Summary: Gartner forecast on January 15, 2026 that worldwide AI spending will reach $2.52 trillion in 2026, 44% above 2025’s $1.76 trillion, and $3.34 trillion in 2027. AI infrastructure accounts for $1.37 trillion, while AI cybersecurity nearly doubles to $51.3 billion. Gartner also places AI in the Trough of Disillusionment throughout 2026 and says incumbents will often sell AI into enterprises.
Gartner released a fresh worldwide AI spending forecast on January 15, 2026, and its headline is $2.52 trillion for 2026 — 44% above the $1.76 trillion of 2025 — with the total set to reach $3.34 trillion in 2027. Compute, services, software, models, and security all roll into that single sum, making it the fullest one-number baseline available to anyone deciding what AI products to build or buy this year.
The forecast also pins AI to the Trough of Disillusionment on Gartner’s Hype Cycle for the entirety of 2026 — the stretch in which pilots and rollouts keep falling short of their promises and market expectations cool. Spending has not stopped; what is shifting is how it gets spent. Read next to our opening outlook on AI in 2026, the report adds the quantitative grounding that story was missing.
The Breakdown: Which Segments Capture the Cash
- AI infrastructure: $1.37 trillion in 2026, up from $965 billion in 2025 — one segment swallowing more than half the total — and projected to reach $1.75 trillion in 2027
- AI services: $588.6 billion, versus $439.4 billion a year earlier
- AI software: $452.5 billion, versus $283.1 billion a year earlier
- AI cybersecurity: $51.3 billion, close to twice the $25.9 billion of 2025
- AI models come in at $26.4 billion; data science and ML platforms at $31.1 billion
AI-optimized servers on their own are set to expand 49% in 2026 and claim 17% of the year’s total AI outlay. Drawing on Gartner’s numbers, ITPro adds that AI-related semiconductors — processors, high-bandwidth memory, and networking components — already made up close to a third of all semiconductor sales in 2025. The upstream of the stack was repriced around AI some time ago.
How Purchasing Works While the Trough Lasts
John-David Lovelock, Gartner distinguished VP analyst, is blunt about the consequence: with AI stuck in the Trough of Disillusionment for all of 2026, it “will most often be sold to enterprises by their incumbent software provider rather than bought as part of a new moonshot project.” The corollary he attaches: until ROI becomes more predictable, AI cannot genuinely scale within the enterprise.
A sobering datapoint drawn from MIT research appears in ITPro’s coverage: 95% of organizations surveyed in 2025 said their generative AI projects had returned nothing. Hold the two signals side by side and the infrastructure-heavy mix stops looking odd — the foundation is treated as spend that happens whichever application wins, while returns at the application layer remain unproven. Gartner adds a caution of its own: AI adoption is shaped fundamentally by the readiness of human capital and organizational processes, with the dollars invested only a secondary factor.
Why Security Spending Doubled
Nothing in the forecast grows faster than AI cybersecurity, which doubles from $25.9 billion to $51.3 billion. The driver is the reality of models running in production at scale: prompt injection defenses, data leak prevention, and output compliance are moving from afterthought patches onto budgeted line items. Wire AI into existing workflows and the protection costs land on the same purchase order — a plainly favorable wind for security vendors.
Takeaways for Developers and Procurement Teams
One: as infrastructure and security take a bigger slice of the budget, platform and tooling teams gain leverage inside their companies, so headcount planning should start early. Two: “sold, not bought” means a large share of AI capability will arrive as upgrades from your incumbent SaaS vendors — audit those vendors’ AI roadmaps before pitching anything custom, and redraw the build-vs-buy line from there. Three: proven ROI now gates scaling — run measurable, small-scope deployments first, bank the numbers, and only then argue for expansion. Teams working with scarce resources should treat that discipline as an advantage rather than a constraint.