AI companies are borrowing hundreds of billions of dollars to buy chips, and Wall Street is supplying much of the money, according to a report from South Korea's Chosun Ilbo. The figure is a scale marker, not a rounding error: the debt now underwriting AI compute purchases sits in the hundreds-of-billions range, and it comes from lenders and investors rather than from the operating profits of the firms doing the buying.

The mechanism matters more than the headline. AI developers need advanced processors in volume, and those processors are expensive and supply-constrained. A company that cannot wait for its own revenue to accumulate can borrow against future revenue instead, buy the hardware now, and train or serve models sooner. Wall Street, in this picture, is not a spectator. It is the balance sheet that makes the purchase schedule possible.

Who is actually paying for the chips?

The buyer at the counter is an AI firm. The wallet, increasingly, is a lender. Chosun Ilbo's framing puts the two parties in a single sentence: AI firms borrow, Wall Street fuels. That division of labor has a simple commercial consequence. The risk of a downturn in AI demand no longer sits only with the chipmakers and their customers. It also sits with the institutions that extended the credit.

Takeaway: the semiconductor demand boom is partly a credit story, and credit has a repayment schedule that model demand may not share.

What does borrowed demand mean for the supply chain?

Chipmakers sell into this demand as if it were cash demand, because at the point of sale it is. A fab planning capacity does not distinguish between a wafer bought with retained earnings and one bought with a syndicated loan. If the borrowing is broad enough — and hundreds of billions suggests it is — it can pull forward orders, tighten near-term supply, and lift prices across the advanced-node segment.

The same mechanism works in reverse. Debt-funded purchases are committed in advance but repaid over time. If AI revenue growth slows before the repayments mature, the same lenders who expanded demand can contract it abruptly.

Takeaway: leverage can amplify a chip boom on the way up and the same boom's correction on the way down.

What to watch next

The number to follow is not the borrowing total itself but its direction: whether credit to AI buyers keeps expanding at this scale, or whether lenders begin tightening terms. That single variable will shape how much of the current order book translates into sustained semiconductor demand, and how much of it was, in effect, rented.