xAI adding 200,000 GB300s a month, 10GW of compute planned next year
gaganghotra_ · x · 2026-10-22
Brad Gerstner revealed SpaceX/xAI will build out roughly 25% of all new US compute next year. Musk says xAI is now standing up 200,000 GB300s per month — what it did in a year a few months ago — hitting 3GW by year-end and targeting 10GW next year, 'unheard of' scale. gaganghotra poses the open question: Elon is winning the compute game yet still behind on models, with Claude outperforming Grok on many tasks.
Related event: Musk reveals xAI scaling to 200,000 GPUs per month(2 posts)→
More from Infra
- OpenAI's Jalapeño Inference Chip Beats Nvidia Blackwell on Perf-per-Watt, Says Hardware VP Richard Ho — firstadopter · 2026-10-22
- Kobe Steel quietly makes top-tier niche semiconductor gear, keiretsu breadth on display — PAstynome · 2026-10-22
- npm hits install and publishing outage, theo jokes he's 'retiring to the woods' — theo · 2026-10-22
- LLM inference prices fell ~1600x in two years, from $32 to $0.02 per million tokens — FinanceYF5 · 2026-10-22
- MiniMax M3.1 Flash serving picks vLLM, with Inferact as API launch partner — woosuk_k · 2026-10-22
- Why Nvidia keeps its market share: deep systems integration and Dell partnerships ASIC rivals lack — PTrubey · 2026-10-22