AI Labs: Who Has the Most Compute Power?

A ranking of the organizations with the largest current and planned H100 GPU clusters, the real-world limit on AI progress.

The top 3

  1. Meta's GPU Cluster Dominance: Meta currently leads the industry with a massive cluster comprising over 600,000 H100 GPUs, significantly boosting their Llama training capacity. Meta
  2. Microsoft-OpenAI Infrastructure Synergy: Through their partnership, Microsoft provides the Azure compute backbone that supports OpenAI's large-scale training, representing one of the largest combined footprints. Microsoft
  3. xAI's Memphis Supercluster: xAI recently launched 'Colossus,' a cluster of 100,000 H100 GPUs, making it one of the most powerful training systems commissioned in the shortest timeframe. xAI

Sources

Open the full topic