Goldman flags $1.5tn AI debt wave, warns supply pressure hurts credit

Goldman Sachs has cautioned that the heavy pipeline of AI-related bond issuance is creating a significant technical headwind for credit markets, even as the financial health of hyperscalers stays sound. New offerings may need to carry concessions to entice buyers, adding strain on spreads for investment-grade names exposed to artificial intelligence.

The equity capital raises by Oracle and Meta indicate that some of the funding weight is shifting to shareholders, potentially fueling dilution concerns for stocks already battered by last week's downturn. For equity investors, the emphasis has moved from demand to returns, which should drive greater differentiation among AI companies, rewarding those that achieve margin and free cash flow improvement. Combined with calls from wealth managers to diversify, this strengthens the argument for wider market leadership.

Earlier:

• WSJ: Anthropic targets Nasdaq debut next week under ticker ANT despite tech selloff

The AI infrastructure build is increasingly a financing story, and Goldman's view is that balance sheets and returns, not demand, will determine the winners.

Summary:

• Goldman Sachs reports more than $110 billion of AI-linked bond supply announced since October 29, including roughly $25 billion from Microsoft and $10 billion from Oracle
• The bank forecasts AI-related investment-grade financing at about $1.5 trillion over three years, covering bonds, data centre shell financings and chip-backed deals
• Secondary spreads on the Microsoft bond widened five to ten basis points; Goldman cites supply imbalances and concerns over monetisation speed
• Hyperscalers have combined operating cash flow near $740 billion and safeguard ratings via equity, such as Oracle's $10 billion issue and Meta's planned $95 billion raise
• Upside earnings surprises are declining; investors now focus on turning demand into margins, free cash flow and returns
• Goldman remains selective on AI credit and expects capital discipline and return on invested capital to define the next winners

Goldman Sachs anticipates AI-related investment-grade financing will total roughly $1.5 trillion over the coming three years. The bank warns that the flood of debt hitting the market is becoming a strong technical drag on credit, despite healthy hyperscaler fundamentals.

In a note examining AI from a public investing standpoint, Goldman stated that over $110 billion of AI-related bond supply has been announced across multiple currencies since October 29. This includes around $25 billion from Microsoft and $10 billion from Oracle. Future financing is expected to cover a variety of instruments, from unsecured hyperscaler bonds to data centre shell financings and chip-backed transactions.

The market is absorbing the supply, but at a cost. Goldman noted that secondary spreads on the Microsoft deal widened by five to ten basis points after the announcement. The bank attributed such moves to supply-demand imbalances and some uncertainty over how quickly AI spending will generate revenue.

The bank offered more reassurance on balance sheets. It noted that hyperscalers maintain a large cash cushion, with combined operating cash flow of nearly $740 billion, and seem committed to preserving their credit ratings by diversifying funding sources. Goldman pointed to Oracle's $10 billion equity issue alongside its bond deal and Meta's planned $95 billion equity raise as evidence that companies are managing the ratings impact of their capital plans. Nonetheless, it said it remains selective on credit within the sector.

On equities, Goldman said this earnings season has brought diminishing upside surprises as AI spending settles into a steady, multi-year ramp. Strong demand alone is becoming less of a differentiator. Investors increasingly want to know whether that demand can be met and how efficiently it converts into margins, free cash flow and shareholder returns.

Monetisation is uneven, the bank said. Newer consumer AI formats currently earn less than traditional search. Still, Goldman noted that Google's AI Overviews has reached 1 billion users and monetises at core search rates, and that a quarter of Meta's Muse agent users make purchases through it. Enterprise adoption is steady rather than accelerating sharply. Demand for custom AI infrastructure is strong, while uptake of off-the-shelf tools is more gradual.

Goldman concludes that AI is becoming a test of capital and operational discipline as much as of technology. It expects the next generation of winners to be determined by balance-sheet capacity, pricing discipline and return on invested capital. That favours companies that can turn large order books into clean margin expansion.