Gartner Says AI Will Be 41% of All IT Spending in 2026. Only 1 in 4 Projects Hits Its ROI Target

Gartner forecasts worldwide AI spending will reach $2.59 trillion in 2026, up 47% year over year, representing roughly 41% of all IT spending, up from about 32% in 2025. Enterprise AI spending specifically is projected at $407 billion, up nearly 35% from the year before, with more than 45% of that going to infrastructure rather than software or services.

The number worth putting next to that: only 25% of AI initiatives met their expected returns in 2025, according to IDC and Microsoft research covering the same period, despite 74% of executives reporting they saw first-year ROI at all. The median time to reach positive ROI on an AI initiative is 14 months, with an average return of $3.70 per dollar spent among the initiatives that did succeed.

Two numbers, one honest picture

Spending is scaling faster than success rate. That's not necessarily irrational, a lot of AI investment right now is infrastructure and capability-building that won't show a clean ROI figure for a while, and the 25% figure covers a wide range of project types and maturity levels lumped together. But it does mean "everyone else is spending more on AI" is a weak justification on its own for a specific initiative, since the base rate for that spending actually paying off within a year is closer to one in four than three in four.

What this is useful for

If you're the person on a team being asked to justify why you haven't adopted more AI tooling, or being pushed to adopt something faster than you'd like, these two numbers together are a legitimate counterweight: the spending trend is real, but it is not the same thing as a success guarantee, and the gap between the two is wide enough that "we should be spending more because everyone else is" doesn't hold up as a standalone argument. A specific measurement plan for what success looks like, checked against a 14-month realistic timeline, is a more defensible basis for a new AI initiative than matching the industry spending curve alone.