AI Will Eat a Fifth of Corporate IT Budgets by 2027

AI will consume 20% of corporate technology spending in 2027, according to a new Forrester study — and more than half of IT executives are still expected to miss their return-on-investment targets, The Register reports.

The numbers land hard because the expectations attached to them are even harder. In Forrester's survey, 45% of AI decision-makers say an investment must show a positive return within one year to count as successful. Yet only 22% have actually seen a positive impact on their company's operating profit over the past year.

Do the arithmetic: nearly half demand payback inside twelve months, while barely a fifth of organizations can point to profit-line evidence today. That gap — between a one-year clock and a reality where four in five firms have nothing to show on the bottom line — is where most AI budgets will live or die.

Why the returns aren't arriving

Forrester's diagnosis is organizational, not technical. Companies are pouring money into AI without changing how their organizations and work processes actually run. The technology arrives; the operating model stays the same.

The result is a patchwork. Financial gains do materialize, but they stay trapped inside individual use cases — a pilot here, a departmental tool there — and never scale into enterprise-wide returns. Buying models is easy. Rewiring workflows around them is the part that costs real effort, and it's the part most firms skip.

So the spending keeps climbing toward that one-in-five share of the IT budget, while the payback stays scattered and thin.

What does this change for CIOs planning 2027 budgets?

The practical answer sits in Forrester's cost forecast: expect a growing split in your model portfolio. To keep costs down, companies will increasingly run open AI models, which are cheaper to operate at scale. The expensive, advanced models won't disappear — they'll be reserved for critical tasks where the premium buys something the cheaper options can't.

That's a two-tier architecture, not a single-vendor bet. Budget owners who plan for it now — open models for volume workloads, frontier models only where the stakes justify the bill — are the ones positioned to stretch that 20% share into visible returns.

The harder homework is the organizational one. If gains only show up where processes change alongside the technology, then the ROI problem won't be solved by another model purchase. It gets solved by the unglamorous work of redesigning workflows, retraining teams and killing the use cases that never leave the pilot stage.

Forrester's message to IT leadership is effectively a deadline: the money is already committed, the one-year payback clock is already ticking for nearly half of you, and the profit evidence so far exists at just 22% of companies. Between now and 2027, the separation won't be between firms that spend on AI and firms that don't. It will be between firms that change how they work and firms that just buy.