One of the biggest questions on investors’ minds in the back-half of 2026 asks if any of the hyper-scalers will guide capex lower.

Not so long ago, investors were prone to bidding up mega-tech shares whenever management reiterated a commitment to throwing ever more money at the AI arms race. The more unequivocal that commitment, the fiercer the bid.

$220 billion (give or take) in new hyper-scaler debt later, “Mr. Market” isn’t so cavalier. I wouldn’t call him discerning either, but visuals like the chart shown below are giving some investors pause.

Yes, valuations are reasonable on an earnings multiple, but we’ve thrown caution completely to the wind on the free cash flow front.

The 90-degree inflection in the hyper-scalers’ P/FCF multiple is, of course, a direct result of plowing money into data centers and access to compute.

Despite growing concerns, just 28% of professional capital allocators expect a capex cut this year, as illustrated by the figure on the left, below, from this month’s installment of BofA’s global fund manager poll.

The figure on the right’s more interesting: Nearly half of survey panelists said hyper-scaler capex, and the associated debt binge, is the most likely source of a systemic credit event.

The same poll found 34% of fund managers saying management should return cash to shareholders, while another 36% said cash flow should be used to “improve balance sheets.” That left just ~a quarter of survey panelists arguing for more capex.

What investors want, they usually get. Which means the next big question might be this one: Will a hyper-scaler capex cut be greeted as good news (i.e., a sign of prudence and discipline) or bad news (i.e., evidence that the ROI on money already spent isn’t as high as management hoped)?