Francesco Di Costanzo
Back to shorts

(19) AI Debt Is Moving Into Private Markets

/5 min read

Public capacity is the constraint

AI infrastructure financing will migrate into private markets because public bond investors are likely to reach concentration limits before hyperscalers reach their borrowing limits.

The scale alone can change market structure. Apollo’s 2026 credit outlook says current projections approach $5 trillion of investment over five years, with more than $2.7 trillion of cumulative outlays by five hyperscalers from 2025 through 2029. Goldman Sachs expects the leading technology companies to invest $5.3 trillion in technology and data centres from 2025 through 2030. Its credit analysts expect liquid-market saturation and issuer concentration limits to become somewhat more binding, although they acknowledge that the magnitude remains debated.

The constraint is portfolio construction. A bond manager can consider Alphabet, Amazon, Meta and Microsoft sound credits while declining the next issue because the portfolio already owns too much exposure to the same investment cycle.

Strong credits can still crowd a market

The opposing case has strong evidence. J.P. Morgan Asset Management estimates that large hyperscalers entered 2026 with average net debt equal to about 0.8 times EBITDA, against 2.6 times for the typical investment-grade issuer. It expects their bond issuance to rise from $121 billion in 2025 to $250 billion in 2026. The Bank of England reported that the five largest AI hyperscalers represented only 3% of outstanding US investment-grade debt at the end of 2025. Their bonds had been readily absorbed through the first half of 2026, with credit spreads still compressed.

Those figures explain why the migration will be gradual. They do not remove the concentration problem. By early May, the same five companies already accounted for more than 15% of year-to-date US investment-grade issuance. Apollo calculates that if debt funds only 20% of forecast hyperscaler capital expenditure through 2029, Amazon would become a top-three investment-grade issuer and Meta, Microsoft, Oracle and Google would enter the top ten.

Credit quality answers whether a borrower can repay. It does not determine how much of one correlated theme a diversified fund can hold.

Private capital supplies the missing balance sheet

Private markets can divide the financing requirement by asset, contract and maturity instead of placing another unsecured bond into a crowded index. Goldman Sachs counted more than $1.7 trillion in private infrastructure assets and almost $400 billion of undeployed capital in September 2025. Private real-estate funds held another $2.1 trillion, including $600 billion available for investment. The Bank of England, citing OECD analysis, said private credit’s share of AI financing rose from 9% in 2024 to 34% in 2025.

Recent transactions show the model. Meta’s 2025 filing describes a Louisiana data-centre venture in which it retained a 20% interest while the parties committed roughly $27 billion of development funding. Meta agreed to future leases and residual-value guarantees, giving outside capital contractual claims linked to Meta without placing all the project debt on Meta’s consolidated balance sheet. In June 2026, Apollo announced an initial $35 billion financing for Broadcom’s AI XPV platform, backed by Apollo-managed funds, Blackstone and banks.

The Bank for International Settlements describes the common structure as a vehicle that owns the data centre or equipment, borrows privately, and services the debt with leases or capacity contracts. That structure widens the buyer base to infrastructure funds, private-credit vehicles and insurers.

The risk moves with the debt

Private financing is a release valve, not free capacity. Public markets could absorb more than current estimates if wider spreads, foreign currencies or different maturities attract buyers. High government yields have also made the all-in return on high-grade corporate bonds attractive. The Bank of England found no evidence by July that AI issuance had crowded other borrowers out of public markets.

Private structures still solve a real matching problem. Long-lived buildings can sit in infrastructure portfolios; chips can support shorter, amortising asset-backed debt; contracted leases can suit insurers seeking predictable cash flows. The price is weaker public visibility and a harder assessment of aggregate exposure.

The BIS calls many of these arrangements “shadow borrowing” because debt-like obligations remain outside the technology company’s consolidated balance sheet while guarantees and contracts preserve the economic link. The IMF warns that tenant concentration, obsolescence and connections among banks, private credit and insurers can transmit stress if AI spending reverses.

The practical implication is a change in measurement. Credit investors and regulators should aggregate AI exposure across public bonds, private loans, infrastructure funds, securitisations, leases and contingent guarantees. Moving the financing away from public indices increases capacity only if the new holders bring independent risk budgets rather than another claim on the same balance sheets.

Sources

  1. Goldman Sachs, "Private Markets Are Expected to Have a Growing Role in Data Center Financing" https://www.goldmansachs.com/insights/articles/private-markets-expected-to-have-growing-role-in-data-center-financing

  2. Apollo Global Management, "2026 Credit Outlook: From Scarcity to Selection—The Return of a Buyer’s Market" https://www.apollo.com/content/dam/apolloaem/documents/insights/outlook/apollo-global-2026-credit-outlook.pdf

  3. Bank of England, "Financial Stability Report - July 2026" https://www.bankofengland.co.uk/financial-stability-report/2026/july-2026

  4. J.P. Morgan Asset Management, "Artificial Intelligence" https://am.jpmorgan.com/us/en/asset-management/liq/insights/market-themes/artificial-intelligence/

  5. Egemen Eren, Ingomar Krohn and Karamfil Todorov, "Financing the AI Infrastructure Boom: On- and Off-Balance Sheet Borrowing" https://www.bis.org/publ/qtrpdf/r_qt2603u.htm

  6. Meta Platforms, Inc., "Annual Report on Form 10-K for the Year Ended December 31, 2025" https://www.sec.gov/Archives/edgar/data/1326801/000162828026003942/meta-20251231.htm

  7. Apollo Global Management, "Apollo Leads $35 Billion Capital Solution for Broadcom AI XPV Platform in Partnership with Blackstone and Leading Global Banks" https://ir.apollo.com/news-events/press-releases/detail/629/apollo-leads-35-billion-capital-solution-for-broadcom-ai

  8. International Monetary Fund, "Global Financial Stability Report, April 2026: Global Financial Markets Confront the War in the Middle East and Amplification Risks" https://www.imf.org/-/media/files/publications/gfsr/2026/april/english/text.pdf