Anthropic signed a deal for SpaceX’s Colossus 1 data center (220,000+ NVIDIA GPUs, 300 MW) and immediately raised Claude Code and API rate limits.
Key Takeaways
Claude Code five-hour rate limits doubled for Pro, Max, Team, and seat-based Enterprise plans, effective today; peak-hour reductions removed for Pro and Max.
API rate limits for Claude Opus models raised considerably; specific tier table published in the announcement.
Colossus 1 access joins an already large compute stack: up to 5 GW with Amazon, 5 GW with Google/Broadcom, $30B Azure capacity with Microsoft/NVIDIA, and $50B with Fluidstack.
International expansion targets regulated industries (financial services, healthcare, government) with in-region inference in Asia and Europe via the Amazon collaboration.
Anthropic also expressed interest in orbital AI compute with SpaceX, framed as a multi-gigawatt long-term possibility.
Hacker News Comment Review
The five-hour rate limit doubling drew immediate skepticism: weekly caps remain unchanged, so heavy users may simply exhaust the same total tokens faster rather than getting meaningfully more capacity.
Colossus 1’s environmental record is a flashpoint. Commenters cited illegal power use, air quality violations near Memphis, and grid stress risk during extreme weather, putting Anthropic in the same category as other hyperscalers criticized for similar practices.
The business logic of the SpaceX deal reads differently depending on framing: SpaceX monetizes Grok/xAI infrastructure ahead of a likely IPO, while Anthropic gets fast capacity it cannot build quickly enough itself; commenters see this as mutual opportunism, not ideological alignment.
Notable Comments
@mirzap: argues the rate-limit doubling is “a marketing stunt” if weekly limits stay fixed, letting users hit the same ceiling in fewer days.
@antipaul: notes Anthropic is getting Colossus 1 only; SpaceX/xAI retains Colossus 2 with double or more the GPUs, so xAI is not exiting AI compute.
