In 2026, the average business’s monthly AI spend has grown roughly 4x in a single year, yet only 6% of organizations can point to a measurable earnings impact from that spend (Ramp, 2026; McKinsey, The State of AI, 2026). The businesses closing that gap aren’t spending less on AI — they’re auditing what they already pay for before adding anything new.
Nobody decided to spend this much on AI. It happened one free trial, one bolted-on feature, one “let’s just try it” API key at a time. Six months later, there’s a ChatGPT Team plan, a Claude subscription, three AI features hiding inside tools already being paid for, and an API bill nobody remembers approving.
That pattern isn’t unique to any one business — it’s the industry default in 2026. The problem isn’t that AI costs money. It’s that almost nobody is checking whether what they’re spending still matches what they’re getting back. Here’s what the data says about how fast that gap is opening, and a simple way to audit it before it gets away from you.
How Fast Is AI Spend Actually Growing in 2026?
In 2026, the average business’s monthly AI spend grew roughly 4x between February 2025 and February 2026 (Ramp, 3 Steps to Manage AI Spend, 2026). Zoom out further and the number gets bigger: worldwide AI spending is forecast to grow 47% in 2026 alone, reaching an estimated $2.59 trillion globally (Gartner, 2026). Among businesses that spend on AI at all, the median company now dedicates close to 15% of its total software budget to AI tools alone.
Why the Average Company Now Runs 16 AI Tools
The median company runs 9 different AI models in production; the average is 16.5, pulled upward by heavy users running 26 or more, who carry a median monthly AI bill of $26,562 (Ramp, How Much Do AI Tokens Cost Businesses?, June 2026). Separately, AI-native app spending grew 108% year-over-year across the board — and ChatGPT is now the single most-expensed app of any kind, ahead of every traditional SaaS category (Zylo, 2026 SaaS Management Index).
Isn’t that the part nobody budgets for? A tool sprawl problem doesn’t announce itself as a single scary invoice — it arrives as sixteen reasonable-looking charges nobody has looked at side by side. That’s the exact same fragmentation we broke down in our post on the AI tech stack purge, just moved one layer up: from apps in general to the AI features layered on top of them.
The ROI Blind Spot: Why Spend Keeps Rising While Results Stay Flat
Only 6% of organizations qualify as AI “high performers,” meaning at least 5% of EBIT is directly attributable to AI. Just 37% attribute any earnings impact to AI at all — a figure that hasn’t moved since the year before (McKinsey, The State of AI, 2026). Meanwhile, 59% of CFOs name balancing the pressure to deploy AI quickly against managing its risk as their single biggest governance challenge (Deloitte, Q2 2026 CFO Signals Survey).
Spend and results aren’t supposed to move independently of each other. When they do for long enough, that gap has a name — and it’s not strategy.
The two numbers side by side tell the real story: AI now claims a real, growing slice of the budget, but fewer than 4 in 10 companies can point to a bottom-line reason why. This is exactly the gap we built a fix for in our 90-Day AI ROI Framework — define the metric before the spend, not after it.
The Waste Hiding in Your AI and SaaS Stack
Organizations leave an average of 36% of software licenses unused, wasting roughly $19.8 million a year at scale, and 78% of IT leaders reported unexpected charges tied to consumption-based AI pricing in the past year alone (Zylo, 2026 SaaS Management Index). Separately, 68% of companies say at least some AI initiatives ran over budget, and a third say it happens “mostly or always” (WitnessAI, via CFO Dive, July 2026).
None of this is a security story, though it rhymes with one — unmanaged AI tools are exactly how shadow AI takes root in a business, spend and risk growing together in the same blind spot.
Falling Token Prices, Rising Bills: The Part That Doesn’t Add Up
Here’s the counterintuitive part. The price to get equivalent LLM performance has fallen 9x to 900x per year, depending on the task, with a median around 40x a year for PhD-level science benchmarks (Epoch AI, LLM Inference Price Trends, 2025). Tokens should be getting dramatically cheaper — and per unit, they are.
Yet total dollar spend on AI still climbed roughly 497% over about a year (Ramp, 2026). The math only makes sense one way: businesses aren’t buying fewer, cheaper tokens — they’re buying access to more models, more seats, and more overlapping tools, and the savings from cheaper compute get absorbed by sprawl before they ever reach the bottom line. Cheaper AI didn’t make anyone’s bill smaller. It just made it easier to justify adding one more tool.
Building a Simple AI Spend Audit (Without Hiring a FinOps Team)
A useful audit doesn’t require new software or a finance hire. It requires three columns: every AI tool or AI feature currently being paid for, the specific business outcome each one is supposed to move, and who actually owns that outcome. Anything missing an owner or a measurable outcome goes on the cut-or-consolidate list first — before a single new tool gets added.
Run that audit quarterly, not once. Spend that’s growing 4x a year needs a review cadence that keeps pace with it, or the next audit just rediscovers the same sprawl with a bigger number attached. Vetting which of those tools are worth keeping long-term is its own discipline, covered in our post on the AI vendor shakeout.
Our AI Strategy Consulting service builds this exact audit and ROI model before any new tooling investment gets approved — so spend and results move together instead of drifting apart. For businesses where nobody currently owns that decision, our Hire Your AI CEO service puts fractional AI leadership in charge of exactly this kind of budget accountability.
A free AI audit with Aifyze is the fastest way to see what your current AI stack is actually costing — and what it would take to make that spend show up on the bottom line.
Frequently Asked Questions
How much are businesses actually spending on AI tools in 2026?
Monthly AI spend at the average business grew roughly 4x between February 2025 and February 2026, and among companies that spend on AI at all, the median now dedicates close to 15% of its total software budget to AI tools (Ramp, 3 Steps to Manage AI Spend, 2026).
Why is AI spend rising even though token prices are falling?
Because tool sprawl is outpacing the savings. Per-token API prices have fallen 9x to 900x per year depending on the task (Epoch AI, LLM Inference Price Trends, 2025), yet total AI dollar spend still grew 497% in roughly a year — driven by the average company now running 16.5 separate AI models instead of fewer, cheaper ones (Ramp, 2026).
How common are AI budget overruns?
Very common. 68% of companies report at least some AI initiatives ran over budget in the past year, and 33% say it happens “mostly or always,” according to a survey of 300 business executives (WitnessAI, via CFO Dive, July 2026).
What percentage of companies can actually prove AI is paying off?
Only 6% of organizations qualify as AI “high performers” with at least 5% of EBIT attributable to AI, and just 37% attribute any EBIT impact to AI at all — a figure unchanged from the year before (McKinsey, The State of AI, 2026).
Where should a small business start an AI spend audit?
Start by listing every AI tool and AI feature currently being paid for, including bundled add-ons inside existing software, then match each one to a specific business outcome it’s supposed to improve. Anything without a clear owner or measurable outcome is the first thing to cut or consolidate.
