Meta details Prometheus, the Ohio supercluster heading for 1 million GPUs

The Next Web
Read the full article
Meta's Prometheus supercluster in Ohio currently operates hundreds of thousands of Nvidia GPUs and aims to scale up to about 1 million GPUs. The cluster supports training of Meta's Muse models and has achieved significant reliability improvements despite early challenges with hardware faults and power management.

More on this story


More in Chips & Compute

Blackstone Bet on Chips Drove a Debt Boom That’s Upending Credit

Covered by 2 sources
Chips & Compute4 min read

On the ground at Nvidia's AI showpiece: Jensen-mania, robot ducks and neocloud distress

CNBC
Chips & Compute6 min read

Google turns the infrastructure behind Gemini’s agents into a cloud service

Covered by 2 sources