Computing power has become an indispensable fundamental resource, Bridgewater: AI computing power giants should be regulated

As the artificial intelligence industry accelerates toward concentration within a handful of leading companies, the focus of AI regulation is shifting from issues such as model safety and data usage to the core infrastructure supporting AI development—computing power.

According to The Information, Bridgewater Associates Co-Chief Investment Officer Greg Jensen stated that the world’s top AI computing power holders need to be subject to additional oversight similar to major financial institutions.

If a company controls more than 5% of computing power resources in the United States or globally, it could be considered a “systemically important institution” and subjected to stricter regulation.

In an interview, Jensen explained that regulators could apply a framework similar to the oversight of systemically important banks to such companies—that is, once an institution holds a certain proportion of key resources, it must undergo additional regulation.

Furthermore, he suggested that any AI model entering the U.S. ecosystem should be subject to regulatory standards.

Jensen: Highly Concentrated AI Computing Power, Possibility of a “Computing Power Cap”

Jensen believes the AI industry is forming a highly concentrated structure of computing power, which has become a critical foundational resource for training and running AI models.

He noted that these companies can be viewed as “systemically important institutions” and subjected to regulation similar to that of systemically important banks.

“We should bring any company with more than a certain proportion—such as 5%—of global or U.S. computing power under regulatory oversight.”

This concept suggests that future AI regulation may not be limited to model developers themselves, but could also extend to companies controlling significant amounts of computing resources.

Jensen proposed another, more direct idea: regulators could consider limiting the amount of computing power a single company is allowed to own, to prevent a small number of firms from controlling an excessive share of computing resources.

He commented that he does not consider setting caps on how much computing power a company can hold to be unacceptable, drawing an analogy to position limits in parts of the financial markets.