Former SEC Chair Gary Gensler Talks AI Boom & More

Gary Gensler · MIT Sloan School of Management

AI boom will decelerate and then reverse within 18-30 months; depth depends on debt in the system.

Gensler expects a deceleration of the AI boom followed by some form of reversal in the next 18 to 30 months. He says the open question is timing and depth, and that depth depends on how much debt is in the system, where it sits and who holds it. He cites history (railroads, canals, the internet) and a game-theory dynamic in which everyone wants a piece of the boom, so capital markets tend to overfund in exuberance. As evidence he points to Meta raising $90 billion in equity and Broadcom announcing a $63 billion package of complex guarantees that was not in its S-1, all off balance sheet. He says he believes the economy is in the late stages of a capital expenditure boom, which makes the Fed's job harder if the boom decelerates. He says he can be positive about the technology and still expect capital to overshoot.

“Well, I think that we will have a deceleration of this AI boom and then some form of a reversal in the next eighteen to thirty months.”at 2:35

Gary Gensler · MIT Sloan School of Management

Capital markets should stay merit-neutral; disclosure of AI investment risks is the key safeguard.

Gensler argues that the US succeeds at keeping government out of investment choices, which are merit neutral. If the public wants to invest in Anthropic, that is the public's choice, as long as the risks are materially disclosed. He notes that a cleanup crowd usually arrives afterward. His implied position is that disclosure rules should be the main regulatory tool for AI investment risk, rather than government picking winners.

“I think we're really successful in The US keeping the government out of the choices we make that we're merit neutral. And if the public wishes to invest in anthropic, that's the public's choice as long as there's the material disclosures of all the risks.”at 5:29

Gary Gensler · MIT Sloan School of Management

Fed will likely raise rates in December, then may pause, given the AI capex deceleration risk.

Gensler says that if he were at the Federal Reserve, he would be asking what a deceleration in AI investment means for inflation and growth in 2027. He expects a rate increase in December. He also says the Fed might pause after December, so the question is whether it raises again in March. He frames this as a hard choice, since the Fed has to weigh the late stage of a capex boom and the risk of a deceleration against inflation.

“I think they're going to have a rate increase come December.”at 6:08

Gary Gensler · MIT Sloan School of Management

US has limited fiscal room; high debt constrains response, though its economy is larger than France's.

Gensler argues the US already runs very high debt relative to its economy. He points to France, which has even higher debt and no fiscal room to step in. He says the US also has limited fiscal room, but that it is a bigger and more significant economy, which gives it more capacity than France. He adds that a fiscal stimulus package could come in 2028 if a deceleration occurs, and he sees Social Security and fiscal challenges as the long-term structural issue.

“So we're already running very high debt to our economy, but see what's happening in France... We have limited fiscal room as well, but we're a bigger, I would say, more significant economy.”at 7:48

Gary Gensler · MIT Sloan School of Management

Presidents turn to foreign affairs in their last two years; Democrats may push affordability and AI safety.

Gensler says history shows that presidents in their last two years, and this president will have already had six years, tend to turn more to foreign affairs because they have less leverage with Congress. He expects Democrats in Congress to still try to get things done, focusing on affordability, housing and AI safety. The midterms are a week away, and he expects 15 to 20 presidential candidates to announce by early next year.

“History tells us that presidents tend in the last two years, and remember this president will have had six years already, will turn more to foreign affairs. They have less leverage with congress.”at 8:32

Gary Gensler · MIT Sloan School of Management

AI data center financing adds systemic debt risk, including off-balance-sheet guarantees and insurer exposure.

Gensler is concerned about systemic risk from AI borrowing. He says more than $500 billion of new AI-related debt has been raised this year, and adding Broadcom's $63 billion package takes the total toward $600 billion. He says he worries about who holds that debt, pointing to insurers that have picked up a lot of it directly or through private credit. He also worries about guarantees that Broadcom, Nvidia and AMD have written into data centers, which sit off balance sheet. He says these arrangements use more special purpose vehicles than Citicorp did in its day. The implication is that the buildout's funding structure could amplify a downturn.

“I do concern myself that the risk in the system in artificial intelligence, we've raised over $500,000,000,000 of new debt this year. And then you add this Broadcom 63,000,000,000.”at 10:57

Gary Gensler · MIT Sloan School of Management

AI chips bought today may lose most value by 2032 given rapid scaling, despite 4-7 year useful lives.

Gensler says a lot of the funding is going to chips, including memory chips, which may have four to seven year useful lives on the books. He acknowledges that someone in another country may still use them later. However, he argues that scaling and technological advances are so fast that the chips of 2026 will not be very valuable in 2032. He says he does not have a clear picture of whether the market understands this depreciation or obsolescence risk, which is why he is worried about the financing built on these assets.

“And a lot of the funding is funding the chips, the memory chips, and those chips might have four to seven year useful lives... the chips of 2026 won't be that valuable in 2032.”at 12:08

Gary Gensler · MIT Sloan School of Management

More optimistic about AI productivity gains over 5-10 years, though data centers may be overbuilt.

Gensler says some at MIT think AI will bring only modest productivity gains and significant labor disruption. He is more optimistic and expects productivity gains over the next five to ten years. He qualifies this by noting that the buildout could still mean some data centers have been overbuilt, so the productivity upside does not rule out an investment bust.

“Some think there's only gonna be modest productivity gains and there'll be significant labor disruptions. I'm more optimistic about the productivity gains over the next five to ten years”at 14:28