Washington, D.C./ Science, Tech & Medicine

Goldman Sachs Sees AI Spending Hitting $1.4 Trillion by 2027, Nearly 4% of US Economy

AI Assisted Icon
Published on November 14, 2026
Goldman Sachs Sees AI Spending Hitting $1.4 Trillion by 2027, Nearly 4% of US EconomyData Center Servers
Domaintechnik / Unsplash

Annualized U.S. spending on AI-related hardware and intellectual property is on track to climb to $1.4 trillion by the end of 2027, equal to nearly 4% of the entire American economy, according to a new forecast from Goldman Sachs. The investment bank says artificial intelligence will remain the single biggest driver of U.S. business investment growth heading into 2027, even as rising borrowing costs and tighter investor scrutiny loom over the sector.

The forecast, reported by The Sacramento Bee, projects U.S. business investment growth at 8% on a quarterly basis in 2027 — nearly double the consensus forecast of 4.2%, according to Channel NewsAsia. Goldman estimates that AI-related investment alone will add 0.6 percentage points to U.S. GDP growth next year, a figure that includes semiconductor investment.

That semiconductor piece matters because of how it's counted. Goldman says semiconductor investment is undercounted in official business-investment statistics, and most information-processing equipment is imported and therefore netted out of GDP entirely. Strip semiconductors out of the picture, and the bank's math changes substantially: AI-driven investment would still boost 2027 capital-expenditure growth by 4.3 percentage points, but the GDP contribution shrinks to just 0.2 percentage points.

Debt Is Fueling the Build-Out

Much of this spending boom is being financed with borrowed money. Companies are increasingly relying on debt-funded expansion to keep pace with AI infrastructure demands, the same report notes. Goldman Sachs Research separately projects global hyperscaler debt issuance will reach $420 billion in 2027, more than 60% above its estimate for fiscal 2026, per the bank's own analysis in Navigating a System of AI Bottlenecks.

Goldman also expects hyperscalers to issue roughly $400 billion in global investment-grade bonds in 2027 against an estimated $1.14 trillion in capital expenditures — about a 35% share funded by debt, according to Yahoo Finance. The same report notes that the five leading hyperscalers — Meta, Microsoft, Alphabet, Amazon and Oracle — reported combined capital expenditures of $405 billion in 2025, a figure Goldman expects to balloon to $750 billion by the end of 2026.

Lenders Start Getting Nervous

Not everyone is comfortable with how much of this build-out rides on borrowed capital. Banks and credit investors are upping their scrutiny of AI-related debt deals while showing reluctance to increase exposure to borrowers, according to Finance & Commerce. That outlet reports JPMorgan had $29 billion in loans backed by other income-producing properties — including data centers — outstanding at the end of the third quarter, representing 12% of its commercial real-estate exposure.

Some AI-focused firms are already feeling the squeeze differently than the tech giants. JPMorgan's James Sullivan said neocloud companies face substantially higher costs for additional capital than hyperscalers because their borrowing spreads have widened, according to CNBC. Nebius illustrated that shift firsthand: the company abandoned plans for a high-yield bond issue it had explored in September and instead entered a $3.5 billion senior secured loan facility with Citigroup in October for data-center development, GPU procurement and general corporate purposes, Investing.com reports.

Midterm Results Add a Wrinkle

The forecast lands just as the political landscape shifted. Democrats won U.S. midterm elections earlier this month, and their victories raised concerns about tighter regulation of data-center development. Still, Goldman Sachs assessed that the election outcome poses only limited risks to AI investment, per the Channel NewsAsia report.

Markets reacted quickly to the midterm results regardless: Microsoft and Amazon shares each surged more than 6%, while AI infrastructure provider CoreWeave plunged around 10%, CNBC reported. Nomura, cited in the same CNBC piece, said much of the data-center capacity expected to come online in 2027 is already under construction, while projects planned for 2028 face more exposure to permitting delays and regulatory uncertainty.

Bigger Numbers, Bigger Questions

Goldman's broader infrastructure estimates dwarf even its 2027 projections. The bank's research arm estimates aggregate AI capital expenditure of about $7.6 trillion across compute, data centers and power between 2026 and 2031, and projects global data-center power demand will rise 170% from 2025 to 2030, with more than 60% of that growth coming from within the United States. Utility capital plans tied to that buildout are expected to total $1.3 trillion between 2026 and 2030, per the same Goldman Sachs research.

Not every forecaster shares Goldman's confidence that the spending pace holds steady. Capital Economics expects hyperscalers to spend $800 billion on AI in 2026 and at least $1 trillion in 2027, but it projects that investment will keep rising at a slower pace before falling outright by around 10% in 2028, according to Fortune. Whether the AI investment boom proves durable or starts cooling by 2028 remains an open question that will shape how much of Goldman's $1.4 trillion projection actually materializes.