Large And Difficult To Frame – AI Issuance or [Risk and Value in AI]

This is an excerpt from the Mawer ‘Art of Boring podcast’ and features Sandro Morassutti and Marty Lee, both credit analysts.

Sandro Morassutti – A year ago we discussed the theme of AI-linked credit risk. Today, issuance has definitely shaped the market which is now much bigger.

In a McKinsey report they projected a US$5.2 trillion of data centre spend over the next five years. If we assume that roughly half is going to be debt-financed, that’s about US$500 billion per year in new bond supply.

To put this into perspective, that’s roughly the entire US corporate bond supply’s issuance on a net basis in an average year. So, this one sector and one trend could double corporate bond supply overall in the US.

In early July, we saw Amazon, Alphabet, Meta, and Oracle all come to market and issue US$194 billion of new bonds. That’s versus about US$108 billion in all of 2025.

Today, US corporate supply is running at about US$1.7 trillion per year, and that’s up 27% year on year. For context, back in 2020, total issuance was US$1.75 trillion, which was a record year. We expect a new high watermark in 2026.

The size of this new issuance is difficult to frame.

In terms of individual trades, Amazon raised US$37 billion across 11 tranches in a single day in March.

In February, Alphabet issued £1 billion in a 100-year bond as part of a broader US$31 billion offering. These are sovereign sized transactions from companies who have not participated in markets anywhere near this scale in prior years. You would expect the level of supply to push spreads out, but this has not been the case. Spreads continue to trade at the tights for example investment grade spreads are roughly 80 basis points. High yield spreads are 14 basis points higher to date even with 1.7 trillion U.S. dollar supply that has come to market.

So far the market has been able to digest all of this additional supply without any major issues. The benchmarks have widened slightly, so overall yields are higher, driving demand through waves of supply.

Marty Lee – It’s still early days in terms of AI build-out, but there are no signs of any slowdown. Given how much Treasuries have moved and how little spread compensation there is, we continue to lean towards higher-rated credit and short duration.

We still see value in some areas and participate in new deals from Amazon and Alphabet. These are rare opportunities where AA-rated companies are pricing wider than single A-rated companies.

We have added lower-rated credit, but more in one-off situations.

Supply will keep coming, and we expect spreads to widen. We would rather have the opportunity to invest in future than be fully invested at this point.

The biggest issue with AI is issuing in the long end, that is, the 10- to 30-year bucket. This issuance is competing with government sovereign issuance. We don’t really know what the world will look like in 30 years. Will these AA companies still be rated AA?

In summary, we think we’re being compensated in the short end and don’t need to go out along the yield curve. It’s a lot to try and underwrite what the future will look like. We believe the best values are in short dated, high quality credit and that investors don’t need to take on outsized credit or duration risk.