The short answer
Gartner expects worldwide end-user spending on AI models and platforms to reach $64 billion in 2026, up 63.4% from $39 billion in 2025. Inside that number, generative-AI models grow 117%, AI platform spending rises 36.9%, and the standout is domain-specific and specialized models at 210%. The forecast, published on 20 July 2026, describes a market that is still expanding fast but where buyers have started asking harder questions about cost, latency, performance and provable returns.
For teams building and buying AI, the signal underneath the headline growth is a shift from experiment to accountability. The money keeps flowing, but it is moving toward smaller, task-specific models and toward vendors who can show measurable value and enforce policy across where and how AI is used. The winners of the next phase will not be whoever ships the biggest model — they will be whoever helps enterprises spend efficiently and prove it.
What did Gartner actually forecast?
On 20 July 2026, Gartner projected that worldwide end-user spending on the AI platforms and models market will total $64 billion in 2026, a 63.4% increase over the $39 billion spent in 2025. That is a market that has nearly doubled in two years and shows no sign of a sharp slowdown, even as the wider conversation has turned from capability to cost. The growth is broad, but it is not evenly distributed across the categories Gartner tracks.
Underneath the total, the mix is what matters. Spending on generative-AI models is forecast to grow 117% in 2026, while spending on AI platforms — the tooling to build, serve and operate models — rises a more modest 36.9%. The fastest-growing slice by far is domain-specific language models and specialized models, forecast at 210%. That last number is the tell: enterprises are increasingly reaching for smaller models tuned to a task or industry rather than paying frontier prices for general intelligence they do not need. Teams standing up their first GenAI integration now have a genuine choice between a general model and a focused one, and cost is pushing them toward focus.
Gartner also flagged a change in buyer behavior that reframes the whole forecast. "Enterprise AI budgets face heightened scrutiny with emphasis on usage efficiency, cost management and quantifiable returns," said Arunasree Cheparthi, Senior Principal Research Analyst at Gartner. The firm's read is that spending is shifting toward providers who can demonstrate clear value across cost, latency, performance and reliability, and who offer the evaluation tooling, cost monitoring and policy enforcement enterprises need to keep AI deployments under control. In Gartner's framing, the biggest winners will be vendors that help enterprises manage where and how AI is used — not simply those with the strongest raw model.
Why does a 63% jump matter now?
The headline number matters because it settles a live debate. For a year, the open question has been whether enterprise AI spending was a durable line item or a wave of experimental budgets that finance would claw back. A forecast of $64 billion, up 63%, says the demand is structural: AI has moved from pilots into systems that companies plan to fund for years. If you are still treating AI as a discretionary skunkworks cost, the market has already moved past you.
The composition matters even more than the total. A 210% jump in specialized and domain-specific models — against a comparatively restrained 36.9% for platforms — shows where the value is settling. Enterprises are learning that a smaller model tuned for their domain often beats a giant general model on the metric that actually matters: cost per useful outcome. That is exactly the calculus behind the rise of fine-tuned and domain-specific models, where a focused model on cheaper infrastructure delivers better accuracy and lower latency for a defined task than an oversized generalist.
And the scrutiny theme matters because it changes who wins the budget. When money was easy, the biggest model won attention. As budgets tighten, the deciding questions become: what does this cost per request, how fast is it, how reliable is it, and can we prove the return? That is a governance and FinOps problem as much as a modeling one, and it rewards teams that instrument their AI spend rather than those that simply increase it. It is worth being precise about scope, too: this is a market forecast, not guaranteed revenue, and Gartner's categories evolve — treat the direction as firm and the exact splits as directional.
What does this change in your AI budget?
The first thing it changes is the default answer to "which model?" A year ago, reaching for the largest frontier model was a defensible reflex. Now, with specialized models the fastest-growing segment and finance watching cost per outcome, "use the biggest model for everything" is an expensive habit rather than a safe one. The risk is not spending on AI; it is spending without a model-selection discipline that matches each workload to the cheapest option that clears the quality bar.
The second is that AI cost becomes a first-class operational metric, not a footnote. As spending scales into the tens of billions across the market, individual companies will feel the same pressure internally: unbounded token consumption, idle fine-tuning jobs and over-provisioned inference quietly compound. Without cost monitoring, evaluation harnesses and clear ownership, an AI budget can grow 63% and still deliver little — the classic pattern of rising spend with unproven return. For regulated FinTech and healthcare teams, this intersects with governance obligations: knowing which models touch which data, at what cost, is now both a finance and a compliance artifact.
The third risk is the quiet one: mistaking budget growth for progress. A larger AI line item is easy to approve and hard to justify after the fact if no one measured what it produced. The teams that struggle are not the ones that spend too little or too much in aggregate — they are the ones that cannot tie spend to a measurable outcome when asked. Folding evaluation, cost attribution and policy enforcement into your AI and digital transformation program early is far cheaper than reconstructing it under a budget review.
What it means for US & EU software teams
If you are scaling AI spend, build the measurement layer before you build the next feature. Put cost monitoring, a simple evaluation harness and per-workload ownership in place now, so every model call has an owner, a cost and a quality target. The point is not to slow down — it is to make sure the 63% more you are likely to spend next year buys 63% more value, and that you can prove it when finance asks. That discipline is what separates a durable AI program from a line item at risk in the next review.
If you build AI features, treat the 210% jump in specialized models as permission to right-size. For most production tasks — classification, extraction, routing, structured generation — a fine-tuned or domain-specific model on modest infrastructure will beat a frontier general model on cost and latency while matching it on quality. Reserve the biggest models for the genuinely hard, open-ended work. Matching each job to the smallest model that clears the bar is the single highest-leverage cost move available, and it is exactly where a focused AI and data engagement pays for itself.
There is a delivery read, too. Gartner's message that winners are the vendors and teams that help manage where and how AI is used applies inside your organization as much as to the market. A capability that owns model selection, cost governance and evaluation as routine work — in-house or through a dedicated engineering team — turns a forecast like this into a plan. Teams without that capability will keep approving bigger AI budgets and keep being surprised, at the next review, that they cannot say what the money bought.
What to do this quarter
Turn a market forecast into a short, concrete posture pass rather than a bigger blank check.
- Inventory your AI spend. List every model, platform and endpoint you pay for, what it costs per month, and who owns it — you cannot govern spend you have not mapped.
- Instrument cost per outcome. Add token and inference cost tracking tied to a business metric, not just a raw usage dashboard, so spend maps to value.
- Right-size your models. For each production workload, test whether a smaller or fine-tuned model clears the quality bar at lower cost and latency before defaulting to a frontier model.
- Stand up evaluation. Build a lightweight eval harness so model or vendor swaps are decided on measured quality, cost and latency — not vibes.
- Set policy and ownership. Define which models may touch which data, enforce least-privilege on keys, and give every AI workload a named owner accountable for its cost.
- Decide build-vs-buy deliberately. Choose where a platform, a specialized model or an in-house build fits each use case, and avoid paying frontier prices for commodity tasks.
A 63% market jump is confirmation that enterprise AI is here to stay — and a warning that the easy-money phase is ending. Teams that pair rising budgets with real measurement, right-sized models and clear ownership will turn this growth into compounding value. Teams that treat a bigger AI line item as its own reward will learn, at the next budget review, that spending more and getting more are not the same thing.
Frequently asked questions
How much will enterprises spend on AI platforms and models in 2026?
Gartner forecasts worldwide end-user spending on AI models and platforms will total $64 billion in 2026, up 63.4% from $39 billion in 2025. The forecast, published on 20 July 2026, covers the AI platforms and models market and reflects continued strong momentum even as buyers pay closer attention to cost and return on investment.
Which parts of the AI market are growing fastest?
Gartner expects spending on generative-AI models to grow 117% in 2026 and AI platform spending to rise 36.9%. The fastest-growing segment is domain-specific language models and specialized models, forecast to grow 210% as enterprises move from general-purpose models toward smaller models tuned to a particular task or industry.
What are domain-specific language models (DSLMs)?
Domain-specific language models are AI models tuned to a narrow task, industry or knowledge domain rather than trying to do everything a large general model does. Because they are smaller and focused, they can be cheaper to run, lower latency and more accurate on their target task, which is why Gartner expects the segment to grow 210% in 2026 as teams optimize for cost and measurable outcomes.
What does Gartner say enterprises should prioritize?
Gartner says enterprise AI budgets face heightened scrutiny, with emphasis on usage efficiency, cost management and quantifiable returns. Senior Principal Research Analyst Arunasree Cheparthi notes that spending is shifting toward providers that can demonstrate clear value across cost, latency, performance and reliability, and that offer evaluation tooling, cost monitoring and policy enforcement across deployments.
Sources
Gartner — Forecasts Worldwide AI Platforms and Models Market to Grow 63% in 2026
AIwire (HPCwire) — Gartner Forecasts Worldwide AI Platforms and Models Market to Grow 63% in 2026
Kaohoon International — Gartner Forecasts Worldwide AI Platforms and Models Market to Grow 63% in 2026