Hyperscalers: Now also a credit story

Hyperscaler debt is becoming a central and relatively defensive way to access the AI investment cycle, with wider spreads offering carry but free-cash-flow recovery after 2027 determining credit outcomes.

• AI-related investment could reach USD 5,500 billion by 2030, while hyperscaler operating cash flow and equity issuance are expected to fund only about one-quarter of the total.
• Investment-grade bonds are expected to provide USD 2,100 billion of external capital, and J.P. Morgan estimates that hyperscalers could issue USD 300 billion of investment-grade debt over the next year.
• Hyperscaler capex is projected to rise from USD 156 billion in 2022 to more than USD 1,000 billion in 2027, while disclosed data-centre lease obligations total around USD 1,400 billion, including roughly USD 1,100 billion currently off balance sheet.
• Hyperscaler spreads have widened about 30 basis points year to date versus 2 basis points for the broader USD investment-grade index; duration-adjusted spreads are around 105 basis points against roughly 80 basis points for the broader market.
• Balance sheets remain stronger than the wider investment-grade industrial sector, and the combined hyperscaler weight in the USD investment-grade index could approach 10% by 2030, increasing concentration and supply risks.

Takeaway: The additional carry is attractive because the repricing reflects supply and complexity more than deteriorating business models, but investors should monitor free-cash-flow recovery from 2028, AI monetisation, capex discipline and off-balance-sheet exposures through active security selection.