Goldman Sachs and Morgan Stanley dissect the AI debt binge as the credit party cools
Wall Street's two biggest investment banks see record AI borrowing, rising hyperscaler leverage, and investors getting pickier about who they lend to
The AI boom runs on chips, power, and data centers. Increasingly, it also runs on debt.
New analyses from Goldman Sachs and Morgan Stanley show AI-related borrowing has hit levels that would have sounded fanciful a year ago. Both banks also flag a shift in mood. For some borrowers, the credit party for AI funding has already ended.
The numbers behind the borrowing spree
Goldman Sachs estimates that $489 billion in AI-related debt has been issued so far in 2026. That already sits above the $322 billion figure Goldman had forecast in 2025.
Morgan Stanley takes the view out to year-end. The bank projects nearly $570 billion in global AI-related debt issuance for full-year 2026.
The driver is no mystery. Hyperscalers, the giant tech firms that run cloud platforms and train frontier models, are pouring money into data centers, semiconductors, and power infrastructure. Morgan Stanley expects hyperscaler capital expenditure needs to likely exceed $1 trillion in 2027.
The growth is even sharper in the riskier corners of credit. Goldman reported AI-linked leveraged finance issuance at $88 billion year-to-date in 2026. In the same period last year, that figure was $20 billion.
High-yield AI infrastructure supply tells a similar story. It reached $40 billion in 2026, already more than the $12 billion total for all of 2025.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Leverage is climbing fast
According to the analyses, aggregate gross leverage among hyperscalers has surged from approximately 0.9x to 1.8x in the last six months. Gross leverage compares total debt to earnings. A doubling means these companies now owe roughly twice as much relative to what they generate as they did half a year ago.
The analyses point to off-balance-sheet liabilities totaling approximately $3 trillion, including lease and purchase commitments.
Why investors are getting choosier
Both Goldman and Morgan Stanley describe diminishing risk appetite, particularly for lower-rated borrowers. Investor interest is clustering around investment-grade offerings, the debt issued by companies with the strongest credit ratings.
In practice, that creates a two-speed market. A top-rated hyperscaler can still raise enormous sums with ease. A smaller, more leveraged AI infrastructure player may find lenders asking harder questions and demanding higher yields.
What this means for the AI buildout
The most immediate implication is a bifurcated credit landscape. Investment-grade tech giants look well positioned to keep funding their capex plans. The pressure falls on lower-rated borrowers, many of which rely on debt markets far more than the hyperscalers do.
High-yield AI infrastructure issuance is already more than triple last year’s total. The leverage trend is worth watching closely. A move from about 0.9x to 1.8x in six months is not a crisis on its own. But if hyperscaler capex does climb past $1 trillion in 2027 as Morgan Stanley expects, that ratio has room to keep rising.
The roughly $3 trillion in off-balance-sheet commitments adds another layer of uncertainty. If analysts and rating agencies start treating more of those commitments like debt, perceived leverage could jump further.