Anthropic Raises Claude Code Limits as Compute Pressure Keeps Building

For Claude and Claude Code users, this is one of those quietly important infrastructure stories: Anthropic is not just shipping model updates, it’s actively reshaping access to make the product usable at scale. If you’ve hit usage caps, peak-hour throttling, or watched multi-agent workflows chew through your allowance, this announcement matters more than most model-launch headlines.

Key Points

Anthropic announced a deal with SpaceX to use the full compute capacity of SpaceX’s data center in Memphis, Tennessee.
Anthropic says the deal will help increase usage limits for Pro and Max subscribers.
Claude Code limits for Pro and Max users were doubled in the five-hour window.
Anthropic removed peak-hours limit reductions for Claude Code on those same accounts.
API limits for the Opus model were also increased.
Anthropic says the SpaceX deal provides access to more than 300 megawatts of new compute capacity.
SpaceX said Colossus 1 includes over 220,000 NVIDIA GPUs, including H100, H200, and GB200 accelerators.
Anthropic also said it is interested in working with SpaceX on “multiple gigawatts” of orbital compute capacity, though that idea is still speculative and unproven.
The deal is notable because Elon Musk had recently been publicly hostile toward Anthropic, but later said he’d spent time with Anthropic leaders and was impressed.
Anthropic says demand for Claude Code and related products has surged, driven by migration from OpenAI, broader adoption in professional software teams, and more compute-heavy multi-agent workflows.
The company has already tried to manage demand with tighter limits, including peak-hour restrictions and a short-lived test involving the Pro plan.
Anthropic also credits earlier compute deals with Microsoft, Google, Amazon, Nvidia, and others for helping it raise limits, though some of that capacity will take time to come online.

My Take

What strikes me is that this is less about a flashy product launch and more about the mundane reality of successful AI software: compute becomes the product bottleneck. That’s especially true for Claude Code, where user demand is likely to be spiky, expensive, and very sensitive to latency and quota frustration.

I think the most meaningful part of the story is the limit increase itself. If you’re building with Claude Code, higher caps and fewer peak-hour penalties are the sort of change you actually feel day to day. A model benchmark is abstract; fewer “come back later” moments is real developer experience improvement.

At the same time, this is also a reminder that AI companies are now in an infrastructure arms race. The article makes clear that Anthropic is leaning on huge external compute deals to keep pace, and that tells me the economics are still rough underneath the polish. I’d be curious whether these higher limits hold up under sustained demand or whether users eventually run into a new ceiling once the next wave of adoption arrives.

The SpaceX angle is the weirdest and most interesting part. Elon Musk’s public shift from hostile critic to impressed partner is very on-brand for the current AI era: alliances are fluid, and capacity is king. The orbital compute talk feels more speculative to me, maybe even a little overhyped, but it does point to a real concern: terrestrial compute, power, land, and cooling may become strategic constraints before the software side does.

If I were using Claude Code heavily, I’d be happy about this and watch closely for how the limits behave in practice. Better access is great, but what developers really want is predictable access.

In short: Anthropic is trying to turn scarce compute into a better developer experience. That’s the right move, and probably the only move if Claude Code is going to keep growing.