Banks get creative with debt as AI borrowing surges past $725 billion

Investment banks are restructuring how they sell corporate debt as AI infrastructure spending forces record volumes into global bond markets, pushing issuers beyond the U.S. dollar and into lease-backed structures.

"Alphabet and Amazon have diversified into other global markets in Europe, Canada, Asia," said Teddy Hodgson, global co-head of investment-grade debt at Morgan Stanley. The large transactions have reshaped global bond markets and established new records for bond sales in euros, sterling and yen.

Amazon raised €14.5 billion ($16.56 billion) in March from an eight-part deal, the largest ever in the euro corporate bond market, according to LSEG. Alphabet smashed records across yen, Canadian dollar, Swiss franc and sterling deals, and sold the first 100-year bond from a tech company since 1997. The two hyperscalers alone have issued $60 billion in multi-currency bonds over the past 12 months.

Capital expenditures for hyperscalers this year are estimated at about $725 billion, according to BNP Paribas, nearly double the level seen in mid-2025. Spending is rising faster than operating cash flow, analysts said, creating the need to access external funding sources. If AI investment continues at this pace, the bond market may need to absorb $1 trillion or more in annual issuance from the sector within two years.

The scale of the shift is visible across the credit spectrum. SpaceX, the newly public space and AI infrastructure company, tapped the bond market for $25 billion in a sale that priced less than two weeks after its record-breaking IPO. The deal drew close to $90 billion in orders across five tranches with 5, 7, 10, 20 and 30-year maturities. The benchmark 10-year notes were priced at just 1.4 percentage points above U.S. Treasuries, a spread typically reserved for single-A or strong triple-B issuers with long, predictable cash flow histories.

Dell Technologies also joined the wave, completing about $3.0 billion in senior unsecured notes maturing between 2031 and 2037 while securing a new $6.0 billion revolving credit facility. The company is expanding its role in AI infrastructure by launching Vera Rubin-based PowerEdge servers and shipping rack-scale AI systems to partners such as CoreWeave.

Data center lease-backed deals gain traction

Beyond traditional bonds, bankers are structuring deals around pre-arranged data center leases to provide more visibility on future cash flows. The latest example was an $810 million note issued by Stingray Compute, owned by Cipher Digital, earlier this month. The offer was nine times oversubscribed, said Cody Gunsch, head of North America leveraged finance capital markets at Morgan Stanley. The lending was backed by a data center lease to Amazon.

Gunsch said the first deals of this kind, with structures inspired by construction loans, began last year and about 15 have since been sold to high-yield investors. The approach allows AI startups and data center operators to raise debt before construction is complete, using pre-arranged tenant commitments as collateral.

The surge in AI-related debt issuance carries implications for bank revenues and credit markets. Underwriting fees from the wave of bond sales are boosting investment banking revenue at firms such as Morgan Stanley, which has led multiple landmark deals. At the same time, the rapid growth in corporate leverage tied to a single theme raises questions about concentration risk. If AI spending slows or a major hyperscaler pulls back, the debt market could face a sudden repricing of risk.

This article is for informational purposes only and does not constitute investment advice.