As companies start writing next year’s IT budgets, two surveys published a week apart in September point to a tougher conversation ahead. BCG’s IT Spending Pulse found IT budgets expected to grow 5.8% in 2026, with AI and machine learning the top spending priority.
Days earlier, Futurum’s second-half 2026 CIO survey found that 46.9% of organizations were running their AI spending over budget.
If you have to explain the AI line to finance, the challenge isn’t making the case for more investment. It’s figuring out what to fund, what to trim, and which existing priorities get squeezed to pay for it—because AI isn’t drawing from a new pot of money, and the receipts are showing up across survey after survey.
READ MORE:
Post-mortem of an AI pilot: An exit checklist before scaling to production
The one line that’s growing
Start with the BCG report. IT buyers expect 2026 budgets to grow 5.8%, but the growth is lopsided: 66% expect to increase AI and machine learning spending, more than any other category, while net buyer intent for IT services swung down 19 points and CRM and ERP both turned negative.
The budget grew — but most of the growth intent went to AI. The spending isn’t pointless, either: BCG found measured ROI on GenAI and AI agents rose to 13.8%, up from 11.2% a few months earlier, with high-maturity firms reporting about 19% against 8% to 9% for the least mature. The problem isn’t that AI spending is wasted. It’s that most organizations can’t yet tell which spending is the useful kind.
BCG surveyed mid-size and large companies, so if you work at a smaller one, assume the pressure is at least as sharp.
The same pattern shows up elsewhere. Enterprise Technology Research found AI’s share of the average IT budget climbed from 12.1% to 14.2% in a year. We looked at one version of that squeeze — consolidation as a funding strategy — in our recent piece on single-vendor SASE.
One caveat, and it is a significant one: not everyone reads this the same way. Gartner’s analyst John-David Lovelock disputes the reallocation story outright. “CIOs got some net new money for AI back in 2024, and a little in 2025,” he told CIO.com, “so there wasn’t a diversion, and now more of their spending is about rebranding than diversion.” His point: existing spend is being relabeled as AI rather than moved to it. Keep that in mind when you look at Gartner’s own forecast — $2.7 trillion in worldwide AI spending in 2026, up 49.5%, a market-level figure dominated by chips, servers, and hyperscaler buildouts rather than any one company’s budget.
So what actually gets cut?
The clearest answer comes from Goldman Sachs’ June CIO survey. Only 35% of AI spend is net-new budget; the rest is reallocation. And the order is revealing. The biggest sources are application software, with 18% of AI funding pulled straight from it, followed by labor at 11%, cloud infrastructure at 10%, and data and analytics at 9%. In plain terms: the software you already buy is first on the list.
READ MORE:
Attackers are calling your help desk to reset their way in
Futurum’s survey, covering 1,636 enterprise decision makers, goes further. When the overrun lands, most organizations don’t slow down: 47.6% ask for more money, 43.3% absorb it and settle later, and only 17.2% pause or reduce the AI work itself. (Respondents could pick more than one answer, so those add up past 100%.) When they reallocate within IT, external contractors and consultants are cut first — 60.9% of the organizations doing it. And where AI-driven headcount reductions happen, they land on IT support and helpdesk (60.2%) and software and application development (58.4%).
For nearly a quarter of over-plan organizations (23.1%), the money comes from a non-IT business unit’s budget — evidence, Futurum says, that AI funding has begun to reach beyond IT’s own.
Not everything is on the block, though. A small Jefferies survey of 30 IT leaders found 71% expect to reallocate IT services budgets toward AI, but 0% expect cuts to cybersecurity spending. Security, so far, has a defender.
None of this is hypothetical. CIO.com documented the tactics in March: one team cut its data-validation software budget 40% and merged three project management tools, freeing about $47,000 a year for AI quality-control software. Another trimmed contractor spending by 30%. A third delayed server and network upgrades by 12 to 18 months to free about 30% of its infrastructure budget.
Why this lands harder on a small team
Enterprise surveys describe the mechanism, but small IT teams feel the pinch differently — and often worse.
When Goldman says application software is first to be cut, that is your tooling budget: the monitoring license, the help desk platform, the backup tool you have renewed without thinking for years. When Futurum says external contractors go first, for a lean team, that is your MSP or the consultant you bring in for projects you can’t staff. When it says AI-driven cuts land on IT support and helpdesk, in a small shop that is not a department being trimmed. It is a person — often the same person who does everything else.
There is a subtler trap, too. In a large company, that non-IT budget is a different department absorbing the cost. In a small one, there may be no other budget to raid — so the money comes from somewhere closer to home, or from the maintenance you keep deferring.
Which is the real risk. The money being pulled to fund AI often comes from the same foundations AI depends on: identity, data quality, reliability, and the upgrades nobody notices until they are overdue. Defer maintenance long enough and you are not funding AI. You are borrowing against next year’s stability.
How to keep your line off the block
Aarti Anand, who advises CEOs of $10 million to $50 million companies, put the problem in a LinkedIn post: the AI line often survives budget cuts “on ambiguity, not value” — not because it is the strongest line, but because nobody is sure what breaks if it is cut. Her test for any line: what does this return, who owns it, and what specifically breaks if it is gone?
Dion Hinchcliffe, chief research officer at The Futurum Group, says in his own commentary that this is not truly a budgeting issue. The real problem is structural: procurement, budgeting cycles, and governance were designed for slow-moving technology, but AI moves fast. He calls the solution rethinking an organization’s “AI metabolism” for CIOs.
The research keeps circling around the same gap. Flexera’s 2026 IT Priorities Report found 80% of organizations increased AI spending, but only 19% say proving its effectiveness is a top priority. Dataiku’s survey of 685 CIOs found 47% have already decommissioned 20 or more AI agents this year — a hint of how much is being tried and dropped.
So before the budget season closes, do what the data says most organizations are not doing. Put a number on what your AI spend returns. Find its owner. And check what is on the chopping block to pay for it — because if you cannot say what breaks when a line is cut, you have just volunteered it.
Categories: Articles, IT Spending
Your Essential Daily Tech Brief
Sign up for the Spiceworks Newsletter and stay ahead with a curated mix of expert advice, technical tips, and the trending discussions your peers are having right now.
Leave a Reply