- What changed
- Meta completed the first phase of Prometheus, its gigawatt-scale supercluster featuring closed-loop liquid cooling, and used it to train the Muse foundational models.
- Why you should care
- Gigawatt-scale AI superclusters utilizing liquid cooling demonstrate the physical scale required for training next-generation foundational models.
- Your move
- Watch. Monitor hardware reliability data from large-scale clusters.
- What to watch next
- Further expansion phases of the Prometheus supercluster toward the million-GPU target.
- Event
- breakthrough
- Event date
- Oct 19, 2026
- Relevant to
- General AI readers
