Are credit rating agencies getting fed up with hyperscalers?

S&P Global Ratings credit outlook reports tend to be full of measured and nuanced credit geekery. So when we read the ratings agency’s latest hyperscaler ‘temperature check’ we almost fell off our chairs.

You know the context: hyperscalers have spent vast sums on capex projects to tool up AI models and build out data centres, and are planning many more.

S&P reckons a full half of the economic growth coming from the US private sector was linked to AI-centric activity over the past year. And they assume the top six US hyperscalers will collectively spend more than $7tn in the five years through 2030. So the fate of the US economy, not to mention the stock market, credit, infrastructure and real estate, is increasingly tied to the AI show staying on the road.

And the AI show is staying on the road, right? S&P:

Every time we take a deep dive into this sector, we find that capex is rising faster than we anticipated, financings are becoming more complicated and less transparent, and that returns on investment will take years to realize.

Every time we take a deep dive into this sector, we find that capex is rising faster than we anticipated, financings are becoming more complicated and less transparent, and that returns on investment will take years to realize.

🔥🔥🔥

The credit rating process involves analysts diving deep into the workings of companies, typically with access to a ton of non-public information. But the authors write that, despite AI return on investment being critical to the rating judgments, “the big six hyperscalers don’t quantify their returns on investment on AI”. Maybe, the report’s authors speculate, this is because it’s difficult to work out. Or maybe, “simply, entities just choose not to share that data.”

Left with a rather large and conspicuous missing piece to their rating puzzle, the agency has to assume something. So they tell us that they assume — perhaps not unreasonably — that the largest companies in the world probably know what they’re doing and that their current and future investment all works out. But they do set out some guardrails on their ratings. Because even investment-grade-rated companies can’t keep piling on debt, leasing commitments and power purchase agreements without limit.

Here’s a chart showing how much debtlike commitments each of the top six US hyperscalers are projected to be running over the next few years as a multiple of EBITDA:

Different companies have different thresholds based on their distinct businesses and also their ratings. The amount of leverage you can take while still preserving the very top triple-A rating is a lot less than the amount that breaches what S&P Global Ratings calls its downgrade thresholds into junk territory.

So Microsoft can build debt, lease commitments, the works, that sum to only a year of EBITDA before its rating is under threat. Meanwhile, Oracle can build adjusted debt to the tune of 4.5x EBITDA before the downgrade threshold is breached on its BBB- rating.

It looks from the chart that Amazon is on track to lose its AA rating unless earnings pick up or capex slows down by 2027. But Alphaville understands that it’s not that mechanistic. Rating decisions are all made by committee, and there could be ameliorating circumstances that justify no cut.

And here’s a chart showing how much headroom each company has for EBITDA disappointment or more debt / lease commitments / power purchasing agreements, without tripping downgrade thresholds:

Alphabet stands out for quite how much room it has in its current AA+ rating. But for some hyperscalers, the path to get from ‘we hope this works out’ to ‘it’s worked out!’ without getting downgraded looks pretty thin. For double-A-rated Amazon to move to high single-A wouldn’t be much of a disaster. But in Oracle’s case, getting downgraded means getting junked. And with around $117bn of index-eligible US dollar bonds, that would be quite the event.

Moreover, just stating nominal dollar headroom numbers doesn’t really convey how S&P’s base case assumption — that these companies know what they’re doing and all the investment makes its way into lovely, lovely ebitda — sets up quite a high ebitda growth bar against which each of their progress will be measured.

The lines in the chart show how much ebitda is projected to grow versus last year’s under the rating agency’s base-case projections, and how much or little room there is before any ebitda misses translate into difficult conversations with credit rating analysts (assuming that we hold other things — like capex, lease obligations, forward power purchase commitments, etc — steady).

Normally at this point, Alphaville would go to the effort of snarkily grouping all the risks outlined in the report into a Rumsfeldian quadrant. But, emojis aside, S&P did the work for us:

Regular readers will recognise a number of these themes, from circular financing, through to risk of overbuilding. We’ve included a link to Joachim Klement’s Substack, in which he discusses the risk from open models just to round things out.

The report also features a circular financing taxonomy, which includes residual-value guarantees, take-or-pay agreements, chip financing, lease liabilities, power purchase agreements, backstop guarantees, lease guarantees, and direct equity investments. Does S&P think this might prove problematic?

Interconnected financing structures could amplify volatility if demand weakens unexpectedly. The failure of one entity would have implications for entities that have lent it money, have guaranteed the value of its assets, or are just expecting payment for goods delivered.The broader question is whether circular financing is creating leverage collectively that is individually manageable but could become highly correlated should demand fall sharply. . . . The scale of overlap and interconnectedness is vast. In a downturn, even the best capitalized and most profitable firms may incur substantial pain.

Interconnected financing structures could amplify volatility if demand weakens unexpectedly. The failure of one entity would have implications for entities that have lent it money, have guaranteed the value of its assets, or are just expecting payment for goods delivered.

The broader question is whether circular financing is creating leverage collectively that is individually manageable but could become highly correlated should demand fall sharply. . . .

The scale of overlap and interconnectedness is vast. In a downturn, even the best capitalized and most profitable firms may incur substantial pain.

Better hope that base case works out.

Further reading:
— Nvidia’s $200bn ‘balance sheet-as-a-service’ (FTAV)
— Just how big is the hidden leverage of AI hyperscalers? (FTAV)
— If this is true, the hyperscalers are toast (Klement on Investing)
— SpaceX considered as a leasing company (FTAV)
— SpaceX bond yields rocket towards junk (FTAV)