Nvidia’s processors are key to training the large language models that power AI bots like OpenAI’s ChatGPT and Google’s Bard. That makes the company’s products “an incredibly practical chokepoint that would allow the U.S. to impose itself on all other actors” in AI, Mustafa Suleyman, cofounder of DeepMind and Inflection AI, told the Financial Times in an interview published Friday.

In July, Inflection joined six other companies, including Google, Microsoft, Meta, and OpenAI, at the White House to commit to managing AI risks. The pledges include promises to rigorously test AI models before releasing them to the public, invest in cybersecurity, and develop measures to reveal when content is AI-generated.

“The U.S. should mandate that any consumer of Nvidia chips signs up to at least the voluntary commitments—and more likely, more than that,” Suleyman told the Financial Times, referring to the promises made at the White House.

Nvidia declined to comment.

In 2010, Suleyman cofounded DeepMind, which Google acquired in 2015. The company’s AlphaGo program famously beat a world champion Go player in 2016. 

He then left Google in early 2022, and founded a new AI company, Inflection AI, a few months later. Earlier this year, Inflection AI announced its chatbot Pi, designed to serve as a “digital assistant.” Inflection AI is now valued at $1.3 billion following a June funding round that included Nvidia, Microsoft, and Bill Gates.

Who is regulating A.I.?

Despite Suleyman’s insistence that the U.S. can set the agenda, Washington is behind its peers in Europe and China when it comes to thinking about how to regulate the new technology. Suleyman himself admitted to the Financial Times that the U.S. was falling behind, calling the odds of passing legislation “very low.”

The European Union is currently considering the so-called A.I. Act, which would judge the risk level of different uses of AI. Programs considered high-risk, such as those whose decisions can harm a person, will have to go through multiple rounds of testing before release. Some uses of AI, like facial recognition, could be banned entirely.

European business leaders are worried that overregulation may put Europe at a disadvantage compared to the U.S. The European Parliament is expected to pass some version of the act later this year.

Yet Suleyman praised Europe, saying it was “heading in the right direction” in his interview with the Financial Times. 

China has also taken an early lead in setting rules on AI. Last December, regulators imposed rules on deepfake technology, barring its use for fake news and requiring a notification that an image or video was altered. Censors followed that up with regulations on generative AI last July. Chatbots need to pass a security review and uphold “core socialist values,” yet the rules also promise to support further innovation and do not refer to penalties for companies that breach the rules.

The lighter-than-expected rules were a boon to China’s tech companies, with Baidu CEO Robin Li calling them “more pro-innovation than regulation” in an earnings call in late August. Regulators gave Baidu, along with several other Chinese tech companies, the green light to release chatbots to the public last week.  

Some U.S. tech leaders have warned against stringent U.S. regulation of AI, citing a fear that it would give China a lead in the new technology. 

Nvidia as a ‘chokepoint’

Nvidia may not be too keen on being used as leverage for any particular agenda.

The company’s chips are critical for any company working on AI. Demand for Nvidia processors helped to boost sales at its data center segment (which correlates to demand for AI) by 171% year on year for the most recent quarter. 

But that’s also put the company in the crosshairs of U.S. regulation, as the Biden administration tries to limit China’s ability to develop AI.

Last September, both Nvidia and fellow chipmaker Advanced Micro Devices revealed that the U.S. barred them from selling their most advanced chips to Chinese companies.

Earlier this year, the Biden administration banned new U.S. investments into Chinese companies working on strategic technologies like semiconductors and AI. The White House is also reportedly considering further controls on chip sales to China, as well as limiting Chinese access to U.S.-based cloud computing services.

Chinese companies are now hurriedly snapping up billions of dollars worth of Nvidia chips to get ahead of new controls, the Financial Times reported last month.

Nvidia makes up to 25% of its data center revenue from China, according to CFO Colette Kress on the company’s earnings call in August. While the company doesn’t see an “immediate material impact” from expanded export controls, Kress warned that restrictions in the long term will “result in a permanent loss of opportunity for the U.S.”

Still, the White House isn’t concerned about blowback to U.S. chip companies.

The U.S. is “trying to choke [China’s] military capacity,” U.S. Commerce Secretary Gina Raimondo told NBC’s Meet the Press on Sunday, following her recent visit to China.

That leaves a lot of room for companies to sell chips that are “not the leading-edge,” Raimondo said. “Billions of dollars” in sales would create “revenue for American companies, which they can plow back into research and development, which allows us to lead the world in innovation,” she continued. 

Update, September 5, 2023: This article has been updated with a response from Nvidia.