Anthropic adding computing capacity from Musk’s SpaceX

Anthropic has secured a deal with SpaceX to access computing power from its Colossus 1 data center, which contains over 220,000 Nvidia GPUs. The agreement, announced Wednesday, aims to boost capacity for Anthropic’s AI models, Claude Pro and Claude Max, which have seen increased demand. SpaceX, now operating as SpaceXAI after acquiring Elon Musk’s AI firm xAI, stated the cluster is optimized for large language models, multimodal systems, and frontier-scale generative AI. Anthropic also expressed interest in collaborating with SpaceXAI to develop orbital AI compute capacity, potentially reaching multiple gigawatts. The partnership marks a shift for Musk, who previously criticized Anthropic on social media, calling its AI model ‘misanthropic and evil’ and questioning its alignment with humanity. Musk later clarified his change of heart, citing interactions with Anthropic’s leadership, whom he described as ‘highly competent’ and committed to ethical AI development. In a post-meeting statement, Musk said he no longer detected ‘evil’ in Anthropic’s work, provided the company engages in critical self-examination. He added that SpaceXAI had already transitioned AI training to Colossus 2, making the lease agreement feasible. The move follows a surge in downloads for Anthropic’s AI tools over the past year. Anthropic also announced increased usage limits for certain subscribers, including doubling Claude Code’s five-hour rate limits for select plans and raising API limits for Claude Opus models. The company’s decision to partner with SpaceX reflects broader industry trends toward leveraging advanced computing infrastructure to support AI innovation.