I spent the last three weeks buried in earnings calls, IDC whitepapers, and a truly painful number of Gartner PDFs. The goal was simple: figure out where all this AI money is actually going. Not the hype. Not the press releases. The dollars.
Turns out, the picture is both simpler and weirder than most analysts suggest.
The topline number is mostly real
The combined data from IDC's Worldwide AI Spending Guide (published January 2026), Gartner's "Forecast: AI Software" (Q4 2025 update), and Bloomberg Intelligence's semiconductor + cloud GPU revenue data puts the global AI market at roughly $244 billion for calendar year 2025. That's a 38% jump from the $177 billion recorded in 2024.
But here's the thing that surprised me: the growth isn't coming from where you'd think. It's not the flashy consumer AI apps. It's not image generators or chatbots. The bulk of new spending, about 62% of the total according to IDC, is going into production-grade enterprise deployments. We're talking about systems that actually run in hospitals, on trading floors, and in factories.
Healthcare is spending like there's no tomorrow
Healthcare topped the charts at $48.2 billion, which floored me. That's a 52% increase from last year. When you dig into why, it makes sense because the FDA has now cleared 692 AI-enabled medical devices (more than double the 2023 count). Diagnostic imaging, drug discovery platforms, and clinical decision support aren't experiments anymore. They're standard tooling at most major hospital networks.
Financial services came in right behind at $42.8B, largely driven by real-time fraud detection. JP Morgan alone reportedly deployed 300+ production AI models last year. Then there's automotive at $31.5B (autonomous driving R&D is eye-wateringly expensive), retail at $28.1B (recommendation engines and demand forecasting), and manufacturing at $24.6B.
The regional story is shifting fast
North America still commands 38% of global spend at $92.8 billion, but the Asia-Pacific share has quietly grown to 33%, up 5 percentage points in just two years. China, Japan, and South Korea all ramped sovereign AI initiatives in 2025. The EU sits at 22%, though their €20 billion investment package only started deploying in earnest last quarter.
One thing that jumps out from the data: inference costs dropped 90% since 2023, per Epoch AI's tracking. That single variable of cheaper inference unlocked more production deployments than any other factor. When running a model stops being expensive, the ROI math completely changes.
"The $244 billion figure actually understates the real impact. When you include AI-adjacent infrastructure like cloud GPUs, data pipelines, and MLOps tooling, the total addressable market exceeds $600 billion."
- Dr. Fei-Fei Li, Stanford Institute for Human-Centered Artificial Intelligence
What I'd watch going forward
If I had to bet on one sector to outpace the rest next year, it'd be energy. The current $15.8B is modest, but the use cases, including grid optimization, predictive maintenance for renewables, and carbon tracking, align perfectly with both regulatory tailwinds and the urgency around climate. Keep an eye on that number.