<img height="1" width="1" style="display:none;" alt="" src="https://px.ads.linkedin.com/collect/?pid=1448210&amp;fmt=gif">
Free Research

Cloud AI External Workload Capacity Consumption by Type: 2026 to 2035

Checking your access...

SOURCE RESEARCH

AI Cloud Workloads

Market Data | 2Q 2026 | MD-AICW-101

Get The Report

Inference workloads are set to surpass training workloads in total capacity consumption by 2033 and will exceed 45 GW of compute capacity by 2035.

Training workloads will continue to scale aggressively during the forecast period, surpassing 35 GW in 2035. This will be driven not only by foundation model training, but by fine-tuning workloads as well.

Inference capacity consumption growth will begin to rapidly accelerate in 2030, due to better token economics for enterprises, and improved model capabilities, unlocking higher impact and stronger Return on Investment (ROI) use cases for end users