AlixPartners, the global consulting firm, said project delays, labor shortages, and tightening capital markets are combining with community resistance to new developments to shape that outlook, even as demand for AI computing keeps climbing. "Two-thirds of the market expects distress within 18 months, and the investors and lenders are the most convinced of anyone," said Andrej Danis, a managing director and partner at the firm who co-leads its digital infrastructure practice. "That's not a demand problem. That's a margin problem." The firm's broader outlook work has flagged smaller, less-established neocloud operators as the most exposed to that squeeze, since they carry less balance-sheet cushion than the hyperscalers absorbing most of the capital coverage.
The financing environment described in the survey is already visible in individual deals. QTS, backed by Blackstone, issued $4.6 billion in green bonds in April to finance a major data center site in Fayetteville, Georgia, a project that has faced local opposition. Total bond and loan issuance financing AI infrastructure has already exceeded the full-year total for 2025 partway through 2026, according to reporting cited alongside the AlixPartners survey, illustrating how much of the year's capital-raising has concentrated in a market that executives are simultaneously describing as increasingly fragile.
Bond Issuance Has Gone From $28 Billion a Year to $121 Billion
The scale of the borrowing behind that scrutiny has grown quickly. The five largest hyperscalers, Amazon, Alphabet, Meta, Microsoft, and Oracle, issued a combined $121 billion in U.S. corporate bonds in 2025, according to a January BofA Securities report cited by Reuters, compared with an average of roughly $28 billion a year between 2020 and 2024. BofA expects that group to borrow around $140 billion annually over the next three years, a figure analysts said could exceed $300 billion in some years. Barclays separately forecast total U.S. corporate bond issuance reaching $2.46 trillion in 2026, up nearly 12% from 2025, and attributed the increase largely to AI hyperscaler capital needs rather than the broader refinancing and M&A activity that typically drives issuance.
Nearly 80% of Recent Data Center Bonds Are Trading Wider
That borrowing is now meeting a more selective market. Bloomberg reported this month that more than $500 billion in debt has financed AI infrastructure buildout, and that nearly 80% of data center securities sold since early last year are now quoted at a wider credit spread than at issuance. Spreads widen for many reasons, including supply, duration, and general market liquidity, and this data does not show that investors are pricing in defaults. What it shows is that they are demanding more compensation to hold the debt than they did when it was first sold, which raises the cost of the next round of borrowing for the same projects.
The industry's construction challenges are no longer separate from its financing costs. Delayed equipment, labor shortages, permitting disputes over who bears the cost of new grid infrastructure, and community opposition increasingly influence not only when projects get built, but also the price investors demand to fund them. That is consistent with a broader pattern in how capital is reaching energy and infrastructure projects generally: capital is increasingly rewarding projects that have already cleared their execution risk over development-stage bets on the come, and the same competition for scarce labor and equipment already reshaping how manufacturers and data centers compete for power and skilled trades is now a variable bond investors are pricing directly. As lenders and bond investors become more selective, the industry's execution problems are becoming a credit story.
