Gartner Forecasts $4T in AI Spending as Its Keynote Tells CIOs to Hit the Brakes
Infrastructure takes 59% of Gartner's 2027 AI forecast. Its keynote says buyers see productivity gains of 16%, against the 50% or more vendors claim.

TL;DR
- Gartner forecast on November 12 that worldwide AI spending will reach $4 trillion in 2027, up 47.5%.
- Its keynote coaches CIOs to treat vendor pushes for more AI consumption as a risk to manage.
- Gartner expects spending on AI agents and assistants to more than double in 2027, to $67 billion.
Gartner forecast a 47.5% rise in AI spending for 2027 on November 12, from its IT Symposium in Barcelona. Three days earlier, the same conference's opening keynote told CIOs to rein in AI.
Gartner's forecast puts the 2027 total at about $4 trillion. Infrastructure takes $2.4 trillion of it, or 59%. Gartner Fellow Daryl Plummer, who gave the keynote with Kristin Moyer, has told CIOs to apply the brakes on AI and said vendors want buyers to buy more tokens.
AI sellers who grow on usage are pitching CIOs whose analyst firm tells them to limit "careless consumption." Plummer called a vendor's push for more tokens "bad momentum" for the buyer.
Plummer says a successful AI project costs 18 times more than two years ago
Only one AI project in five succeeds, Plummer said, unchanged from last year, while 58% of more than 1,000 functional leaders in a Gartner survey report some positive value. In ITPro's report, Plummer tied the cost to that one project: "you're spending 18 times more to get it than just two years ago." The 18x figure was in the opening keynote in Orlando on October 19, as The National CIO Review reported, and Plummer used it again in Barcelona on November 9.
Gartner's releases on both keynotes leave the figure out, and it hasn't published a baseline or method for it.
The advice is aimed at consumption
Plummer put the productivity gain buyers see at 16%, against the 50% or more vendors claim. Plummer added: "The biggest drag on AI momentum is using AI when you don't need to." Gartner measured that gain itself, and it sells research both to CIOs and to the technology providers selling to them. Gartner research cited in the keynote found 86% of CIOs see AI risk growing faster than value, and AI use cases rose more than 400% in a year.
Banks have already set up AI FinOps teams to manage agent token spending.
Gartner's forecast doesn't show buyers pulling back. Gartner expects AI software spending to grow 42% in 2027, to $649 billion, and spending on AI agents and assistants to more than double, to $67 billion. Organizations are "boosting their software spending as vendors add AI capabilities to existing products," said Gartner analyst John-David Lovelock.
AI providers are building ahead of their revenue
Plummer tied the push for more tokens to the providers' own build-out and warned of a "massive asset write-down." Gartner's forecast puts hyperscalers and neocloud providers behind the rise in AI-optimized server spending, to $957.2 billion in 2027 from $527.6 billion. By 2030, Gartner expects AI providers to spend almost $7 trillion on infrastructure, against about $3.6 trillion in AI data center revenue, as The National CIO Review reported.
The biggest drag on AI momentum is using AI when you don't need to.
The November 12 numbers are Gartner's 3Q26 edition of its AI spending forecast. The next edition will show whether spending on AI agents and assistants still more than doubles in 2027.



