AI giants’ debt issuance forecast to reach $420 billion in 2027
Key Facts
Banks and credit investors have increased scrutiny of AI-related debt deals, with some banks reluctant to finance additional transactions, according to a Reuters report on November 9, 2026. That comes as Goldman Sachs Research expects global hyperscaler debt issuance to reach $420 billion in 2027. The estimate is more than 60% above Goldman’s estimate of $250 billion for fiscal 2026. The expected borrowing volume puts the bond market’s capacity to absorb accelerating issuance at the center of the credit debate. It also highlights investors’ focus on whether companies can turn AI spending into returns that support debt servicing.
Goldman’s earlier estimates reflect the growing reliance on borrowing to finance AI infrastructure. On August 5, 2026, the firm said AI-related debt issuance had approached $500 billion from the start of the year through that date. Hyperscalers accounted for about 40% of the issuance Goldman tracks. Those companies issued $108 billion of debt globally in 2025, followed by $194 billion in the period through August 5, 2026. The comparison points to accelerating borrowing during the year, within a broader market that includes other parties financing AI projects and data centers.
Goldman forecast that debt would fund about 33% of hyperscalers’ capital spending in 2026, rising to 35% in 2027. In addition to companies’ own borrowing, the firm said financing for projects and data centers could reach $300 billion in 2027. This adds funding sources to the wave of direct issuance and increases the importance of the credit market’s capacity to absorb it. Goldman warned of a gap between what hyperscalers can issue while maintaining investment-grade ratings and what the bond market can readily absorb. “There is a big difference between how much the hyperscaler group can issue while remaining investment grade and how much the bond market can readily absorb.”
Yield spreads show a greater widening in AI-related technology company debt than in high-grade industrial companies with no AI exposure. According to Reuters, the spread on 10-year hyperscaler debt widened by about 50 basis points from the start of 2026 through November 9, compared with about 20 basis points for comparable industrial companies. The divergence means investors repriced credit risk more sharply in the AI-related segment. The report followed a selloff that erased $2.5 trillion from the Nasdaq 100’s market value on the previous Thursday and pushed the index into a correction. Reuters linked the climate to increased scrutiny of sector-related debt deals by banks and investors.
Oracle highlights the sensitivity of financing to credit developments, according to an Axios report published on November 3, 2026. The company told investors it would cap new capital-markets financing at $20 billion through fiscal 2030. It then filed to raise $20 billion, half through equity and half through bonds, according to Axios. In the week before the report, five-year credit default swaps on the company’s debt stood at 294.4 basis points, after being below 100 basis points the previous year. The widening spread reflects a higher market cost of insuring the company’s debt over that period.
Convertible corporate bonds posted their worst monthly performance since 2008 in October 2026, falling 6.4%, according to Axios. The market pressure coincides with a time lag between AI companies’ infrastructure spending and their ability to generate income from it. Goldman estimated that the gap could range from three months to two years. The firm said assumptions about AI monetization and the viability of broader expansion would continue to evolve, as would credit spreads and valuations. The ability of these investments to generate revenue and cash flow therefore remains central to assessing the risks of related borrowing.
The physical expansion of data centers adds execution challenges to the financing question. On October 19, 2026, Goldman forecast global electricity demand from data centers would rise 170% between 2025 and 2030, with the United States accounting for more than 60% of the increase. It estimated that the average time from a project’s grid-connection request to commercial operation is about five years, while sourcing high-voltage substations and step-up transformers could take about three to five years. On October 28, Goldman reviewed earlier estimates that AI infrastructure investment would range from $3 trillion to $5 trillion over five years, and cited a Goldman Sachs Global Institute estimate of as much as $8 trillion over four to five years. The outlook for upcoming debt issuance remains tied to the course of these investments and markets’ ability to finance them.