Message from September 25th, the world's top Fulone MaskElon MuskTODAY, 25 SEPTEMBER, A TWEET WAS POSTED ON PLATFORM XDetailed disclosure of the flag SpaceXAI CompanyHashrateDeployment。
SpaceXAI currently operates the two main computational clusters of COlossus 1 and Colossus 2, of which Colossus 1 contains 150,000Nvidia H100, 50,000 H200 AND 30,000 GB200And the Colossus 2 cluster consists of 110,000 GB200 and 440,000 GB300。
MASK SAID HE WAS GOING TO SHIP ANOTHER 220,000 GB300 NEXT WEEK AND PLANS TO SHIP ANOTHER 220,000 GB300 IN NOVEMBER 2026, AND IF ALL DEPLOYMENTS GO WELL, ANOTHER 220,000 GB300 IN LATE DECEMBER 2026This means that the Colossus 2 calculation cluster will be equipped with 1.1 million GB300 by the end of this year。
Note: GB200/GB300 is a new generation of GPU chips launched by Nvidia, a follow-up product to the Blackwell architecture, dedicated to the optimization of AI training and reasoning. The GB series, compared to H100/H200, has significantly improved its energy efficiency and numerical density and is now the main force chip of the new xAI cluster。
For ordinary users, supported by these huge British Weeda calculus, SpaceXAI can create a smarter, stronger, and more powerful A.I. product, training more advanced artificial intelligence models, improving product experience, and predicting more accurate answers, more logical reasoning, more natural voice interactions, etc。
When it comes to competition in the AI sector, Mask has made it clear that SpaceXAI has only three years of research and development, and that competitions such as Anthropic and OpenAI have accumulated over six to ten years. He suggested that if the momentum remained strong, SpaceXAI was expected to lead in about six months。