Anthropic Secures SpaceX’s Colossus 1 Infrastructure, Doubles Claude Code Limits
Anthropic announced that it has secured full access to SpaceX's Colossus 1 data centre in Memphis, and is using the infrastructure surge to immediately double Claude Code rate limits and lift peak-hours restrictions for paid subscribers.
There are business partnerships that make sense on paper. Then there are ones that would have seemed flatly implausible a year ago. Anthropic signing a compute deal with Elon Musk’s SpaceX, whose owner has publicly called Anthropic “misanthropic and evil” and whose merged xAI subsidiary is a direct rival, falls squarely in the second category.
Yet on May 6, 2026, both companies confirmed exactly that arrangement. Anthropic stated that the company would use the full computing capacity of Colossus 1, the Memphis data centre originally built for xAI. The deal adds more than 300 megawatts of capacity and over 220,000 NVIDIA GPUs, expected online within the month.
As a direct result, Claude Code rate limits are doubling across paid tiers effective immediately.
What Colossus 1 Actually Adds to Anthropic’s Infrastructure
Colossus 1 is not a modest facility. As the xAI official announcement notes, the Memphis data centre houses dense deployments of NVIDIA H100, H200, and GB200 accelerators, the same hardware powering some of the industry’s most advanced AI training and inference workloads.
SpaceX claims to have built it in 122 days for xAI’s own model development, using gas-burning turbines to supply power during construction.
As WIRED reports, local residents and activists have raised concerns over the turbines and their impact on air quality in surrounding Memphis neighbourhoods, an issue Anthropic has not publicly addressed.
The 300 megawatts Anthropic gains from Colossus 1 adds to a rapidly expanding compute stack. According to Anthropic, the SpaceX deal joins:
- An up to 5 gigawatt agreement with Amazon, including nearly 1 gigawatt arriving before the end of 2026
- A separate 5 gigawatt agreement with Google and Broadcom, expected online in 2027
- A strategic partnership with Microsoft and NVIDIA covering $30 billion in Azure capacity
- A $50 billion infrastructure investment with Fluidstack
The scale is notable for a company eyeing a $900 billion valuation that, as recently as February, faced compute shortages that visibly throttled subscriber usage.
The Rate Limit Changes That Subscribers Will Actually Notice
Compute deals are invisible to most users until they translate into something tangible. This one does, immediately.
As Anthropic states, Claude Code’s five-hour rate limits are now doubled across Pro, Max, Team, and seat-based Enterprise plans, the upgrade developers running long agentic coding sessions had requested most persistently.
Peak-hours limit reductions, the throttling that kicked in during high-demand windows for Pro and Max subscribers, are removed entirely. API rate limits for Claude Opus models are raised considerably, benefiting enterprise teams running high-volume inference workloads.
For developers timing Claude Code sessions around off-peak hours to avoid usage ceilings, those workarounds are no longer necessary.
The Orbital Compute Ambition Changing the Longer-Term Picture
The most forward-looking part of the announcement received the least attention: Anthropic has formally expressed interest in partnering with SpaceX to develop multiple gigawatts of orbital AI compute capacity.
As xAI mentions, frames space based compute as a way to bypass the two core limits of terrestrial data centres: land availability and cooling constraints. In orbit, neither applies in the same way.
This shift is driven by Terafab, a manufacturing powerhouse slated to dedicate 80% of its total compute output to these orbital data centres.
SpaceX’s recently filed FCC paperwork outlines plans for up to one million satellites as the basis of an orbital data centre network.
Whether Anthropic becomes a formal infrastructure partner rather than just a customer of Colossus 1 remains under early negotiation, but both companies’ stated interest signals that space-based frontier AI compute is now being actively explored.
As Wired noted, that ambition alone, from a company that spent the last three months fighting a US government blacklisting, represents a remarkable acceleration of Anthropic’s infrastructure posture in a very short window of time.
Source: Higher usage limits for Claude and a compute deal with SpaceX
