Anthropic Doubles Claude Code Rate Limits, Leases All of SpaceX's Colossus 1
Anthropic doubled Claude Code's 5-hour rate limits and removed peak-hour throttling for Pro, Max, Team, and seat-based Enterprise plans, then disclosed the source of the new capacity: a lease on the entire Colossus 1 data center — 300 MW and ~220,000 NVIDIA GPUs in Memphis — that SpaceX absorbed when it took over xAI.
How much are Claude Code's rate limits increasing and why?
Claude Code's 5-hour rate limits are doubling for Pro, Max, Team, and seat-based Enterprise plans, and API limits for Opus are rising significantly, enabled by a new compute partnership with SpaceX.
TL;DR
5-hour rate limits doubled for paid Claude Code tiers; peak-hour throttling scrapped on Pro/Max. Opus API limits jump up to +1,500% input / +900% output at Tier 1. The compute unlock is a full lease of Colossus 1 — 300 MW, ~220K NVIDIA GPUs (H100/H200/GB200) — formerly xAI's Memphis flagship, now SpaceX-owned. Orbital compute floated but unsigned. Free tier excluded; weekly caps unchanged.
On May 6, 2026, Anthropic announced it would double Claude Code's 5-hour rate limits for Pro, Max, Team, and seat-based Enterprise subscribers, scrap peak-hour throttling for Pro and Max, and lift ClaudeOpus API rate limits substantially. At Tier 1, Opus gets a +1,500% boost in input tokens per minute and +900% in output, with proportional bumps at higher tiers. The free plan is excluded and weekly caps remain unchanged.
Key facts
5-hour Claude Code limits doubled on all paid tiers (Pro / Max / Team / seat-based Enterprise).
Peak-hour throttling removed for Pro and Max.
Opus API: Tier 1 +1,500% input tokens/minute, +900% output tokens/minute.
The data center is the former xAI flagship, transferred to SpaceX when xAI was absorbed.
Free tier excluded. Weekly caps unchanged.
The SpaceX angle is terrestrial, not orbital
The headline is misleading if you picture rockets or Starlink. The compute unlock is a Memphis-based GPU cluster: SpaceX inherited Colossus 1 when it took over xAI; xAI's training had already migrated to the larger Colossus 2; and Anthropic is now leasing the freed capacity in full. ~220K GPUs and 300 MW makes it one of the largest single-tenant AI compute leases announced this year.
“The capacity unlock isn't Starlink and isn't orbital — it's a 300 MW, 220,000-GPU data center in Memphis that SpaceX inherited when it absorbed xAI.”
Anthropic and SpaceX did separately float a future deal for multi-gigawatt orbital compute, tied to SpaceX's FCC filing for roughly one million satellites. No agreement has been signed — that piece is purely aspirational.
Five compute commitments now stacked
Counting Colossus 1, Anthropic now has five publicly-disclosed compute commitments running in parallel: AWS (≤5 GW), Google / Broadcom (5 GW landing in 2027), a $30 B Microsoft / NVIDIA Azure deal, a $50 B Fluidstack agreement, and now SpaceX. The pattern reads as deliberate capacity diversification ahead of an Anthropic raise reportedly at a $900 B valuation and SpaceX's expected fall IPO.
A notable rapprochement
The deal is striking given the recent backstory. Anthropic previously cut xAI off from its API, and Musk has publicly described Anthropic as "misanthropic". On Wednesday Musk reversed course on X, writing that Anthropic's team was "highly competent and cared a great deal about doing the right thing… Claude will probably be good," while reserving the right to reclaim compute "if Claude engages in actions that harm humanity." In the same thread he announced xAI will dissolve and be renamed SpaceXAI.
Competitive context
Claude Code competes with GitHub Copilot, Cursor, and Replit's agentic features. Doubling rate limits without raising prices is the clearest competitive lever Anthropic has pulled this year — particularly because peak-hour throttling has been one of the most-cited friction points in developer feedback. No competitor has published a public response as of this writing.
What to watch
Whether Colossus 1 actually comes online in the stated month, and whether Claude Code's effective throughput matches the headline rate-limit math.
Whether Anthropic's Q3 2026 ARR disclosure shows Claude Code seat counts crossing 100K — a benchmark the doubled limits make easier to defend.
The orbital compute MoU: it's currently aspirational, but a signed deal would be the first commercial commitment to space-based AI inference at gigawatt scale.
AI-assisted reporting. Generated by gentic.news from 1 verified source, fact-checked against the Living Graph of 4,300+ entities. Edited by Ala SMITH.
Following this story?
Get a weekly digest with AI predictions, trends, and analysis — free.
AI Analysis
The SpaceX partnership is the most interesting signal here, not the rate-limit increase itself. Anthropic already has deep ties with AWS and Google Cloud; adding SpaceX suggests either a capacity bottleneck at traditional hyperscalers or a strategic bet on non-cloud compute for inference. SpaceX's Starlink ground infrastructure could theoretically offer distributed inference nodes closer to users, reducing latency for real-time coding agents. However, the announcement is thin on details — no financial terms, no duration, no technical description of how SpaceX compute is provisioned. This smells like a compute-arbitrage play: Anthropic likely secured reserved GPU capacity at below-market rates in exchange for long-term commitment. If true, it signals that GPU supply remains constrained for frontier labs despite increased production. The rate-limit doubling is a tactical move to retain Claude Code users who might defect to Cursor or Copilot; the SpaceX deal is the structural story worth tracking.
