AI debt sales pick up
Bloomberg reports a surge in AI‑linked debt issuance as investors seek exposure to the AI theme even amid broader market volatility, indicating that capital markets are underwriting AI projects via the credit channel. That suggests some AI investments are being treated like longer‑dated, infrastructure‑style financings rather than short‑term speculative bets. (bloomberg.com)
Artificial intelligence borrowing kept coming this month, with Wall Street arranging tens of billions of dollars for data centers even as broader markets swung sharply. (bloomberg.com) Bloomberg reported on April 11 that investors were still buying high-grade, artificial-intelligence-linked debt as conflict in the Middle East pushed up concerns about energy prices and inflation. The same report said the volatility may be making top-rated artificial intelligence debt look more defensive to some buyers. (bloomberg.com) One of the clearest examples came from QTS, the Blackstone-backed data center company, which launched a $4.6 billion debut bond sale tied to a Microsoft-linked campus in Georgia. Bloomberg said the financing would help fund new artificial intelligence infrastructure, and Moody’s said the build-out could require more than $3 trillion of financing. (bloomberg.com) Another deal hit the market on April 8, when banks including Natixis, Mitsubishi United Financial Group and Société Générale started selling $3 billion of loans for a Meta Platforms-backed data center project in Ohio. Bloomberg said the package covered both the building and related power assets in one transaction. (bloomberg.com) The financing is moving beyond ordinary company borrowing and into project-style structures that look more like toll roads or power plants. The Bank for International Settlements said in March that many artificial intelligence data center deals now use joint ventures or special-purpose entities backed by long-term leases or capacity commitments. (bis.org) That structure lets technology groups add computing capacity without putting every dollar of debt directly on the parent company’s balance sheet. The Bank for International Settlements said most of this issuance has maturities longer than five years, locking in funding for multi-year construction programs. (bis.org) Wall Street has been preparing for a much larger wave. Bloomberg reported on March 16 that Bank of America raised its 2026 forecast for investment-grade debt sales by hyperscalers to $175 billion, and said hyperscaler data-center spending this year could top $600 billion, up about 70% from 2025. (bloomberg.com) Other estimates run higher. Bloomberg Markets reported in February that Morgan Stanley expected $250 billion to $300 billion of 2026 debt issuance from hyperscalers and related joint ventures, with off-balance-sheet project finance becoming a core funding tool for data centers. (finance.yahoo.com) The risks are getting more attention as the sums grow. The Bank for International Settlements said credit-default-swap spreads widened for some hyperscalers, reflecting both the volume of borrowing and uncertainty about how quickly these projects will pay off. (bis.org) For now, the credit market is treating artificial intelligence less like a short software cycle and more like a long construction program, with bonds and loans paying for land, power, buildings and chips over several years. Bloomberg’s April 11 report said that demand was still strong enough to keep those financings moving through a turbulent week. (bloomberg.com)
