Worldwide spending on artificial intelligence is forecast to reach $2.7 trillion in 2026, a 49.5% increase year-over-year, according to Gartner's latest quarterly forecast published on 21 September 2026. The number is so large it risks becoming abstract, so here is what it means in practice: AI is no longer an experimental budget. It is now one of the largest categories of enterprise and government spending on the planet.
The infrastructure build-out is the biggest project in history
John-David Lovelock, distinguished VP analyst at Gartner, described the build-out of AI data centre capacity as "the largest infrastructure project humanity has ever undertaken." Hyperscalers and service providers are buying AI-optimised servers, networking fabric and semiconductors at a pace that dwarfs previous technology cycles. That spending is inelastic, which means it is not slowing down even as memory prices rise.
For enterprise buyers, this matters because capacity creates capability. The reason agents can now run multi-step workflows, process real-time data streams and operate across multiple enterprise systems is that the underlying infrastructure has reached a density and speed that simply did not exist two years ago. If you are evaluating an agent platform today, you are buying into a stack that is improving faster than any technology infrastructure before it.
Where agents sit in the $2.7 trillion picture
Gartner has separated cross-functional agents and assistants from general AI software in its latest forecast, explicitly recognising them as a distinct spending category. That separation matters. It means procurement teams, CFOs and vendor boards are now treating agentic AI as its own budget line rather than a feature bundled inside existing software licences.
Expect this to change how enterprise software is bought. Incumbent vendors are embedding agentic features into their existing products to defend market share, but enterprises are simultaneously running indirect projects to exploit those features and direct transformation projects to build custom agent applications. Gartner forecasts that combination will create a $1.2 trillion services opportunity by 2030.
Services shift from transformation to exploitation
A subtle but important shift is underway in how enterprises use services firms. Two years ago, the typical AI services engagement was a large transformation programme: strategy, change management, full-stack implementation. Now, enterprises are turning to providers for smaller, targeted projects that extract value from the AI features already embedded in their incumbent systems.
This is both good news and a warning for suppliers. Good news because it lowers the barrier to entry: a buyer does not need a board-approved transformation budget to deploy an agent. Warning because it raises the bar on time-to-value. If your agent platform cannot demonstrate measurable ROI inside a single quarter, the buyer will find one that can.
Domain-specific models are opening a new front
Near-term growth in AI application development platforms has been revised up from 28% to 39% for 2026, as enterprises and software providers race to build custom AI applications tailored to specific needs. For model providers, the pressure to offer cost-efficient models aligned to enterprise use cases is creating a growing opportunity for domain-specific language models.
This is particularly relevant for agent builders. A general-purpose model can handle a wide range of tasks reasonably well. A domain-specific model can handle industry workflows with higher accuracy, lower latency and at a fraction of the inference cost. For regulated sectors — finance, healthcare, legal, defence — the compliance and auditability advantages of domain-specific models may prove decisive.
The trough of disillusionment is real, but it is also useful
Gartner places generative AI firmly in the "Trough of Disillusionment" in 2026. That sounds negative, but for experienced technology buyers it is a signal to pay attention. The trough is where hype separates from substance. It is where the vendors with real products, real customers and real unit economics survive, and where the rest fade.
Enterprises are using this moment to grow operational efficiency, automate workflows and improve customer engagement through the simpler embedded AI features that incumbent software vendors are shipping now. The agents that prove themselves during this phase will be the platforms that dominate the next growth cycle.
What this means for Agentic Expo buyers and suppliers
A $2.7 trillion market creates a lot of noise. The buyers who navigate it successfully will be the ones who see agents working in person, compare platforms side by side, and ask hard questions about security, governance and integration. The suppliers who win will be the ones who can articulate exactly where their agents fit inside an enterprise architecture, what they cost to run at scale, and how they prove compliance.
Agentic Expo brings both sides together in one room. If you are buying or supplying agentic AI in 2026, the decisions you make in the next six months will define your position for the next five years. Make them with full information.
Sources: Gartner Inc. worldwide AI spending forecast, published 21 September 2026, as reported by SiliconANGLE and IT-Online. All figures and quotes attributed to Gartner and its analysts.