Earlier this week, I mentioned that forecasts for hyper-scaler capex in 2026 are up more than $20 billion already this year, which is to say up that much versus where consensus stood headed into earnings season.

Wall Street expects the five largest hyper-scalers to throw $561 billion at the AI buildout over the next 12 months, up nearly 40% from the $406 billion they shelled out last year.

Capex estimates will be revised higher, probably by a lot and beginning as soon as this week, with results from Alphabet and Amazon due. As you ponder the outlook for AI outlays, don’t forget that some of that spending will come at the expense of buybacks.

I realize that’s an ad nauseam talking point, but I haven’t mentioned it in a few weeks. The figure below, from the latest Charlie McElligott dispatch, is a helpful reminder of the extent to which this represents a tectonic shift.

As you can see, we’re going from a place where S&P 500 companies devoted almost half their operating cash flow to share repurchases to a place where nearly that same share goes to capex instead.

This is one vector on which the AI story’s going “from tailwind to headwind,” as McElligott put it. (Another’s the cannibalization dynamic going on inside tech, where AI disruption fears are bleeding SaaS like a stuck pig with severe knock-on implications for private credit and private equity.)

“The ‘cashflow for capex burn’ and even more expansive funding needs for the hyper-scalers [have] major implications for credit issuance” and constitute a “tangible” threat to buybacks, Charlie went on, noting that the dialed-up IG issuance is a spread widener impulse “just on the magnitude alone.”

That latter point’s important: It’s not about credit quality — well, maybe it is for Oracle, but not for the rest of the hyper-scalers — it’s about supply and demand. Even if demand’s voracious for high grade corporate debt (and it is), when you ask the market to underwrite magnitudes more supply, that’s going to come at a cost.

Elaborating on the “tailwind-to-headwind” point mentioned above, McElligott wrote that between “credit spread widening, a negative buyback impulse and the prospect of a rebalancing out of crowded legacy positioning in secular growth themes,” there’s now a “real risk to stocks.”

After all, we’re accustomed to just the opposite of those dynamics. For more than 10 years, McElligott said, risk assets benefited from “tight credit spreads [and] massive perpetual passive equity demand from corporate buyback flows,” while more recently, the “extreme concentration [in] AI themes” drove cap-weighted benchmarks ever higher almost single-handedly.