How AI Data Centers
Use ENERGY

AI data centers are scaling faster than the grid can adapt. Hardware choice, energy source, and location are the three levers that determine whether that growth is sustainable — and the efficiency gap between best and worst practice is already enormous.

Our conclusion: Optimizing all three levers together cuts total energy footprint by orders of magnitude more than any single factor alone.

Explore the Data →
PUE Range (Data Centers) 1
1.1–2.4x
Power Usage Effectiveness
Carbon Variance by Grid Region 2
8x
Same workload, but a very different footprint
GPU Efficiency Improvement (Gen-Over-Gen) 3
2.5X
But AI workloads grew 10x in the same period