Anthropic and SpaceX Eye Data Centers in Orbit in Fresh Alliance Against OpenAI

It is an alliance that not everyone anticipated, but which is ultimately quite logical: The AI subsidiary of SpaceX, xAI, and AI company Anthropic have signed a far-reaching cooperation agreement. At the center is access to Colossus 1, one of the world’s most powerful AI supercomputers. The deal comes at a time when both companies are eyeing an IPO this year and competition in the AI sector is intensifying massively.

Both Anthropic, as a spinoff from OpenAI, and SpaceX/xAI with Elon Musk at the helm are in fierce competition with OpenAI — an alliance between the two therefore makes considerable sense. Currently, Musk and OpenAI co-founders such as Sam Altman and Greg Brockman are in court, disputing the origins of OpenAI.

The Deal at a Glance

As part of the agreement, Anthropic receives access to the full computing capacity of SpaceX’s Colossus 1 data center. The facility has more than 220,000 NVIDIA GPUs, including models of the H100, H200, and next-generation GB200 types. The associated power of over 300 megawatts is expected to be available within a month, according to Anthropic, and will directly benefit subscribers of the Claude Pro and Claude Max services.

Colossus 1 was built in record time and is considered one of the fastest infrastructure deployments in the history of the AI sector. The data center supports use cases ranging from language modeling and multimodal systems to scientific simulations at frontier level.

Anthropic’s Investor Network: Google, Amazon, Microsoft, and Nvidia

The SpaceX deal is part of a series of strategic capital partnerships that Anthropic has concluded in recent months. The company has thereby built one of the broadest infrastructure bases among independent AI providers.

• Amazon: Agreement for up to 5 gigawatts of computing capacity, with nearly 1 gigawatt by end of 2026, including inference capacities in Asia and Europe
• Google and Broadcom: Another 5-gigawatt agreement, to be gradually brought online from 2027
• Microsoft and Nvidia: Strategic partnership with $30 billion in Azure capacity
• Fluidstack: $50 billion investment in American AI infrastructure

Anthropic operates its models on a heterogeneous hardware base, including AWS Trainium, Google TPUs, and NVIDIA GPUs. This diversification is intended to reduce dependencies on individual providers and ensure scalability.

AI Data Centers in Orbit

Particularly noteworthy is a project that goes far beyond conventional data center infrastructure: Anthropic and SpaceX have expressed their interest in jointly developing several gigawatts of orbital AI computing capacity. No concrete contracts have been signed yet, but the letter of intent is part of the current agreement.

Both companies cite the physical limits of terrestrial infrastructure as justification: power, space, and cooling are increasingly becoming bottlenecks. SpaceX sees itself as the only organization with the necessary launch frequency, the mass costs to Earth orbit, and the experience in operating satellite constellations to turn orbital computing capacity from a research concept into a concrete engineering program. Solar energy in space is considered a virtually unlimited and sustainable energy source without burdening the Earth.

That building AI data centers in orbit is of course not without its problems has been widely reported. Meta, for example, another competitor in the AI field, is taking a different approach and wants to “beam” solar energy from space down to Earth.

IPOs 2026: Pressure and Ambitions

Both SpaceX/xAI and Anthropic are considered hot candidates for an IPO in 2026. For both companies, building a credible and scalable infrastructure is a key signal to potential investors. The partnerships and capacity commitments now communicated are likely also strategically placed with upcoming roadshows in mind.

The time pressure is real: In the AI infrastructure race, competitors such as Meta and OpenAI are also investing massively in their own data center capacities. Meta has announced multi-billion-dollar investments in data centers, while OpenAI is working on its own infrastructure offensive in the US as part of the Stargate project together with SoftBank and Oracle.

International Expansion and Social Responsibility

Anthropic plans to roll out some of the new capacities internationally. Corporate customers from regulated industries such as financial services, healthcare, and the public sector increasingly require regional infrastructure to meet compliance and data protection requirements. When selecting locations, Anthropic says it prioritizes democratic countries with stable legal and regulatory frameworks and secure supply chains.

Furthermore, Anthropic has committed to compensating for any electricity price increases for private households near its US data centers. This model is to be extended to international locations in the future.