By 2026, global enterprise AI spending is set to reach $2.59 trillion, a 47% jump over 2025, as recently predicted by Gartner. That number would make AI the fastest-growing technology expenditure category in enterprise history. The infrastructure funding that spend represents is real: Nvidia’s quarterly revenue has exceeded $81 billion, Micron’s high-bandwidth memory production for 2026 is already sold out, and data center construction pipelines now run continuously from Virginia to Singapore.
Despite that spending, few enterprises can actually justify it. And on that question, the data from the world’s largest research firms tells an uncomfortable story: enterprises are writing checks faster than they are producing evidence those checks were worth writing.
The Gap, By The Numbers
Based on a survey of 1,719 executives across 97 countries, McKinsey’s 2026 State of AI report found that 37% of respondents attribute at least some EBIT impact to AI use, a figure unchanged from the 2025 survey. More strikingly, only 6% of organizations qualify as “AI high performers,” meaning they attribute 5% or more of EBIT to AI and describe the impact as significant. While the share of large companies scaling AI agents jumped from 27% to 40% year on year, that 6% figure hasn’t moved in the same period.
In other words: adoption is accelerating fast, but financial proof of that adoption paying off is not.
McKinsey frames this tension directly, noting that AI conviction is growing across organizations faster than the financial returns companies can actually attribute to it. That’s a polite way of describing what finance departments are now openly worried about. According to Gartner research, 84% of finance leaders say they struggle to measure AI ROI at all – not that the ROI is bad, but that measuring it in the first place has become genuinely difficult as AI costs spread across hundreds of individual subscriptions, token invoices, and cloud line items that were never designed to be tracked as a single spending category.
Deloitte’s State of AI in the Enterprise survey, covering 3,235 leaders across 24 countries, found that 66% of respondents report productivity gains, but only 20% see AI-driven revenue growth. Just 25% of AI initiatives delivered the ROI executives expected in 2025. And in a separate 2026 Forbes Research survey, while over 70% of organizations report “positive” AI ROI in some form, fewer than 1% report significant ROI of 20% or more. Most companies land in the 1% to 5% range, and even that is frequently measured as a productivity gain rather than a financial one, an important distinction when a board is asking whether the AI line item is earning its place in next year’s budget.
PwC’s 2026 survey of 4,454 CEOs across 95 countries puts a number on how widely this uncertainty has spread at the very top of organizations: 56% of CEOs report no significant financial benefit from AI to date.
Where The Money Is Actually Going
The uneven spending tells its own story. IDC and Microsoft research puts the average enterprise return at $3.70 for every dollar spent on generative AI, with mature programs reaching $4.60 per dollar according to Accenture. But that average conceals a wide spread: pilot-phase programs return only $1.20 per dollar, and the median time to positive ROI sits at 14 months. Only 45% of organizations can quantify their AI ROI at all, which means over half of the enterprises currently increasing their AI budgets are doing so without a reliable way to prove the last round of spending worked.
The spend itself skews heavily by industry and by company size. Enterprise AI spending is projected to hit $407 billion in 2026, up 34.8% year over year, according to Gartner, with financial services leading at $68 billion and 79% adoption, and technology companies posting the highest adoption rate at 88%. Education, by contrast, trails at 34% adoption, a reminder that “enterprise AI spending” is not one uniform curve but several very different ones layered on top of each other. Per-employee spending tells a similar story of concentration: the median company spends under $200 per employee on AI, while the top 10% of spenders exceed $2,800, and financial services firms average $3,200 per employee, more than double the cross-industry figure.
Why The Proof Keeps Lagging The Spend
Part of the answer is structural. The AI procurement decisions made across most large enterprises through late 2025 were driven by technology leadership – CTOs, CIOs, and AI strategy teams – with comparatively limited scrutiny from finance. The competitive argument was consistent everywhere: move slower than rivals on AI adoption, and the resulting productivity and cost-structure gap becomes permanent. That logic created a genuinely permissive environment for AI spending, one that largely bypassed the cost-benefit review cycle that normally governs IT expenditure.
That environment has started shifting. An Axios investigation published in May 2026 identified what a poorly governed AI deployment can look like in practice: one large, unnamed enterprise client spent $500 million in a single month on AI services, a figure that illustrates just how far spending discipline can slip once technical teams operate without financial guardrails. Cases like that are pushing finance departments back into the AI conversation, not to slow adoption outright, but to insist that budgeting rigor catch up with technical ambition.
Alongside that sits a deeper measurement problem hiding inside the ROI numbers themselves. Half of surveyed companies use data quality improvements as their AI success metric, and 48% use employee productivity, according to Forbes Research. Far fewer tie AI investment directly to profit-and-loss impact or margin. That distinction matters more than it sounds: an organization can be measurably more efficient on paper while still losing market share or eroding margin elsewhere, and a dashboard full of productivity metrics will not surface that problem until the P&L does.
What Separates The 6% From Everyone Else
The AI high performers, as classified by McKinsey, are not distinguished by how much they spend but by governance discipline that most enterprises still lack. According to Deloitte, only one in five companies report a mature governance model for autonomous AI agents, even as agent adoption itself is growing 82% year over year. That gap between fast-moving deployment and slow-moving oversight is where the accountability risk concentrates, and increasingly where the security risk does too: IBM’s 2025 study found 13% of organizations had experienced an AI-model breach, with average breach costs in financial services reaching $5.56 million.
For the CFOs and boards watching this unfold, the practical takeaway isn’t that AI spending should stop – McKinsey’s survey found 80% of individual workers report real productivity gains from AI. The takeaway is narrower and more disciplined: instead of funding AI initiatives on competitive anxiety alone, without the measurement infrastructure to prove they worked, enterprises need to build that infrastructure first. The $2.59 trillion being spent globally in 2026 will eventually need to answer for itself in the same currency it was spent in. Right now, for the overwhelming majority of enterprises, it still hasn’t.
