Wall Street Now Sees More Than $1 Trillion of AI Spending in 2027. What Has to Go Right
Goldman Sachs expects the five biggest cloud spenders to put about $1.2 trillion into AI next year, up from about $800 billion. Demand, revenue, cash and memory prices all have to cooperate.
Part of the running storyFree Cash Flow Is Now the Scorecard for AI Spenders: What the Evidence SaysAlphabet (GOOGL) on Wednesday and Amazon (AMZN) on Thursday told investors that next year’s AI bill will be bigger than this year’s. Alphabet said its capital spending will rise significantly in 2027, and Amazon Chief Executive Andy Jassy said the company expects to spend “meaningfully more,” Reuters and NBC News reported. Meta (META) said a week earlier that it expects “to continue investing aggressively,” and none of them put a number on it.
Wall Street already has one. Goldman Sachs strategists led by Ryan Hammond forecast in September that Amazon, Alphabet, Microsoft (MSFT), Oracle (ORCL) and Meta will spend about $1.2 trillion on AI infrastructure in 2027, up from about $800 billion this year, PYMNTS and Briefs reported.
Who expects what
Goldman’s figure is the highest of the forecasts we could confirm, but not by much. It sits above the roughly $1.1 trillion consensus the strategists cited, and Goldman expects the pace to cool sharply, to 12% growth in 2028.
| Forecaster | What it counts | 2027 | Said when, and where |
|---|---|---|---|
| Goldman Sachs | AI infrastructure at the five biggest US spenders | About $1.2tn | 25 September, reported by PYMNTS, Briefs and The Decoder |
| Wall Street consensus | The same five, as Goldman cited it | About $1.1tn | 25 September, same reports |
| Evercore and Bank of America | Big Tech AI spending | Above $1tn | After April’s earnings, reported by CNBC on 30 April |
| JPMorgan’s Jamie Dimon | The big cloud companies | About $1tn | CNBC interview on 21 September, reported by PYMNTS |
CNBC summed up where the forecasts stood on Thursday: spending by the biggest cloud companies is “expected to top $1 trillion next year.”
How big that is
Goldman’s team reached for the railroads. Spending in 2027 would be “a larger share of GDP than any technological investment cycle since the railroad build-out in the late 1800s,” the strategists wrote, according to PYMNTS and Briefs. Briefs said the comparison used consensus estimates, not Goldman’s own higher forecast.
JPMorgan Chief Executive Jamie Dimon made a similar point on CNBC in September, describing spending on that scale as “like a 1% increase to GDP each year,” PYMNTS reported.
Today’s base is smaller than the talk suggests. Investment in AI data centers, the computers inside them and their networking came to about 0.8% of US GDP in the first quarter, the research group Epoch AI estimated in June. The railroad comparison describes where the spending is headed, not where it is.
What would have to be true
- Demand has to keep outrunning supply. In July, Jassy said Amazon would not “have enough capacity to meet all the demand we have in 2026,” adding: “And I believe this dynamic will also be true in 2027, too,” Fortune reported. This week Microsoft said outside capacity it had ordered arrived late, and Alphabet’s finance chief, Anat Ashkenazi, said outside computing power “came online only at the very end of Q3,” CNBC reported. Microsoft’s commercial contracts not yet turned into revenue rose 73% to $677 billion, its results release showed.
- Revenue has to follow. Goldman estimates the five need roughly $300 billion a year of AI revenue to break even on their outlays, Briefs and The Decoder reported. Gartner, a research firm, takes a wider view: it says AI data-center owners as a group need $8.5 trillion of revenue by the end of 2030 to cover their capital and financing costs and earn a healthy return, the Financial Times reported, according to EnterpriseAM.
- The cash has to be found. Goldman’s team noted that spending already exceeds the cash the companies’ operations generate, which means more borrowing, Briefs and The Decoder reported. Amazon’s free cash flow over the past 12 months was negative $42.4 billion, its results release showed. Our narrative on the cash scorecard has the rest.
- More dollars have to buy more computers. Part of this year’s rise is prices, not capacity. Amazon raised its 2026 plan in July “owing to higher memory costs,” Fortune reported, and Meta Chief Executive Mark Zuckerberg attributed most of an April increase to component costs, particularly memory, CNBC reported.
Of the four, demand looks the safest today, on the companies’ own words and their order books. Revenue is the least proven: the break-even is an estimate of what has to come, not a total anyone has reported.
What we’ll learn next
Nvidia (NVDA) reports on Tuesday 17 November, with its call at 5 p.m. ET. Its forecast is the first read on what the five are ordering for next year.
The companies will put their own numbers on 2027 with their next results. Microsoft has already set a marker: Chief Financial Officer Amy Hood said spending this quarter will be “over $50 billion,” on the company’s call.
We think the trillion is more likely to arrive than not. Alphabet and Amazon have said outright they will spend more in 2027, Microsoft expects its spending to speed up in the first half of the year, and the bottleneck so far has been supply, not demand. Whether it pays is the open question, and Goldman’s $300 billion break-even is the number we’ll hold the spenders to.
Follow Amazon
Its stories collect under My stocks on the front page. For the daily email, subscribe to the newsletter.
Mostly Markets is for information and education only and isn't investment advice. Marcus Lindqvist is a pen name. Prices are from Nasdaq and CNBC as of the time stated.













