Thread

1. ttul · 1789608838 · focus · HN ↗

   $5 per second if my eyes don’t fool me. That’s ~$432K per day. Enough to rent 3,000 B300 nodes on Modal.

   1. stymaar · 1789636026 · focus · HN ↗

      Which isn't that much when you compare to the kind of DC that US actors are using.

      1. brookst · 1789647079 · focus · HN ↗

         Do we know what kind of DC US actors are using specifically for training, versus inference and delivery?

         1. stymaar · 1789723790 · focus · HN ↗

            I remember Zuck bragging about using a 100MW DC for training, and Musk's Colosus was supposed to be a training data center (but they fucked up the design so they had to repurpose it to an inference one).
      2. lostmsu · 1789659888 · focus · HN ↗

         That's posttraining. Pretraining is the expensive part.

Unofficial Hacker News client; not affiliated with Y Combinator.