Gartner forecasts AI-optimized IaaS spending will reach $42 billion in 2026
Gartner projects AI-optimized IaaS spending will rise 96.4% to $42.276 billion, with inference set to overtake training.
By Wei-Lin Zhao · AI Correspondent
· 2 min read
Worldwide AI-optimized IaaS spending 2026 will reach $42.276 billion, up 96.4% from $21.529 billion in 2025, Gartner forecasts. The research firm also expects spending on inference, the operational use of trained models to generate responses, recommendations and decisions, to exceed training spending this year, a projected change in mix that Gartner says reflects more production-oriented AI adoption.
The forecast is narrowly scoped. It covers AI-optimized infrastructure as a service, the cloud compute capacity used for large-language-model training and operation, rather than all AI spending, data-center capital expenditure, or the broader AI-infrastructure market spanning hardware, software and services.
Gartner expects the category to expand again to $66.143 billion in 2027, a 56.5% increase. The percentage growth rate would slow from 2026, but the projected dollar increase would be larger: about $23.867 billion in 2027, compared with roughly $20.747 billion this year.
Why is AI-optimized IaaS spending shifting toward inference?
Gartner projects $23.3 billion of global AI-optimized IaaS spending will support inference in 2026, versus $19 billion for training. Inference would account for 55% of the category this year and 59% in 2027, according to the firm.
Hardeep Singh, a senior principal research analyst at Gartner, said organizations are putting fine-tuned and domain-specific models into customer-facing and operational systems. That creates demand for continuous, real-time execution rather than periodic training, in Gartner's assessment. The firm also cites continued LLM-training demand, AI deployment across enterprise applications and workflows, and agentic AI systems that perform multistep autonomous tasks as drivers of cloud consumption.
The distinction matters for providers and enterprise buyers because inference is the running phase of AI. A trained model is used in applications, processes and customer experiences, and Gartner says those uses require ongoing infrastructure capacity. Gartner's figures indicate a forecasted spending-mix change, though they do not independently demonstrate deployment success, utilization, return on investment, or maturity at any individual company.
How large is AI-optimized IaaS relative to cloud infrastructure?
Gartner forecasts total worldwide IaaS spending of $287.347 billion in 2026. On that basis, AI-optimized IaaS would equal about 14.7% of the total, calculated from Gartner's two forecast series. Total IaaS, which includes workloads beyond AI, is projected to grow 29.3% this year, far below the 96.4% rate forecast for the AI-optimized segment.
For teams assessing the numbers, cloud infrastructure is the underlying compute, storage, networking and management layer delivered through cloud services. Gartner's forecast addresses a specialized slice of that spending, not every cost associated with deploying AI.
The main evidence is Gartner's own public forecast and press release. Its projections and explanation of the inference crossover are estimates, not independently validated spending data. The material does not provide company-level deployment, regional, power-use, utilization or cost data.
This story draws on original reporting from CIO Dive.
