Jamie Dimon said hyperscaler AI spending could hit $1 trillion next year. The number is not new in the boardroom. What is new is that it has now entered the bond market, where it is being underwritten by the same banks that issue the debt.
Dimon's figure lands at a moment when the AI capex cycle is no longer being financed just by cash flow. It is being financed by a combination of vendor financing, private credit, and corporate bond issuance. That matters because the $1 trillion is not one check. It is a stack of obligations that lenders have to underwrite, and the terms of those obligations are what will determine whether the buildout slows in 2027. A hyperscaler that can issue investment-grade debt at tight spreads will keep building. One that has to roll maturing debt at wider spreads will start to ration.
The second-order effect is on the banks themselves. JPMorgan and its peers sit on both sides of this trade. They are issuing the bonds, structuring the credit facilities, and in some cases underwriting the private credit that sits behind the data center. That is a new role for a bank that has historically been more comfortable on the other side of the table. It means the bank's balance sheet now has a direct claim on the AI buildout, which is a claim that can move with chip prices, power availability, and regulatory changes.
For markets, the question is whether $1 trillion is a floor or a ceiling. If it is a floor, the 2026 rally in AI names has not yet priced the full cycle. If it is a ceiling, the next wave of earnings will be about where the spend stops, and who is left holding the debt when it rolls.
Image source: i.ibb.co