Cost Cutting Retreats as CFOs Push AI Spending to Record Highs

The year finance chiefs stopped cutting

For most of the past five years, corporate finance chiefs talked mainly about trimming costs. In 2026 that changed. Grant Thornton's first-quarter CFO survey of 233 finance leaders found that 68% expected IT and digital transformation spending to rise over the next 12 months. That was the highest reading in the survey's 21-quarter history, edging past the 67% recorded in late 2025.5 Planned cuts fell at the same moment. Given a detailed list of cost categories, 28% of respondents said they did not intend to cut any of them. The previous high was 18%, in the fourth quarter of 2024.15

The drop showed up across the board. The share of CFOs planning to reduce professional consulting support fell from 52% to 32% in a single quarter. Planned cuts to human capital expenses slid from 43% to 29%, and vendor and supplier cuts from 42% to 29%.5 Grant Thornton said CFOs were holding back on cost reduction more boldly than at any point in five years, and that most were not cutting elsewhere to pay for record technology budgets.15

The headline story is real. The fuller picture is less tidy. Read together, the year's surveys show finance chiefs who did ease off broad cost cutting. By autumn, though, many were turning that discipline toward the AI bills they had run up.

Why the scissors went back in the drawer

Part of the reason is that AI is now seen as something companies can't skip. "Companies can't afford to treat AI as optional," Grant Thornton's Paul Melville said when the first-quarter results came out.1 The firm found that CFOs did not generally treat AI spending as a reason to cut jobs. Only 29% expected possible layoffs in the next six months, which tied a 15-quarter low in the survey.5 In some cases, the report said, companies needed to keep or even add staff to handle the growth AI was producing.5

The shift started before 2026 began. Grant Thornton's fourth-quarter 2025 survey showed 51% of CFOs expecting operating expenses to rise, up from 35% the quarter before and a 20-quarter high. The share planning cuts to long-term strategic initiatives fell to 28% from 36%.26 The firm linked that turn to expected tax savings from the One Big Beautiful Bill Act and to growing impatience with slow interest-rate cuts.21

That second factor is the one to watch. "They're deploying capital now because they need to move forward with these strategic initiatives, even if rates aren't as low as they'd like," Grant Thornton's Mike Desmond said.26 In other words, CFOs did not wait for cheaper money. They decided that waiting for lower rates had become its own risk.

Borrowing costs never left the room

That decision does not mean financing costs stopped mattering. In U.S. Bank's fall 2026 CFO survey, 35% of finance chiefs named high borrowing costs as a top concern and 34% named inflation. Only geopolitical tensions ranked higher, at 38%.7 Grant Thornton's own cost-management guidance described a year of volatility rather than a clean downturn. It cited cautious central banks after the 2025 rate cuts and tight liquidity that raises financing risk for investment and deals.23

Lenders are starting to price AI risk too. A UK poll of 251 lenders and investors by FRP found that 43% of Scottish respondents were not confident borrowers had fully counted AI adoption costs in their business plans. The same share said weak delivery of promised benefits could affect the price or margin of a loan.9 That finding comes from one regional sample, so it should not be stretched too far. Still, it points to where this could lead. If AI savings built into a business plan don't arrive, the shortfall could show up in a company's cost of capital, not just its operating budget.

Where the surveys disagree

The reporting is not unanimous. U.S. Bank's research still ranks cost cutting as finance leaders' top priority. In its spring survey of 1,000 senior finance leaders, 39% put it first, up from 33% in mid-2024, while revenue growth jumped from seventh place to second.10 By fall, cost cutting still led at 37%, with revenue growth close behind at 35%.12

This gap is mostly about how the questions were asked, not a real contradiction. Grant Thornton measured whether CFOs planned cuts in specific spending categories. U.S. Bank asked what they ranked as their top priority. A CFO can call efficiency the main goal and still decline to cut consultants or staff. Deloitte's research supports that reading. Asked which cost levers worked best, setting aside workforce reductions, 53% of North American CFOs chose automation or technology upgrades.6 For many finance chiefs, AI is now how they manage costs, not something that competes with cost management.

The bigger split is over jobs. Grant Thornton's low layoff expectations, and a Datarails finding that only 3% of CFOs are cutting finance staff because of AI, point toward redeploying people rather than letting them go.519 In the UK, Deloitte found that 52% of large-company CFOs expect to cut headcount within a year because of AI. Only a third thought AI would improve margins over the same period.18 One analysis noted that cost reduction rose to a strong priority for 68% of UK CFOs in that survey, from 53% the quarter before. That suggests the layoff expectations may partly reflect a cost-cutting mood, not just what AI can actually do.9 The most likely explanation is a transatlantic difference in timing and mood. It is not evidence that AI affects jobs differently in each country.

The meter starts running

The biggest development since spring is that AI turned out to be expensive in its own right. Fifty-one percent of U.S. finance leaders told U.S. Bank that AI spending had exceeded budget over the past year. More than two-thirds also said it had created commercial opportunities.7 In Pigment's third-quarter CFO index, 83% said usage-based AI costs came in higher than expected. Deloitte found that 60% of finance leaders expect AI costs and complexity to rise sharply through 2027.4

Forecasting is getting worse, not better. A Mavvrik and Benchmarkit study of 396 enterprises found that only 11% predicted their AI spending within 10% of the actual figure, down from 15% a year earlier. Sixty-two percent said a surprise AI cost had changed a business decision, and one in four had delayed or canceled an AI project.16 Eighty-one percent reported a moderate-to-high hit to gross margin.16 A global KPMG survey of 2,145 executives found 29% caught off guard by rising AI costs. The coverage pointed to usage-based pricing and unmonitored employee use as major causes.15

Bloomberg has reported that finance executives are cutting back on AI spending and moving budget control to a central function.4 Datarails found that 32% of CFOs went over their AI budgets by at least 10%. It also found that 13% of AI budgets have no clear owner, which helps explain how usage costs grow without anyone noticing.20

Cutting, but not retreating

None of this adds up to a pullback from AI. In CloudZero's survey, 87% of finance leaders plan to increase AI budgets next year, and 45% plan increases above 20%. In the same survey, 61% said their current AI spending was more than they could justify by results.11 Datarails found that 53% plan to expand AI licenses, while only 7% plan to cut or consolidate tools.17 Grant Thornton's third-quarter survey found 84% of CFOs saying AI returns were meeting or beating expectations. A record 80% expected net profit to rise over the next year.

That confidence deserves some caution. Korn Ferry observed that CFO satisfaction rests mostly on AI holding down costs, not on growing revenue. Only 33% said AI was helping sales.14 If AI is valued mainly for controlling costs, and its own costs keep beating forecasts, that reasoning starts to break down.

The pattern for 2026 looks like this. Finance chiefs gave up broad, across-the-board cuts in favor of what Grant Thornton calls cost optimization: cutting where spending adds little value and moving the savings into areas that do.3 They did it with borrowing costs still elevated and the economy uncertain, betting that falling behind on AI was riskier than spending on it. Their next test is whether they can apply the same discipline to their AI bills, before overruns hit margins and, potentially, their loan terms.