Anthropic and SpaceX formalized a major computing agreement on May 6, 2026. which gives the company behind Claude immediate access to over 300 megawatts of processing power. A surprising alliance, with implications that go far beyond a simple infrastructure upgrade.
A computing shortage that forced their hand
For several months, Anthropic is hitting its own limits. The growth in Claude's usage has consistently exceeded available capacity, leading to repeated outages and access restrictions during peak hours. The company had already signed large-scale agreements to address this: up to 5 gigawatts with Amazon, another 5 gigawatts with Google and Broadcom, 30 billion dollars in Azure capacity through a partnership with Microsoft and NVIDIA, and 50 billion dollars invested in US infrastructure with Fluidstack. The agreement with SpaceX is part of this same urgent logic, but with an unprecedented dimension.
Colossus 1: 220,000 NVIDIA GPUs serving Claude
The agreement concerns access to the full computing capacity of Colossus 1, SpaceX's data center located near Memphis, Tennessee. According to Anthropic's official press release published on May 6, 2026, this partnership provides over 300 megawatts of new capacity, or more than 220,000 NVIDIA processors, available by the end of the month. This additional capacity is intended to directly improve the service for Claude Pro and Claude Max subscribers.
Colossus 1 was inaugurated in 2024 to power the Grok models of xAI, Elon Musk's AI startup, which merged with SpaceX in February 2026. The infrastructure, oversized for Grok's current needs, thus finds a profitable use thanks to Anthropic.
Usage limits raised starting today
The agreement is producing concrete and immediate effects for Claude users. Anthropic announced three changes effective May 6, 2026:
The doubling of the five-hour rate limits for Claude Code, applicable to Pro, Max, Team, and Enterprise plans.
The removal of peak hour limit reductions for Claude Code on Pro and Max accounts.
A significant increase in API rate limits for the Claude Opus models, detailed in the updated table on Anthropic's developer platform.
These adjustments directly address the frustrations of the most intensive users, particularly developers and enterprise teams who regularly encountered ceilings.
The next frontier: Earth orbit
Beyond the immediate aspect, the agreement includes a clause for joint exploration in a yet non-existent field: orbital computing. Anthropic has indicated its interest in partnering with SpaceX to develop several gigawatts of AI computing capacity in Earth orbit. The idea is to overcome terrestrial constraints related to land, energy, and local regulations by deploying data centers in space.
This ambition remains at the exploration stage for now, with no officially communicated timeline or budget. But it illustrates the direction of AI infrastructure as the demand for computing becomes structurally greater than the supply available on Earth.
Why Musk said yes to Anthropic
The political dimension of this agreement deserves attention. Elon Musk and Anthropic do not operate in the same circles: the company founded in 2021 by Dario Amodei has had tense relations with the Trump administration, which at one point banned the use of Claude in US government agencies. Musk, for his part, is an avowed supporter of this administration.
In a post on X dated May 6, Musk explained that he had spent time with Anthropic's management to understand the company's approach to AI safety. He concluded that, as long as the team continued to be self-critical, Claude would have a good chance of contributing positively to humanity. An unexpected endorsement that seems to have facilitated the conclusion of the agreement.
SpaceX, a new essential player in AI infrastructure
This agreement is part of a broader strategy by SpaceX, which is now positioning itself as an infrastructure provider for the AI competing with xAI. The company had already signed a similar agreement with Cursor, a startup specializing in AI-assisted code. In parallel, xAI is expected to soon disappear as an independent entity, with its products being consolidated under the SpaceXAI banner.
This consolidation responds to a clear financial logic. Colossus 1 is a massive infrastructure with fixed costs, while Grok models struggle to establish themselves against Claude, ChatGPT, or Gemini. Renting this capacity to other AI players makes it possible to monetize the investment, improve SpaceX's financial profile before an IPO announced as imminent, and position the company as a strategic partner indispensable to the entire ecosystem.
Anthropic also confirms its commitment to expanding internationally, with some of its new computing agreements including capabilities in Asia and Europe to meet data residency requirements in regulated sectors. With the SpaceX agreement, it's a new step in building a global infrastructure for Claude, and the limits raised as of May 7, 2026, are its first concrete translation for users.
No comments yet — start the discussion!