The Musk Paradox
Elon Musk spent February branding Anthropic as a menace. He tweeted that “Anthropic hates Western Civilization” and amplified a false claim about their AI safety practices. Now he’s their biggest compute sugar daddy. At Code with Claude 2026, Anthropic CEO Dario Amodei announced a jaw dropping deal with SpaceX to use the full capacity of the Memphis data center’s Colossus 1 supercomputer. That machine packs over 220,000 NVIDIA H100, H200, and GB200 GPUs. Musk conveniently forgot his anti Anthropic crusade, tweeting Wednesday that he spent a week with their team and “no one set off my evil detector.” This is the same man who sued OpenAI for cozying up to Microsoft.
Desperate Times, Desperate Measures
Anthropic’s exploding demand for Claude Code is driving them into bed with anyone holding a GPU. The company has been drowning in outages and angry developers complaining about usage limits on Hacker News and Reddit. Their response was insulting: introducing peak hour limits and even testing a Claude Code removal from the $20 Pro plan. Now they’re doubling five hour limits for Pro and Max subscribers and removing peak hour caps. But this SpaceX deal is just one of many. Anthropic reportedly signed massive infrastructure deals with Microsoft, Google, Amazon, and Nvidia last month. The company is racing to keep up as developers flee OpenAI over military contract controversies, but the Faustian bargains are stacking up. When your compute lifeline comes from a man who called Western civilization your enemy, maybe the limits weren’t the real problem.
Source: Arstechnica
