Each point on the 10-year yield would add $41 billion a year to AI's $4.1 trillion debt bill

JPMorgan's $4.1 trillion is cumulative issuance through 2030, and not all of it will be priced at this week's yields ^[2]. The 10-year has risen about one percentage point since the start of the year ^[3]. Applied to the full amount, one point is about $41 billion a year of extra interest once the debt is out, an upper bound for each point the yield climbs ^[1].

A higher coupon slows the building only if borrowers mind paying it. SoftBank paid as much as 9.75% on the seven-year tranche of its $11.1 billion junk sale this week ^[6]. That is roughly 4.6 points over the 10-year yield (the maturities differ, so the gap is approximate) ^[2]. "They basically are price insensitive to that raise, which means they're price takers," said Mark Malek, chief investment officer at Siebert Financial ^[7]. "In my view, a lot of these companies need to be price insensitive. They need to get as much capital as possible to compete," he said ^[8].

A price taker does not cut back when the coupon rises. So the pressure shows up in whether lenders say yes at all. Riley Thompson, a vice president at Mitsubishi HC Capital America, said lenders are getting pickier about the projects they fund even when the borrower agrees to pay a higher rate ^[10]. "Instead of a roster of 50 neoclouds, there's probably 20 that the market's truly interested in," Thompson said ^[11]. A senior private credit investor, who was not named, told CNBC that neocloud deals will get harder to finance because those companies have less cushion to absorb costs ^[9]. Amazon, Google, Meta and Microsoft borrow on investment-grade ratings ^[12], and their capital spending is expected to rise again in 2027 ^[13].

CoreWeave is the company the lenders' argument points at. Its latest quarterly filing says each 100-basis-point rise in rates could add $30 million to interest expense on its floating-rate debt, as of June ^[14]. Divide $30 million by one point and the floating balance comes to roughly $3 billion ^[3]. Its shares still rose almost 8% this week ^[4]. Oracle fell 7% for the week and is down about 30% this year ^[5]. The slide followed a Bloomberg report that Oracle had sent a force-majeure notice on its Project Jupiter campus in New Mexico, seeking to delay payment if the site misses its expected 2028 start ^[15]. Oracle said the project "remains on our planned schedule" ^[16]. One week of equity trading after a company-specific report is weak evidence that the bond market separates the two, and CNBC's report does not include credit spreads for either company.

This week's evidence fits more than one outcome. Rates on CoreWeave's floating debt could keep climbing, at $30 million per point ^[14]. Lenders could keep funding the top tier and stop taking calls from the rest, as Thompson describes ^[11]. Or permits could bind before rates do: Texas Governor Greg Abbott ordered a temporary halt to data center environmental permits on Monday ^[17], and 69% of respondents to an NBC News Decision Desk poll oppose a data center in their area ^[18].

I think the rationing matters more than the coupon. A borrower that must compete for capital will pay 9.75%, and one that cannot find a lender stops building at any rate. The counter-case is SoftBank itself, which raised $11.1 billion at junk yields in a single week ^[6]. If neocloud financings outside Thompson's 20 close in size over the next two quarters, the rationing view is wrong and the higher coupon is the whole cost.