Quick Answer
87% of finance leaders are raising their AI budget for FY2027, 45% of them by more than 20%, even though 61% already spend more on AI than they can justify. Most can't see where it goes: 54% can't break AI spend below a rough category, so budgets default to the total and keep climbing. The teams that allocate spend, tying each AI dollar to the team, product, or outcome behind it, are the ones who can defend the number to a board.
NOTE: This is a deeper look at one finding from Why Finance Caps What It Can’t See, CloudZero’s survey of 252 senior finance leaders.
Finance has stopped debating whether to fund AI, and plenty of teams are now in YOLO mode with the spend.
In CloudZero’s survey of 252 senior finance leaders, 87% are increasing their AI budget for FY2027, and 45% of those are raising it by more than 20%.
The same survey found that 61% already spend more on AI than they can justify with measurable outcomes.
Those are mostly the same people. Among the finance leaders who say they cannot justify what they currently spend, 91% are increasing it anyway. What follows is why that happens, which programs overrun hardest, and what over-budget teams say they would do differently next year.
What is an AI budget?
An AI budget is the planned allocation of money for an organization’s artificial intelligence costs across a fiscal period, covering model API usage, inference and training compute, AI infrastructure, and licensed AI tools. It behaves differently from a traditional software budget because most of the cost is consumption-based, so the total rises and falls with usage.
An AI budget is defensible only when the spend inside it can be broken down below a single line on a bill and traced to what it produced. That last condition is where most finance teams are stuck.
2026 State of AI Spend Report
Why Finance Caps What It Can’t See
Learn how finance leaders are managing AI spend, and how they’re approaching FY2027
Why AI budgets keep rising when finance can’t defend them
AI budgets keep rising because the pressure to spend arrives from above finance, and it arrives without a matching demand for proof.
Boards and CEOs set the direction. Finance executes it and absorbs the consequences. Among companies whose boards push them to invest aggressively, the spending mandate precedes the business case, and the approval clears anyway. Companies whose boards tell them to invest aggressively approved AI spend without reliable projections at 61%, against 39% of companies at every other board posture.
That is a sequencing problem. The mandate lands first, and the projection is expected to catch up.
CEO-mandated AI programs overrun the most
CEO-mandated AI programs blow their budget by more than 20% at 19%, against 9% of every other program.
That is the “act now, reconcile later” directive showing up in the numbers. The instruction is to move, and the budget conversation gets scheduled for whenever finance catches up.
Ownership of the mandate does not change who carries the result. Consequences from poor AI spend visibility hit 86% of companies regardless of who drives the program, and two-thirds absorbed more than one consequence. The two most common are quieter than the dramatic ones: 44% approved AI spend without reliable projections, and 36% could not produce a credible ROI number when the board asked for it.
What over-budget teams say they would change
Asked what they would change in hindsight, 57% of over-budget teams said they would have set a larger AI budget from the start.
Stronger ROI cases came in at 35%. More contingency came in at 41%. The most common lesson drawn from blowing through an AI budget is that the number was too small.
That instinct is rational when the alternative is invisible. A finance team that cannot see which part of its AI spend produced something has no basis for cutting one part rather than another. Raising the ceiling is the move that does not require knowing where the money went.
The AI budget problem is a visibility problem
Most finance teams cannot break their AI spend below a rough category, which is why the budget conversation keeps defaulting to the total.
CloudZero’s survey measured AI cost allocation on a five-rung ladder, from seeing only a total on a bill to allocating spend to a customer, product, or cost center and using it inside the quarter. 54% of finance teams sit on the bottom two rungs. Only 12% can break AI spend down finely enough to act on it within the quarter.
Allocation is also the only one of the three capabilities measured where FY2027 demands more than finance currently has. Forecasting is already slightly ahead of where teams say they need to be. Outcome measurement is further ahead still. Allocation is the gap.
The cost of that gap shows up at the board. Companies that claim board-defensible ROI but cannot allocate the spend behind it report the same board-proof gap as companies claiming no return at all, 56% against 54%. Measuring return tells a board the program produced something. Allocation tells them which part of it did.
The 23% of companies that can do both plan the same FY2027 increases as everyone else, 41% against 38%, and absorb roughly half the damage, with 27% hitting a growth-slowing consequence against 47%. Only 3% of them failed to secure an AI investment, against 19% of everyone else. They can see which spend is working, which makes it possible to put more behind it.
External research points the same way. KPMG’s Global AI Pulse Q2 2026, covering 2,145 senior leaders across 20 countries and territories, found organizations with full visibility into AI operating costs five times likelier to report established AI ROI, 15% against 3%. In that same survey, only 35% said their AI operating costs were fully visible and actively monitored, and 49% had already delayed or scaled back an AI deployment when costs began to outweigh expected value.
What to do before FY2027 budgets lock
Three things separate the teams that get through next year without a board problem.
Diagnose allocation on its own. Across the two cohorts the survey compares most closely, teams adding approval gates and teams tying funding to ROI, forecasting and outcome measurement were level at 48% and 52% strong in both. Allocation was the only capability that differed. Score the organization separately on all three and name the weakest.
Name the lever before pulling it. Approval gates are the single most common FY2027 change at 27%, ahead of a formal budget line at 19% and tying funding to demonstrated ROI at 18%. Gates work without visibility, which is exactly why teams that cannot see their spend reach for them. They also become a permanent tax when nothing is funded to replace them.
Put a funded line behind the capability. 74% say linking spend to outcomes would change specific FY2027 decisions. The report calls that a resourcing gap rather than a conviction gap, and prescribes a named owner and a funded line for AI cost management rather than a working group.
Building an AI budget you can defend
FY2027 is the year the excuse for not measuring AI spending runs out.
The teams that come through the year without a board problem are the ones who can say which part of their AI spend they would defend. That is a different question from how much went to AI, and it is the one the next budget cycle asks.
CloudZero’s full research covers all 13 findings from 252 senior finance leaders, with the complete methodology and every response distribution published alongside it.
Read the full report: Why Finance Caps What It Can’t See
Frequently asked questions about AI budgets
