Gartner puts worldwide AI spending at $2.67 trillion in 2026, up 49.5%, with infrastructure taking more than half

Break Gartner's table apart and the shape gets clearer. AI infrastructure climbs from $981.92 billion in 2025 to $1.4844 trillion in 2026 and a projected $1.9777 trillion in 2027 — that single segment is about 55.6% of 2026's total. AI services rise from $434.05 billion to $576.48 billion, then $745.66 billion in 2027; AI cybersecurity goes from $25.92 billion to $51.35 billion and then $85.997 billion, more than tripling in two years; AI software goes from $288.17 billion to $461.64 billion and then $656.35 billion. The fastest growth sits in the smallest buckets: AI data moves from $826 million to $3.13 billion to $6.48 billion, while AI agents and assistants go from $16.48 billion to $29.22 billion and then $65.47 billion — nearly doubling in a year and doubling again the year after. Generative AI models themselves, meanwhile, come to $28.27 billion in 2026, about 1.06% of a $2.67 trillion total. The figures come from Gartner's September 2026 table, Worldwide AI Spending by Market, 2025-2027.

Lovelock's sentence deserves unpacking. He says demand for AI infrastructure — including AI-optimised IaaS, AI-optimised servers, AI network fabric, and AI processing semiconductors and devices — to support anticipated future workloads remains strong and, in his words, inelastic to pressures from memory-related pricing increases. In other words, even as memory prices push up the cost of a single machine, buyers are not backing off. He attributes that to scale: capacity growth from hyperscalers and service providers purchasing AI-optimised servers will continue to be the largest single area of spending, and he describes the data center buildout as the largest infrastructure project humanity has ever undertaken. The subheading of Gartner's release carries a judgement of its own: AI infrastructure investment and AI being embedded into software and services are converging, and that convergence is driving spending to unprecedented levels. For anyone writing a budget, the practical reading is that the compute bill and the software bill are merging into one sheet.

On the software side, the logic is self-defence. Gartner says vendors across different software types are rapidly embedding agentic AI into existing products to stay relevant and to defend against new cross-functional agents. And in 2026 generative AI sits firmly in the Trough of Disillusionment, so enterprises are reaching for the simpler embedded AI features their incumbent software providers already ship — to grow operational efficiency, automate workflows, improve customer engagement and sharpen decision-making. Lovelock adds a sharp observation: enterprises are turning to service providers less often for business transformation and more often for smaller indirect projects that squeeze value out of the AI features already inside their software. As for vendor lock-in, data sovereignty and runaway costs, Gartner's reading is that those risks are not deterring buyers from adopting proprietary capabilities, and that the combination of transformation and indirect projects will drive a $1.2 trillion AI services opportunity by 2030.

Two revisions in this forecast are worth noting. On the short term, 2026 growth for AI application development platforms was raised from 28% in the previous forecast to 39%, as enterprises, software providers and services firms build custom AI applications tailored to their own needs. Growth for generative AI models was raised from 110% to 117%, which Gartner attributes to pressure on model providers to ship cheaper models aligned to enterprise use cases — opening a small but growing opportunity for domain-specific language models. On the long term, Gartner separated cross-functional agents and assistants out of AI software and added consumer agents and assistants to the AI spending forecast for the first time. Two caveats matter: every figure here is a forecast, not an actual, and Gartner's definition is deliberately broad, spanning infrastructure, services, software, security, platforms, models and data, so it cannot be compared directly with narrower trackers that count hardware only. Third-party summaries circulating at $2.59 trillion and 47% growth come from an earlier version; the current release says $2.67 trillion and 49.5%. Gartner also flagged a free webinar on October 8 about how soaring memory prices are changing technology spending, which tells you memory cost has joined the list of key variables in its model.