AI Spending in 2026: Three Numbers That Don’t Add Up

The headline AI spending forecasts for 2026 measure fundamentally different things, and mixing them is the single most common analytical error in technology budgeting. Gartner forecasts $2.59 trillion in total AI spending. IDC tracks $487 billion in AI infrastructure. Stanford reports $581 billion in AI investment. Each number is correct; none of them are interchangeable.

Pick up any board deck or vendor pitch from the past month and you will likely find at least two of these figures sitting next to each other as if they describe the same market. They do not. Gartner is counting procurement spend across the full AI stack. IDC’s tracker is counting hardware shipments. Stanford is counting external funding rounds into AI companies. The economic flows are orthogonal, and the periods do not even align — Stanford’s $581 billion is a 2025 actual, while Gartner and IDC are 2026 forward forecasts.

This is a reference, not a recap. The value is in the side-by-side explanation of what each forecast actually counts, so you cite the right series for the right argument — and never present a combined number that no analyst would endorse. Every figure below is drawn from the primary releases; where a number is reported only through secondary coverage, we say so.

Start with the three headline figures and the one thing that makes them incompatible: each measures a different point in the AI economy. A buyer reading only the headlines would assume the numbers conflict — $581 billion, $487 billion, and $2.59 trillion cannot all be the size of the same market. They are not. They are three orthogonal measurements that happen to share a subject.

Gartner measures total AI procurement spend across the entire stack: hardware, services, software, cybersecurity, platforms, models, application development, and data. This is what organizations are projected to actually spend on AI in 2026. IDC’s infrastructure tracker measures one slice of that stack — hardware shipments of AI-optimized servers, storage, and networking. It deliberately excludes software and services. Stanford’s AI Index measures something else entirely: external funding flowing into AI companies through venture rounds, private equity, and acquisitions. That is investment, not operational spend, and it is reported as a 2025 actual rather than a 2026 projection.

The table below is the asset most coverage skips. Every press recap reports each figure as if it measures the same thing; none of them line the analysts up on a single axis — “what does this number actually count?” Read across each row and the apparent contradiction dissolves. Note that IDC appears twice, because IDC publishes two distinct AI series that are themselves not additive: an infrastructure-only tracker and a broader full-stack spending guide.

The single most important column is “what it measures.” Once you sort by that, the rest falls into place: Gartner is the only row that captures the whole stack, IDC’s tracker is a hardware subset that maps onto Gartner’s infrastructure sub-segment, and Stanford is a different axis altogether. For the capital-formation side of the story — how venture money is flowing into the labs that build the models — see our breakdown of venture capital flowing into frontier AI labs, which is the Stanford-style investment flow at the level of a single company.

Gartner forecasts worldwide AI spending will total $2.59 trillion in 2026, up 47% year-over-year. This is the broadest of the three definitions — it counts AI infrastructure, AI services, AI software, AI cybersecurity, AI platforms, AI models, AI application development, and AI data. AI infrastructure alone accounts for more than 45% of that total, driven by AI-optimized servers, infrastructure-as-a-service, network fabric, and AI processing semiconductors.

There is a revision story worth understanding before you cite the number. In January 2026, Gartner’s forecast was roughly $2.52 trillion at 44% growth. By May it had risen to $2.59 trillion at 47% growth — about $70 billion added, attributed to accelerating demand for AI infrastructure and agentic tooling. Gartner itself cautions that comparisons between its current and earlier estimates are not meaningful, because the scope widened: AI cybersecurity, agentic AI as a distinct segment, and AI data technology were folded in. Treat different Gartner vintages as different definitions, not a clean time series.

The segment table below assembles Gartner’s category breakdown. One caveat the data demands: the full segment-level figures come from Gartner’s January 2026 detailed release, while the May 2026 update republished only the total plus selected segments (notably the upward revisions to agents and models). We label the January estimates as such; the infrastructure sub-segment in particular may have been revised upward in May without a republished full table.

Two pieces of context make the scale legible. Gartner’s $2.59 trillion AI figure sits inside a total worldwide IT spending forecast of $6.31 trillion for 2026, meaning AI accounts for roughly 41% of all IT spending — up from approximately 32% in 2025. And Gartner has described this as one of the fastest periods of technology spending growth in recorded history. The transition Gartner emphasizes is not just scale but composition: spending is shifting from hyperscaler-led to enterprise-led, which the firm calls the inflection year for enterprise AI adoption.

If you only track one segment, track agent software. Gartner’s figures put purpose-built AI agent software at $86.4 billion in 2025, rising to $206.5 billion in 2026 — roughly +139% in a single year — and then to $376.3 billion in 2027, an +82% step. That 2026 growth rate is nearly triple the +47% growth of the overall AI total, meaning agents are the fastest-growing category even within an already-booming market. A note on the two growth figures: the +139% is the 2025-to-2026 step and the +82% is the 2026-to-2027 step; secondary coverage sometimes conflates them, so keep them separate.

Distinguish agent software from the broader “AI software” category. Gartner’s AI software segment ($452 billion in the January estimate) covers enterprise software with AI features. Agent software is a narrower, faster-growing slice covering purpose-built agentic systems and emerging autonomous workflows — it is not the same line item. The $206.5 billion agent figure is reported through secondary coverage of Gartner’s methodology, so treat it as vendor-stated and verify against Gartner’s primary release before quoting it in a contract or board document.

The models line tells a parallel story. Gartner’s AI models spending was raised to approximately $32.6 billion for 2026 — up about 110% from $15.5 billion in 2025 — with the short-term outlook lifted by roughly $6 billion specifically to reflect accelerating enterprise demand. Both the agents and models revisions point the same direction: the fastest money is moving toward the autonomous and generative end of the stack, not the commoditizing infrastructure base. For the cost side of running these workloads, our FinOps playbook on the cost-side of the AI infrastructure build-out is the operational companion to this spend forecast.

IDC’s most-cited AI number comes from its Worldwide Quarterly AI Infrastructure Tracker, which measures hardware only: AI-optimized servers, storage, and networking. It explicitly excludes software and services. Full-year 2025 AI infrastructure spending totaled $318 billion worldwide, more than double the $153 billion in 2024. IDC forecasts that figure will reach $487 billion in 2026 — up 53% — and exceed $1 trillion by 2029 at a roughly 31% five-year compound growth rate.

The composition is striking and useful for buyers: in Q4 2025 alone, AI infrastructure spending reached $89.9 billion, of which servers were $87.7 billion — about 97.6% of the total — and storage just $2.2 billion. The market is, to a first approximation, a server market. IDC projects that accelerated servers (primarily GPU-based) will account for more than 95% of AI server spending by 2029. This is the hardware reality that underpins the infrastructure sub-segment of Gartner’s broader total.

IDC also flags three risks to this trajectory that buyers should hold alongside the growth numbers: power generation and grid capacity constraints, memory and storage component scarcity, and export controls and data sovereignty regulations. None of these are priced into a headline CAGR, and all three are the kind of supply-side ceiling that can turn a smooth forecast curve into a step function.

The Stanford AI Index 2026, released April 13, 2026 as the seventh annual edition, measures something neither Gartner nor IDC does: actual corporate AI investment. Global corporate AI investment in 2025 reached $581 billion — a 129.9% increase from $253 billion in 2024 — based on data from analytics firm Quid. That figure covers private equity rounds, M&A, and minority stakes in AI companies. Critically, it is a 2025 actual, not a 2026 forecast, which is the single most common way this number gets misused.

The Index defines corporate AI investment as external funding for privately held AI companies raising above $1.5 million — venture rounds, private equity, and M&A — explicitly not operational AI spend on hardware, software, or salaries. That definitional choice is what makes the $581 billion non-comparable to Gartner or IDC: those measure procurement and capex, while Stanford measures capital flowing into AI companies. They are different sides of the same economy. For the full set of 2025 actuals across every chapter of the Index — capability, compute, policy, and economy — see our companion digest of the 2025 AI actuals across all Stanford Index chapters; this post is the forward-forecast counterpart to that backward- looking read.

The regional split is where the Stanford data earns its keep. US private AI investment in 2025 was $285.9 billion — 23.1 times greater than China’s $12.4 billion and 48.5 times greater than the UK’s $5.9 billion. The US attracted 1,953 newly funded AI companies, more than ten times the next closest country. And the generative-AI concentration is even sharper: more than half of all US private AI investment was generative-AI-related at $163.6 billion, while China and Europe’s combined generative-AI investment was just $4.7 billion. One watch-out for anyone citing this data: the $344.7 billion figure is global private investment excluding M&A, not a US number — US-only private investment is $285.9 billion, and secondary coverage frequently mislabels the two.

Most misuse of these numbers follows a handful of recurring patterns. Each one produces a figure or a claim that no analyst would endorse, and each is avoidable once you know the distinction it violates. Here are the five we see most often in decks and pitches.

The throughline across all five mistakes is the same: a number is only as useful as the definition attached to it. Strip the definition and any of these figures can be made to say almost anything. The discipline that keeps you honest is boring but reliable — always state what the number measures, what period it covers, and which analyst and vintage it came from, in that order.

Here is the original read the press releases bury. A $2.59 trillion spending forecast reads like an unqualified boom, but Gartner’s own data tells a more cautious story underneath it. Only about 17% of organizations have actually deployed AI agents to date, even as 60% expect to within two years. And Gartner projects that more than 40% of agentic AI projects could be cancelled by the end of 2027 — citing escalating costs, unclear business value, and inadequate risk controls. The spending surge and the production reality are not the same curve.

Projecting forward, the most likely shape of the next two years is a widening gap between committed capital and realized value. The money is flowing fastest into the autonomous end of the stack — agent software at +139% — precisely where the cancellation risk is highest. That is not a contradiction the market will resolve cleanly. It means the firms that win will be the ones that treat the spend forecast as a signal of where capacity is being built, not as permission to skip the harder work of measuring whether a given deployment actually pays back. Gartner’s own language captures the tension: enterprises are favoring tactical pilots over wholesale transformation, even as valuations price in the transformation.

The practical implication for a buyer is to separate the two questions the headlines collapse together. “Is AI spending growing?” is settled — every series agrees it is. “Is our AI spending returning value?” is a different question entirely, and the cancellation data suggests most organizations have not answered it. The discipline that closes that gap is measurement, not more budget. Our work on how CFOs are measuring returns on this spending and the broader frameworks for evaluating AI ROI are where the spend forecast turns into a decision.

The reason to keep these series straight is not academic — it is that each one is the correct evidence for a different claim, and using the wrong one undermines an otherwise sound argument. Match the figure to the question you are actually answering.

For agencies and engineering teams building toward this market, the most useful move is to treat the forecasts as a map of where capacity and capital are concentrating — and then run your own per-workload economics on top of it, because the aggregate numbers say nothing about whether a specific deployment pays back. If you are translating these spend signals into an actual AI roadmap, our AI and digital transformation engagements start with exactly that kind of grounded, evidence-first scoping — and our work on IDC’s enterprise AI adoption projections connects the agent-software surge to what teams should be preparing for now.