A Credit Perspective on the AI Buildout

Dan Toscano · Morgan Stanley

AI financing is not raising the cost of capital; spreads are at or near all-time tights.

Spreads are tight, and Toscano says they were at all-time tights over the course of 2026, so he finds it hard to argue that AI-related financing is driving up borrowing costs: 'We don't see it in spreads.' He notes that rates move independently and are driven by many factors, many of them positive. The backdrop is a strong economy with issuer earnings growing in double digits quarter after quarter despite energy instability and consumer pressure. He says that earnings strength 'can change', so the conclusion rests on current conditions.

“I mean, look at where spreads are today. They are, in my opinion, tight. ... it's hard for me to sit here and say, all this AI related financing is driving up the cost of capital. We don't see it in spreads.”at 4:03

Dan Toscano · Morgan Stanley

The market impact of trillions in AI-related debt supply has not yet materialised.

Toscano says the debate over the trillions of dollars that must be raised across markets is ongoing, but its effect has yet to show up in the market. He argues that volume is coming at scale and has not yet arrived in full. He points to the non-investment-grade market, where he spends most of his time, as having been quiet for months. The supply shock is therefore still ahead rather than already priced in.

“I think the debate around the impact of the trillions of dollars that need to be raised in all markets has yet to play out in the marketplace. ... volume is coming in scale.”at 4:24

Dan Toscano · Morgan Stanley

Data center project financings hinge on lease and cash-control terms that define who bears risk.

Toscano argues that AI financing is essentially project finance, so the details matter far more than in ordinary leveraged lending. Investors need to read the lease terms and ask what can go wrong, what happens if it does, and who takes which risk. Project finance and structured credit for toll roads and power plants have long worked this way. The key protections are how and when cash is dispersed and who controls those decisions, so that cash does not drain out of the system if something goes wrong.

“details matter in all caps... the fundamental question that underpins Details Matter is what can go wrong? And if something goes wrong, what happens? And who's taking what risk?”at 5:24

Dan Toscano · Morgan Stanley

A slowdown in AI capex is a delay, not a credit problem, given how the spending is managed.

Toscano says capex is measured against companies' own ambitions. If spending slows relative to those ambitions, it 'doesn't really create a problem', because it only means 'gratification delayed'. He thinks the pace can speed up or slow down as quickly as investors are comfortable funding management teams, and that companies can adjust chip and compute purchases without consequence. He cautions that once capital is committed, construction takes 12 to 24 months and the cash flows must be managed through it. That assumes a friendly environment for building data centers, which is where the real risk sits.

“I think the pace of CapEx spend gets measured against companies' ambitions, not anything else... if it slows down relative to someone's ambitions, it doesn't really create a problem in my mind. All it means is gratification delayed.”at 6:46