Goldman Sachs Projects $1.2 Trillion AI Infrastructure Spree by Big Tech in 2027

Goldman Sachs has raised its projection for AI infrastructure spending by the largest technology companies, estimating that Amazon, Alphabet, Microsoft, Oracle, and Meta will collectively invest $1.2 trillion in 2027. That figure is more than 50 percent higher than their combined spending this year, according to the bank's latest analysis. The forecast puts the five firms' capital outlays far above what most Wall Street analysts have penciled in, and it signals that the AI buildout is accelerating rather than plateauing.

The scale becomes clearer when measured against the broader economy. Goldman notes that as a share of GDP, this spending wave would be the largest investment cycle since the railroad construction boom of the 19th century. That comparison underscores how much of the current economic expansion is tied to data centers, chips, and the power infrastructure needed to run them. The five companies involved are already among the most valuable on earth, and their combined budgets now rival the industrial pushes of earlier eras.

The bank's estimate is not a simple extrapolation of current trends. It reflects expectations that demand for AI training and inference will keep rising, forcing these firms to keep buying GPUs, building server farms, and securing long-term energy contracts. Microsoft, Amazon, and Alphabet have all told investors in recent quarters that capital expenditures will climb, while Oracle has staked its cloud strategy on AI workloads and Meta has been open about its aggressive GPU purchases. Goldman's number suggests the market has not fully priced in how far these commitments could go.

Three bottlenecks could slow the pace, according to the analysis: power, labor, and memory chips. Electricity grids in key data-center regions are already strained, and utilities face years-long lead times for new generation and transmission. Skilled construction and electrical workers are in short supply, and high-bandwidth memory used in AI accelerators remains tight. Any one of these constraints could push spending past 2027 or force companies to redirect budgets toward energy and supply-chain fixes rather than raw compute.

For the AI industry, the projection matters because it sets a floor under demand for chips, networking gear, cooling systems, and power contracts. Nvidia and its suppliers would be the most immediate beneficiaries, but the ripple effects extend to utilities, construction firms, and memory makers. It also raises the stakes for companies that cannot match the spending pace, potentially widening the gap between a handful of hyperscalers and everyone else.

The open question is whether the returns will justify the outlay. Goldman's forecast assumes AI revenue growth continues at a clip that covers these investments, but if model capabilities plateau or enterprise adoption slows, the same capital cycle could look like overbuilding. Investors will watch quarterly capex guidance and power procurement deals for signs of whether the $1.2 trillion figure holds up or gets revised again.