Anthropic Doubles Claude Code Rate Limits and Signs the Biggest AI Compute Deal in History

Two announcements from Anthropic yesterday, and they're connected in a way that matters.

First, the immediate impact: Claude Code rate limits doubled across Pro, Max, Team, and Enterprise. Peak-hours throttling removed for Pro and Max. Opus API rate limits got a 1500% input token increase and 900% output token increase at Tier 1. All effective immediately.

If you've been batching Claude Code requests to avoid hitting the ceiling, or queuing up long agent runs for off-peak hours, those workarounds may no longer be necessary. I've personally restructured my pipeline to run research agents sequentially instead of in parallel because of rate limits. That constraint just loosened significantly.

Second, the infrastructure backing it: Anthropic signed a deal with SpaceX to lease all compute capacity at the Colossus 1 data center in Memphis. That's 300+ megawatts, 220,000+ NVIDIA GPUs (H100, H200, GB200), available within the month. At the Code with Claude keynote, Dario Amodei disclosed Anthropic grew 80x in Q1 2026 on an annualized basis, pushing annualized revenue past $30 billion.

80x. Not 80%. 80x. That number puts Anthropic in the conversation for one of the fastest revenue ramps in tech history.

The SpaceX deal also includes something that sounds like science fiction: expressed interest in developing multi-gigawatt orbital AI compute via SpaceX satellites. Details from the filing describe each satellite carrying 100 kilowatts of AI hardware, with several thousand satellites needed for multi-gigawatt capacity. SpaceX itself warned this involves "significant technical complexity and unproven technologies." I don't know if this is real or aspirational, but Anthropic filing paperwork about it suggests someone is at least running the numbers.

The Latent Space analysis frames SpaceX as "the kingmaker picking a side" in the AI compute race. That's not wrong. When your compute provider is also heading toward a $1.75-2T IPO and has a strategic interest in proving AI infrastructure revenue, you're not just a customer. You're a story they're telling Wall Street.

For builders: revisit any architectural decisions you made because of rate limits. Batching, queuing, throttle-back logic. Some of it may now be unnecessary overhead.