Short answer
AI workloads consume significant GPU compute time. This counter estimates the number of GPU-hours used by AI systems today.
What this means
Artificial intelligence is increasingly embedded in digital activity, infrastructure, research and everyday workflows.
This indicator helps quantify one aspect of that transformation and makes the underlying trend easier to follow over time.
Methodology
This estimate uses a rough compute proxy based on assumed global AI GPU-hours per day and converts it into a per-second rate. Methodology.
Related indicators
FAQ
What is a GPU-hour?
One GPU-hour represents one graphics processor running for one hour.
Does this include training and inference?
Yes. The estimate is intended to represent aggregate AI compute activity, including both training and inference.
Why is this estimate uncertain?
Hardware utilization, model size, batching, optimization and infrastructure scale vary widely and are not fully disclosed.
