Global spending on artificial intelligence is forecast to reach $2.7 trillion in 2026, representing a 49.5% increase from 2025, according to Gartner. The growth is being driven not only by AI software and services, but increasingly by the infrastructure required to run AI workloads at scale.
Gartner's latest forecast puts total worldwide AI spending at $2.67 trillion in 2026, up from $1.79 trillion in 2025. AI infrastructure accounts for the largest share of this spending, reaching almost $1.5 trillion this year.
AI infrastructure becomes the largest part of the AI economy
The numbers show that the AI boom is increasingly an infrastructure story.
Gartner forecasts spending on AI infrastructure at $1.48 trillion in 2026, compared with $981.9 billion in 2025. That means infrastructure alone represents more than half of projected global AI spending this year.
This category includes AI-optimized infrastructure such as cloud infrastructure, servers, networking, processing semiconductors and other technologies required to support increasingly demanding workloads.
Gartner describes the buildout of AI data-center capacity as the largest infrastructure project humanity has undertaken, with hyperscalers and service providers continuing to invest heavily in AI-optimized computing capacity.
AI-optimized cloud infrastructure is growing even faster
The growth is particularly visible in cloud infrastructure.
Worldwide spending on AI-optimized infrastructure as a service (IaaS) is projected to reach $42.3 billion in 2026, up 96.4% from 2025. Gartner expects the market to reach $66.1 billion in 2027.
For comparison, total IaaS spending is forecast to grow by 29.3% in 2026. This means AI-optimized infrastructure is expanding at more than three times the growth rate of the overall IaaS market.
Inference is becoming the new cloud workload
Another important shift is happening underneath the headline spending figures: AI inference is becoming more important than model training.
Gartner forecasts global spending on AI inference at $23.3 billion in 2026, compared with $19 billion for training. Inference is therefore expected to account for 55% of AI-optimized IaaS spending this year.
The change reflects the transition from experimenting with AI models to putting them into production. Once AI applications are deployed, models need to process requests continuously — whether they are powering enterprise assistants, recommendation systems, automated workflows, search, image generation or other applications.
This creates a different infrastructure challenge from training. Instead of large but periodic training runs, production AI requires reliable, scalable and cost-efficient compute that can handle continuously changing demand.
Why this matters for cloud infrastructure
The $2.7 trillion forecast illustrates how quickly AI is becoming part of the broader cloud infrastructure economy.
AI applications require significantly more than GPUs alone. Production environments also depend on high-performance networking, storage, virtualization, orchestration, data pipelines, monitoring and reliable power and cooling infrastructure.
For cloud providers, this is shifting the focus from simply offering access to computing resources toward building infrastructure specifically optimized for AI workloads.
For businesses, the challenge is different: deciding how much AI infrastructure they actually need, where workloads should run, and how to control costs as AI applications move from testing into production.
The next stage of the AI infrastructure race
Gartner expects total AI spending to rise to approximately $3.64 trillion in 2027. AI infrastructure is projected to approach $2 trillion next year, while AI-optimized IaaS spending is expected to reach more than $66 billion.
These figures suggest that the next phase of AI development will be defined less by whether companies adopt AI and more by how efficiently they can operate it at scale.
As inference becomes the dominant workload, cloud infrastructure will increasingly become a critical part of AI strategy — and the economics of compute, scalability and resource utilization will matter just as much as the models themselves.
For the cloud industry, the AI boom is no longer simply about building better models. It is about building enough infrastructure to run them.
