Goldman Sachs forecasts hyperscaler AI capex rising 50% to $1.2 trillion

Goldman Sachs forecasts that artificial intelligence capital expenditure by the five largest U.S. hyperscalers will rise 50% to $1.2 trillion in 2027, exceeding consensus estimates. This massive buildout, projected to reach $10.3 trillion by 2032, dwarfs historical booms like railroads. To finance this, tech giants are tapping private credit and outside capital, spreading opacity and financial risks to pension funds while raising concerns of an eventual infrastructure oversupply.

Financing requirements and sources

• ▪Columbia Business School professor Stijn Van Nieuwerburgh stated that high borrowing costs and recent interest rate hikes are not deterring technology companies from aggressively developing artificial intelligence infrastructure
• ▪A research paper presented at the Brookings Papers on Economic Activity on September 25, 2026, indicates that financing the $10.3 trillion artificial intelligence buildout will require massive capital from bond markets, banks, and private credit
• ▪Morgan Stanley estimates that major technology companies such as Meta Platforms Inc. will require approximately $2.9 trillion to expand computing capacity through 2028, with more than half of that funding coming from outside investors
• ▪Meta Platforms Inc. financed the majority of its $30 billion Hyperion data center through outside investors, incurring an interest rate at least 1 percentage point higher than on its own debt, adding more than $5 billion in costs

Brookings research paper details

• ▪Columbia Business School professor Stijn Van Nieuwerburgh authored the research paper on artificial intelligence infrastructure financing presented Friday at the Brookings Papers on Economic Activity
• ▪A research paper authored by Stijn Van Nieuwerburgh, presented at the Brookings Papers on Economic Activity, estimates that artificial intelligence infrastructure must generate $3.7 trillion in annual revenue within six years to achieve a 10% return

Risks of outside capital and private credit

• ▪The growing complexity and opacity of artificial intelligence infrastructure financing structures make it increasingly difficult for regulators and investors to track where financial risks are distributed
• ▪Columbia Business School professor Stijn Van Nieuwerburgh warned that the artificial intelligence industry's influx of outside capital, particularly private credit, distributes artificial intelligence infrastructure investment risks across pension funds and sovereign wealth funds
• ▪Columbia Business School professor Stijn Van Nieuwerburgh warned that the artificial intelligence industry's rapid influx of outside capital risks creating an oversupply of artificial intelligence infrastructure, which could lead to a price collapse

Comparison to historical infrastructure booms

• ▪The $10.3 trillion artificial intelligence infrastructure investment boom dwarfs historical United States development periods that built the nation's highways, electric grid, and telecommunications networks
• ▪Columbia Business School professor Stijn Van Nieuwerburgh stated that the current $10.3 trillion artificial intelligence infrastructure buildout is 50% larger than the United States railroad buildout, the second-largest historical infrastructure boom

Debatable claims

• ▪The revenue targets required to justify AI infrastructure spending are unrealistic
• ▪The rapid expansion of AI infrastructure will lead to a market oversupply