AI data center debt has climbed to the top of Wall Street's credit risk watchlist

AI data center debt has become Wall Street's top credit risk, with 48% of fund managers in Bank of America's July Global Fund Manager Survey identifying hyperscaler capital expenditure as the most likely source of a systemic credit event, up from April. Global AI-related debt issuance is tracking toward $570 billion in 2026, four times the prior year's pace, while borrowers remain free-cash-flow negative and debt structures have shifted to less transparent markets, prompting PIMCO to warn of rising default cycles.

A century-old private bond market built for utilities is absorbing a trillion-dollar AI infrastructure debt binge, and one in three fund managers now call it the most likely trigger for the next systemic credit shock. The private placement market, which has quietly financed American railroads, power grids, and investment-grade industrials for over a hundred years, is doing something it has never quite done before. It's financing AI. In May, IREN Ltd., which builds and operates data centers for training AI models, sold roughly $2.1 billion of bonds directly to life insurance firms seeking longer-dated assets to match their annuity liabilities. According to Bloomberg, issuance in this market hit approximately $81 billion through May 2026, the heaviest pace in data going back to 2016. The borrowers are different. The risk profile is different. And increasingly, the people buying this debt are starting to notice. The numbers have grown hard to ignore. Global AI-related debt issuance is tracking toward $570 billion in 2026 alone, according to Forbes, roughly four times the pace of a year earlier. PIMCO pegs AI-related debt issuance at around $100 billion per quarter across hyperscale operators. Hyperscalers issued approximately $121 billion in corporate bonds in 2025, with total data center debt issuance nearly doubling to $182 billion. A parallel $50 billion shadow market has emerged in 144A bonds, private placements that bypass SEC registration entirely, offering institutional buyers a meaningful yield premium in exchange for absorbing project-specific risks including construction, permitting, and tenant concentration. None of that would be alarming on its own if the underlying borrowers were generating cash. They aren't. The companies financing the AI buildout remain firmly free-cash-flow negative, and the debt structures supporting them have migrated steadily toward less transparent corners of the market, where losses are harder to track and underwriting standards have quietly softened. That last part matters. PIMCO's Richard Clarida, Andrew Balls, and Daniel Ivascyn stated plainly in the firm's latest secular outlook that "the default cycle is reasserting itself," with the firm expecting significantly higher losses in leveraged and private direct lending. They flagged shadow default rates rising, payment-in-kind features becoming more common in direct lending books, and AI-related industry disruption compounding the stress. When optimism and dread coexist Bank of America's July Global Fund Manager Survey put a sharper point on how far sentiment has shifted. Forty-eight percent of respondents now identify AI hyperscaler capital expenditure as the single most likely source of a systemic credit event, according to Benzinga's reporting on the survey results. That's nearly double the share from April. The AI bubble has simultaneously vaulted to the top of fund managers' tail-risk rankings, with 45% citing it as the biggest market threat - a 17-point jump from the prior month, past second-wave inflation and a Fed policy error. And yet the same survey found that 61% of fund managers expect no cuts to AI capex this year, and 82% consider long semiconductors the world's most crowded trade. Optimism and dread are coexisting in a way that rarely ends quietly. Here's the thing: equity and credit markets appear to be reading different versions of the same story. Equity investors are still pricing in the AI demand cycle. Credit markets - or at least the sharper participants within them - are starting to price in what happens if the revenue doesn't materialise fast enough to service the debt. Bond demand has softened noticeably, with coverage ratios on hyperscaler bond deals declining and spreads on some data center bonds widening, as Forbes recently noted. Investors who spent two years buying this paper on the strength of the tenant rather than the structure are relearning what infrastructure project finance actually looks like when something goes wrong. Where the debt lives now The structural shift in where this debt sits makes that reckoning harder. Private credit, not the public bond market, now funds most of the marginal AI infrastructure dollar. Blackstone, Blue Owl, Apollo, and others are expected to supply an additional $800 billion in data center financing over the next two years, per S&P Global Ratings. That is a very large number. Off-balance-sheet structures mean public filings no longer tell the full story of who bears the loss if AI returns are delayed - and they obscure it deliberately. That opacity is precisely the feature that makes systemic risk hard to measure until it's too late to manage. For AI infrastructure startups still in the market for non-dilutive capital, the practical consequences of tightening private credit conditions are arriving faster than most expected. The cheap debt window is closing. Life insurers are the marginal buyer in the private placement market. They're beginning to demand yield premiums that reflect the actual risk - not the ambient optimism of 2024. Direct lenders are adding covenants. Underwriting standards won't stay soft when defaults start posting. The same debt machine that financed the AI buildout at generous terms is now quietly repricing the risk it spent two years ignoring. PIMCO's recommendation is simple enough: own quality. The harder question - which neither the bond market nor the fund manager surveys fully answer - is what happens to the companies that built their capital structures on the assumption that credit would stay accessible and cheap. The AI buildout has been called a generational infrastructure investment, and it may yet prove to be one. But generational infrastructure investments have also produced some of history's most spectacular credit events. The private placement market absorbed the railroad boom. It's absorbing this one too, just with different borrowers, different risk profiles, and a growing share of institutional money that has never had to work through a data center default cycle before. Also read: Tencent stock falls 7% as gaming fears and an AI rotation hammer China tech https://startupfortune.com/tencent-stock-falls-7-as-gaming-fears-and-an-ai-rotation-hammer-china-tech/ • Big Tech is spending $725 billion on AI in 2026 and its free cash flow is nearly gone https://startupfortune.com/big-tech-is-spending-725-billion-on-ai-in-2026-and-its-free-cash-flow-is-nearly-gone/ • BlackRock's IBIT posts longest Bitcoin ETF inflow streak since April as CLARITY Act optimism returns https://startupfortune.com/blackrocks-ibit-posts-longest-bitcoin-etf-inflow-streak-since-april-as-clarity-act-optimism-returns/