📈 How Much AI Compute Power Exists Globally? How Fast Is It Growing?

Since 2019, the total compute power of Nvidia chips has doubled every 10 months on average.

🟢 Hopper GPUs Dominate the Market

Currently, 77% of Nvidia’s AI compute power comes from the Hopper architecture (H100, etc.).

There are approximately 4 million H100-equivalent GPUs in global data centers.

🟢Still Not Fast Enough

Although compute power is growing at a rate of 2.3x per year,

AI model training demands are increasing even faster—by 4-5x annually.

This supply-demand gap could become a bottleneck in AI development.