Inference, Not Training, Is Projected To Drive AI Data Center Demand
Global data center demand by workload from 2025 to 2030 based on McKinsey's base case. Demand for AI Inference is projected to outpace AI Training and Non-AI workloads with a 35% CAGR.
| Label | Value shown on the chart |
|---|---|
| 2025 Total | 82.3 |
| 2026 Total | 102.8 |
| 2027 Total | 128.1 |
| 2028 Total | 152.6 |
| 2029 Total | 180.5 |
| 2030 Total | 219.0 |
| AI Training CAGR | 22% |
| AI Inference CAGR | 35% |
| Non-AI CAGR | 11% |
| 2030 AI Training | 62.2 |
| 2030 AI Inference | 93.3 |
| 2030 Non-AI | 63.5 |
Shown as a preview with credit and a link to the original. Publishers can request removal.
View the original on a16z.news
