Your Claude Code Limits Just Doubled. The Reason? Anthropic Signed a Deal With Elon Musk’s SpaceX

Anthropic just announced a compute partnership with SpaceX that gives it access to the entire Colossus 1 data center in Memphis, Tennessee. That is over 220,000 NVIDIA GPUs and 300 megawatts of raw computing power, available within the month.

The immediate result: Claude Code usage limits have been doubled for Pro, Max, Team, and Enterprise subscribers. Peak hour restrictions are being removed. And API rate limits are seeing increases of up to 1,500% for some tiers.

If you have been hitting usage limits on Claude for the past few months, this is the fix you have been waiting for.

What Actually Changed Today

Starting Wednesday, May 7, 2026, Anthropic is rolling out the following changes across all paid tiers.

Claude Code limits doubled. The five-hour usage caps for Claude Code have been doubled on Pro ($20/month), Max 5x ($100/month), Max 20x ($200/month), Team, and Enterprise plans. If you were running out of Claude Code capacity in an hour before, you now get twice the runway.

Peak hours restrictions removed. Claude Code previously throttled Pro and Max users during high-demand periods. That throttling is gone. You now get your full allocation regardless of when you use it.

API rate limits massively increased. Tier 1 API users are getting a 1,500% increase in maximum input tokens per minute and a 900% increase in maximum output tokens per minute. Higher tiers are getting proportional boosts.

Claude Opus API rate limits raised. Opus users were among the most constrained. Those limits have been raised significantly, though Anthropic has not published exact new numbers.

The only plan not getting an upgrade is the free tier.

The SpaceX Deal: Why Anthropic Went to Its Biggest Rival’s Doorstep

This is where the story gets interesting. Colossus 1 is not a SpaceX facility. It is an xAI facility — the AI company founded by Elon Musk, who has been one of Anthropic’s most vocal critics.

In February, Musk wrote on X that Anthropic “hates Western civilization.” By May, he was saying he spent a lot of time with senior Anthropic team members and was “impressed.”

The math behind this unlikely partnership is straightforward. Colossus 1 houses over 220,000 NVIDIA GPUs — a mix of H100, H200, and GB200 accelerators. xAI’s total GPU fleet is around 500,000. By handing over Colossus 1, Musk is giving Anthropic roughly half of xAI’s total compute. In exchange, SpaceX gets revenue from infrastructure that was built for Grok but can generate cash from a competitor willing to pay market rates.

For Anthropic, this deal solves a crisis that was becoming existential. Over the past three months, Claude users had been complaining in growing numbers about hitting usage limits far too quickly. Claude Code Max subscribers reported running out of capacity in as little as 19 minutes. Pro users reported being maxed out within hours of their weekly reset. The backlash was driving developers to OpenAI’s ChatGPT Codex and other alternatives.

Why Anthropic Was Running Out of Compute

The core problem is that Anthropic’s subscription pricing does not cover the actual cost of compute consumed. A Pro subscriber paying $20 per month can easily consume $200 worth of tokens in a single Claude Code session. The gap between revenue and compute cost was unsustainable at scale.

In February 2026, Anthropic signed a $25 billion deal with Amazon for 5 gigawatts of compute capacity. But building data centers takes time. That capacity will not come online for months or years. In the meantime, every new agentic product Anthropic ships — Claude Code, Claude Design, Cowork — burns inference at rates that were exceeding available supply.

The SpaceX deal is a bridge. It gives Anthropic 300 megawatts of immediate capacity while the Amazon infrastructure ramps up. It is expensive, but it is available now, and “now” is what Anthropic needed.

Space-Based Compute: The Long-Term Play

Buried in the announcement is a detail that sounds like science fiction but is apparently serious. Anthropic has “expressed interest” in partnering with SpaceX to develop multiple gigawatts of orbital AI compute capacity.

The logic is not as crazy as it sounds. Data centers are increasingly limited by three constraints: power supply, cooling, and land. Space solves two of those — solar power is unlimited and cooling is free in the vacuum of space. The bottleneck becomes launch costs and latency, both of which SpaceX is uniquely positioned to address.

This is a long-term exploration, not a near-term product. But it signals where Anthropic thinks the compute race is heading.

What This Means If You Use Claude

Claude Code users: Your limits just doubled. If you were on Max 5x and running out in 2-3 hours, you should now get 4-6 hours before hitting caps. Peak hour throttling is gone, so late-night coding sessions will feel the same as 2 AM sessions.

API developers: If you were building apps on the Claude API and getting rate-limited, your throughput just jumped dramatically. Tier 1 users went from constrained to having 15x the input capacity. Check the updated rate limit documentation for your specific tier.

Pro subscribers ($20/month): You get doubled Claude Code limits and no more peak-hour restrictions. This is the biggest quality-of-life improvement since the Pro plan launched.

Free users: Nothing changes for you. The free tier is not getting any limit increases from this deal.

The Bigger Picture

This deal is happening just weeks before Anthropic’s expected IPO in June 2026. Solving the compute shortage before going public is not a coincidence. Investors need to see that the product works reliably at scale, and “our users run out of capacity in 19 minutes” is not a story you want in your S-1 filing.

It is also a signal that the AI compute war is entering a new phase where rivals become suppliers. Musk and Dario Amodei have been publicly antagonistic for years. The fact that they cut a deal anyway tells you how much money is on the table — and how desperate Anthropic was for capacity.

For Claude users, none of the politics matters. What matters is that the tool you are paying for should now work the way it is supposed to. More compute, fewer limits, no more “you have been rate limited” in the middle of a coding session.

The changes are rolling out starting today.