AI’s biggest names were once bystanders in bond markets. Now they’re among the largest, longest-dated issuers in investment grade credit. Rathbone Greenbank Global Sustainable Bond Fund Manager Stuart Chilvers explains why that makes active credit selection more important than ever.
Long ago, before AI dominated what feels like every single corner of financial markets, bond investors hardly batted an eyelid when the ‘Mag7’ acronym first began to be bandied about, before getting back to focusing on the yield curve.
Back then, these gigantic companies might have been becoming an ever more dominant force in equity indices, but they were barely on most bond investors’ radar.
These businesses were generally highly profitable, had little debt (and, as such, were infrequent issuers in bond markets) and so tended to have very high credit ratings.
As a titchy part of the bond universe, mostly trading on very tight credit spreads, they got little attention from bond investors. But, as with almost everything else (or at least it feels that way right now!), AI has changed all that in the last 12-18 months.
From equity market darlings to bond market heavyweights
That’s because many of the Mag7 (primarily Amazon, Google’s parent Alphabet, and Meta) have, alongside Oracle, morphed into so-called AI hyperscalers.
They dominate the arms race to build the computing power, servers and infrastructure needed to train and deploy AI at scale. Every quarter, they seem to revise their capex needs higher.
And it’s become abundantly clear that their operating cashflows alone aren’t sufficient to meet the capital they require to build out AI.
Instead, they seem to be tapping more or less every corner of financial markets to find the investment they require.
The impact on investment grade credit markets has been nothing short of remarkable in both its speed and its scale, even if headline fatigue sometimes dulls the response to the mega-deals that seem to be coming to market with remarkable frequency.
The hyperscalers’ headline issuance numbers, and the weightings they command within bond markets, tell only part of the story.
A lot of the new hyperscaler debt has been issued at longer maturities, making it more sensitive to changes in credit spreads.
As a result, any tightening or widening in credit spreads will have a larger monetary impact on the value of these bonds than on shorter-maturity debt.
If we use DV01 (the metric that measures how much a bond’s price will move if yields were to shift by 0.01%) to provide a risk-adjusted reflection of the US investment grade corporate bond market, Goldman Sachs research shows that, as at the start of May, Oracle is now the largest risk-adjusted name within the US IG universe.
Amazon comes in fifth and Meta eighth (a massive turnaround from 19th and 51st respectively a year before).
Here, there and everywhere…
The US IG market is comfortably the world’s largest and most liquid corporate bond market, thereby enabling these issuers to issue debt here at scale. But this doesn’t mean that hyperscaler issuance has been confined to the US.
Amazon built out a euro curve in one fell swoop earlier in the year, issuing €14.5bn in an eight-tranche deal, while we’ve also seen quite a bit of issuance in the sterling, Swiss franc, Canadian dollar and yen markets. Because they’re smaller, these markets can be altered even more quickly by issuance from the AI behemoths.
For example, Alphabet’s issuance in the sterling market in February means it now accounts for more than 1.4% of the UK corporate bond market off the back of just a single deal (albeit one that contained five tranches, including a century bond).
It is not just the ‘standard’ investment grade market where we have seen issuance, either. As mentioned, capital is being sought in every corner of financial markets. We are seeing issuance through special purpose vehicles (SPVs) – standalone companies set up specifically to hold certain assets like data centres and to issue debt against them.
Because the SPVs don’t sit on the hyperscalers’ balance sheets, they don’t carry the debt themselves – instead, investors are repaid from the cashflows of the underlying assets. The stand-out example is the $27.3bn deal from Beignet to fund Meta’s Hyperion data centre – the largest single-tranche IG bond ever issued.
In addition, AI-related issuance accounts for more than 18% of US dollar high yield issuance in the year to date. In several cases, the hyperscalers are providing backstops, for example, once a data centre build hits a specified milestone.
With these more complex structured deals, the devil really is in the detail: investors need to understand exactly what risks they’re being paid to take.
At the moment, the hyperscalers generally remain exceptionally strong credits. But we think the risks are skewed towards AI-related issuance continuing to get revised higher still in coming quarters.
With the issuers tending to prioritise size when they come to market, we’re wary about the potential spread widening this could cause in existing bonds.
Combined with the sharp rise in concentration in investment grade indices, we think the message to bond investors is clear.
The concentration risks that have played out in equity markets over recent years are creeping into credit markets, but without the uncapped upside potential. That’s why we firmly believe that active credit selection matters more than ever.
Past performance is not a reliable guide to future returns. You may not get back the amount originally invested, and tax rules can change over time. The writer’s views are their own and do not constitute financial advice.
This information should not be relied upon by retail clients or investment professionals. Reference to any particular investment does not constitute a recommendation to buy or sell the investment.
Main image: sean-sinclair-gai1YB3UmDA-unsplash
