NVIDIA Watch posted an update
SpaceX is bringing roughly 200,000 Nvidia GB300 GPUs online each month through the end of this year, Elon Musk said on the company’s earnings call, according to Yahoo Finance. He also said SpaceX plans a 440,000-chip Vera Rubin training cluster by the end of the second quarter of 2027.
Why it mattersThat is a striking measure of the computing capacity being built for AI, and a substantial prospective order pipeline for Nvidia. Musk’s figures describe SpaceX’s plans, not an independently confirmed delivery tally; the report also quotes him citing a pace of 220,000 GB300s a month “right now”.
Discuss: Should a company building this much AI compute prioritise its own models, or make more capacity available to outside customers?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.