Anthropic has secured a multibillion-dollar agreement to utilize the full compute capacity of SpaceX’s Colossus supercomputer. The deal, announced Wednesday, not only provides immediate infrastructure relief for Anthropic’s Claude AI platform but also signals a radical shift toward moving artificial intelligence (AI) data centers into Earth’s orbit.
The partnership grants Anthropic exclusive access to the Colossus 1 facility in Memphis, Tenn., a powerhouse boasting more than 220,000 NVIDIA Corp. graphics cards and 300 megawatts of capacity.
The impact was instantaneous: Anthropic confirmed that rate limits for Claude Pro and Claude Max subscribers doubled as of Wednesday, addressing recent performance bottlenecks and infrastructure strain that had plagued the platform during peak hours.
Beyond terrestrial hardware, the agreement outlines a futuristic roadmap for “multiple gigawatts of orbital AI compute capacity.” SpaceX, led by Elon Musk, argued that the energy demands of next-generation AI are “outpacing what terrestrial power, land, and cooling can deliver.” By moving data centers into space, the companies aim to leverage near-limitless solar power and natural vacuum cooling, theoretically reducing the environmental footprint on Earth.
This move comes as the Memphis facility faces local scrutiny. The rapid 122-day construction of Colossus involved the installation of dozens of natural gas turbines, sparking protests over air quality and the massive water consumption required to cool the high-end accelerators.
The deal marks a dramatic detente between Musk and Anthropic. Following the merger of SpaceX and xAI into SpaceXAI earlier this year, Musk had been a vocal critic of Anthropic, labeling the firm — founded by OpenAI defectors — as “misanthropic” and a risk to Western civilization. However, after a week of meetings with Anthropic’s leadership, Musk reversed his stance on X (formerly Twitter), praising the team’s competence and stating they did not trigger his “evil detector.”
“I see this as xAI and SpaceX realizing that they have spare capacity that they can monetize,” Steven Dickens, CEO of HyperFRAME Research, said in an email. “Musk has proven he can build infrastructure at scale for AI rapidly, and crucially secure power. The frontier labs are craving this infrastructure as inference scales. The two are coming together. This is a great signal for hyperscaler adoption so I don’t see it as a zero-sum game, in fact the exact opposite.”
For Anthropic, the SpaceX partnership is a critical diversification of its supply chain. While the company maintains massive compute agreements with Google and Amazon.com Inc., it has faced significant headwinds with the U.S. government. In March, the Pentagon blacklisted Anthropic as a supply chain risk, leading to ongoing litigation between the startup and the Trump administration.
Conversely, Musk’s SpaceXAI has enjoyed deeper integration with the Defense Department, which recently embraced xAI’s Grok model.
As Anthropic currently negotiates a fresh funding round at a staggering $900 billion valuation, this alliance provides the raw horsepower necessary to compete with OpenAI. By tethering its future to SpaceX’s orbital ambitions, Anthropic is betting that the next frontier of intelligence isn’t just silicon. It’s satellite based.
“By securing over 300 megawatts of capacity and 220K NVIDIA GPUs, Anthropic transitions to a diversified, multi-cloud infrastructure strategy that reduces dependency on hyperscalers such as AWS and Google,” HyperFRAME Research analyst Ron Westfall said. “The partnership provides a roadmap for orbital DCs (data centers), a shift I anticipate can bypass the power and cooling constraints on Earth by deploying multiple gigawatts of computer capacity directly into space.”
