Anthropic 签约使用 SpaceXAI Colossus 超级算力
anthropic 與 spacexai 簽署協議，使用位於孟菲斯的 colossus 1 超級計算機，新增 300 兆瓦容量和 22 萬顆 nvidia gpu 以緩解 claude pro 和 max 訂戶的速率限制。今年早些時候合併了 xai 和 spacex 的 spacexai 計劃下個月上市，並表示 anthropic 有興趣開發未來的軌道 ai 計算。此前有報道稱 anthropropic 向谷歌雲承諾了 2000 億美元。
Anthropic agreed to use computing resources from SpaceXAI's Colossus 1 supercomputer in Memphis, Tennessee. SpaceXAI, the merged entity of SpaceX and xAI, operates a facility containing approximately 220,000 Nvidia GPUs (H100, H200, GB200) with 300 megawatts of capacity. Anthropic intends to direct this power toward reducing rate limits for Claude Pro and Claude Max subscribers, noting that developers average 20 hours per week on Claude Code.
The partnership supports SpaceXAI's goal to go public as soon as next month, offering a customer anchor for its expensive infrastructure. SpaceXAI revealed Anthropic expressed interest in developing "orbital AI compute capacity." This deal supplements Anthropic's existing reliance on Amazon and Google, following reports of a $200 billion commitment to Google's TPUs.