Goldman Sachs projects Big Tech will pour $1.2 trillion into AI infrastructure by 2027, far exceeding Wall Street's forecasts

Five Big Tech giants are on track to pour a combined $1.2 trillion into AI infrastructure in 2027, according to Goldman Sachs. The estimate covers Amazon, Alphabet, Microsoft, Oracle, and Meta and exceeds Wall Street's consensus of $1.1 trillion by roughly $100 billion. If it materializes, the scale of investment relative to GDP would be the largest since railroad construction reshaped the American economy in the 19th century.

The Numbers Behind the Forecast

Bloomberg reported the figures, citing Goldman Sachs strategist Ryan Hammond. The $1.2 trillion projection for 2027 sits more than 50 percent above the roughly $800 billion expected to be spent this year. It is not a sudden reversal for the bank: back in June, Goldman had already warned that consensus estimates for AI infrastructure spending were far too low.

The headline number comes with a caveat, though. Growth is decelerating sharply. After spending nearly doubles in 2026 at close to 100 percent, the pace drops to 54 percent in 2027 and then falls further to 12 percent in 2028. In other words, 2027 may mark the peak of the buildout's expansion rather than the start of a new acceleration.

Can the Revenue Catch Up?

Recouping $1.2 trillion in capital expenditure requires serious income. Goldman calculates the five companies would need about $300 billion a year in AI revenue to justify the outlays. Current earnings still fall short of that bar.

There are, however, signs the demand side is moving in the right direction. Cloud revenue growth jumped from 25 percent in 2024 to 48 percent in Q2 2026, a sign that enterprise adoption of AI services is accelerating. The bigger open question involves the AI labs themselves. Goldman notes it remains unclear whether revenue growth at companies like OpenAI and Anthropic is fast enough to validate the spending, and both firms sit at the center of the expectations—and the financial instruments—propping up the entire buildout.

Debt, Power, and Other Bottlenecks

Goldman also flags a structural shift: spending by these companies now exceeds what they generate from ongoing operations. That gap implies more debt financing, which raises the stakes if AI revenue disappoints.

Physical constraints could slow the boom as well. The bank points to bottlenecks in power, labor, and memory chips as factors that could further limit how fast data centers come online.

With 2027 spending set to peak and growth cooling to 12 percent by 2028, the next signal to watch is whether AI lab revenues—and cloud growth—keep climbing fast enough to close the $300 billion annual gap Goldman says is needed.