Trump administration said they planned to allow Nvidia to resume shipping the chips to China. However, the H20s are now being met with scrutiny in China, with regulators reportedly mandating companies against buying the chips until a national security review is completed.

Chip analysts have told CNBC that companies like Huawei that have been seeking to build an alternative AI chip ecosystem in China could benefit from a lack of Nvidia’s H20s in the market. 

DeepSeek said Thursday that its V3.1 came with “major changes,” including faster response times, and a hybrid reasoning architecture that allows the model to support both reasoning and non-reasoning modes. Reasoning models can execute more complicated tasks through a step-by-step logical thought process.

Starting Sept. 6, the company will also adjust the pricing for using the model’s API, which allows developers of other apps and web products to integrate DeepSeek on their platforms. 

Elon Musk, the world’s richest person, asked Meta CEO Mark Zuckerberg to help him finance a $97.4 billion takeover of OpenAI in early 2025, court filings on Thursday revealed.

The filing is part of a legal case between Musk and OpenAI that was initiated last year. The case is proceeding in a federal court in Northern California, and a judge recently said OpenAI can move ahead with counter claims against Musk, who co-founded OpenAI as a non-profit with Sam Altman and others in 2015.

When Musk floated his proposal to buy OpenAI in February, he was incensed that the company and Altman, OpenAI’s CEO, were pushing to transform the business into a for-profit entity. Altman and Musk, who were longtime friends, have become bitter adversaries since OpenAI’s emergence as a leader in generative AI with billions of dollars in funding from Microsoft.

Musk started xAI in 2023 and was pushing for it to be a direct competitor to OpenAI. Musk later sued OpenAI, alleging a breach of contract, and tried to stop OpenAI from converting to a for-profit company.

In its counter claims, OpenAI has alleged that Musk and xAI’s “sham bid” harmed its business and that Musk has engaged in “harassment” through litigation and attacks on social media and in the press.

As part of its complaint, OpenAI has filed to subpoena Meta for communications between the company, its CEO and Musk about the bid.

In a statement to the court published Thursday, OpenAI said that when Musk and xAI were trying to form a consortium of investors to finance a takeover, they approached Zuckerberg with a letter of intent and asked “about potential financing arrangements or investments.”

Neither Zuckerberg nor Meta signed the LOI, the filing said.

A Meta spokesperson declined to comment. Marc Toberoff, Musk’s attorney in the case, didn’t respond to a request for comment.

The statement in the filing said that Meta has been “spending heavily to develop its own Al capabilities” and has been “offering pay packages of $100 million or more to leading Al researchers and attempting to poach OpenAI employees.” 

Meta has argued that OpenAI’s requests for documents are overly burdensome, and that OpenAI should obtain relevant communications from Musk and xAI, instead.

WATCH: How Sam Altman is tackling a growing threat to the future of OpenAI: Elon Musk