Bridgewater Co-CIO Suggests Extra Scrutiny for Firms Controlling Over 5% of Global Compute

Greg Jensen, co-chief investment officer at Bridgewater Associates, said in a recent interview with The Information that major holders of AI computing power should face additional oversight similar to that applied to large financial institutions. He suggested that companies controlling more than 5% of U.S. or global compute resources could be designated as “systemically important institutions” and subjected to stricter regulation.

Jensen likened the proposed framework to the supervision of systemically important banks. “We should bring into regulatory scope any company that owns more than a certain percentage, say 5%, of global or U.S. compute resources,” he said. The idea would extend AI regulation beyond model developers to firms that control significant computing infrastructure.

He also floated a more direct measure: regulators could consider capping the amount of compute any single company may own, drawing an analogy to position limits in some financial markets. Jensen said such a cap is not unacceptable.

In addition, he argued that any AI model entering the U.S. ecosystem should be subject to regulatory standards. He noted that the AI industry is forming a highly concentrated compute structure, and that compute power has become an indispensable foundational resource for AI model training and operation.