SpaceX Swallows xAI to Anchor $1.75 Trillion Valuation

On May 6, Elon Musk executed a structural pivot that redefined the trajectory of his artificial intelligence and aerospace empires in a single stroke. The announcement was two-pronged and characteristically disruptive: xAI, the independent AI lab founded to challenge OpenAI, will no longer exist as a standalone company. Instead, it has been absorbed into SpaceX and rebranded as SpaceXAI. Simultaneously, SpaceX has entered into a massive infrastructure agreement to lease its Colossus 1 data center—a facility housing 220,000 Nvidia GPUs and drawing 300MW of power—to Anthropic, a competitor Musk had previously described in disparaging terms.

This maneuver comes exactly 33 days before the scheduled start of the SpaceX IPO roadshow on June 8. To the casual observer, leasing one of the world’s most powerful AI training clusters to a rival like Anthropic while dissolving your own AI company might look like a retreat. However, for those of us tracking the mechanical and economic integration of these industries, it is a masterclass in balance-sheet engineering. By merging xAI into SpaceX, Musk is not killing the technology; he is changing the narrative of the underlying assets. He is moving from the volatile, high-burn world of AI model development into the more stable, infrastructure-heavy domain of orbital compute and industrial automation.

The Industrial Scale of Colossus 1

To understand the gravity of the Anthropic lease, one must look at the hardware specs. Colossus 1 was built in an unprecedented 122 days, a feat of mechanical and electrical engineering that many in the industry thought impossible. At its core are 220,000 Nvidia GPUs, likely a mix of H100 and H200 architectures, supported by a 300MW power envelope. In the world of high-performance computing (HPC), these are not just chips; they are industrial-grade heat generators that require massive cooling infrastructure and specialized power distribution units. This facility was the physical proof-of-concept for xAI’s capability to iterate faster than legacy labs.

The decision to lease this entire cluster to Anthropic within the month suggests that SpaceX has already completed its next stage of infrastructure development. Musk confirmed that SpaceXAI has migrated its primary training workloads to Colossus 2, a successor facility designed for even greater density. From a pragmatic engineering perspective, an idle data center is a depreciating liability. By leasing Colossus 1 to Anthropic, SpaceX converts a capital-intensive asset into a massive recurring revenue stream. This provides immediate liquidity and a reinforced bottom line just as the company prepares to pitch its $1.75 trillion valuation to global investors. Anthropic, which is locked in an arms race for compute, gains immediate access to a turnkey cluster that would take years to build from scratch elsewhere.

Why the $1.75 Trillion Target Demands a Merger

The Engineering Logic of Orbital Compute

The most technically intriguing aspect of this merger is the “Orbital AI Compute” thesis. SpaceX’s official communications have noted that the compute requirements for the next generation of AI now exceed the terrestrial availability of land, electricity, and cooling within necessary timeframes. In a vacuum, cooling is a significant challenge because convection is not an option; everything must be handled through radiation. However, space offers an infinite heat sink and nearly 24-hour solar energy if positioned correctly. If SpaceX can solve the thermal management and radiation hardening of GPU clusters in orbit, they bypass the bottleneck of terrestrial power grids.

Integrating Grok and the xAI software stack directly into the Starlink network turns every satellite into a potential node in a distributed global brain. This reduces latency for AI edge applications and provides a secure, sovereign compute layer that is not reliant on any single nation's infrastructure. While Anthropic’s interest in “multi-GW level orbital AI compute” may be largely rhetorical for now, it serves as a powerful endorsement for the SpaceX IPO roadshow. It signals to investors that the world’s leading AI labs are already looking at SpaceX not just as a transportation company, but as the future landlord of the digital world.

The Death of the Public Benefit Narrative

The dissolution of xAI also marks the end of its identity as a Public Benefit Corporation (PBC). When xAI was founded in 2023, it carried a mandate to prioritize societal benefit over shareholder profit. That status was quietly dropped in May 2024, and the merger into SpaceX completes the transition into a purely commercial vehicle. For investors, this is a positive development. It removes the legal ambiguities of a PBC and aligns the AI development team with the rigorous, result-oriented culture of SpaceX’s engineering divisions. The goal is no longer just to build a “truth-seeking” AI; it is to build an AI that can manage orbital logistics, navigate Starships to Mars, and monetize the massive datasets generated by the Starlink constellation.

Musk’s reversal on Anthropic, moving from calling them “evil” to stating they don’t set off his “evil detector,” is a classic pragmatic pivot. In the lead-up to an IPO of this magnitude, ideological purity takes a backseat to strategic alliances. If Anthropic is willing to pay top dollar to lease Colossus 1, they are no longer an enemy; they are a high-value tenant. This deal essentially allows SpaceX to subsidize its own R&D on Colossus 2 with competitor capital. It is a ruthless, highly efficient use of hardware that reflects the engineering ethos of maximizing utility while minimizing waste.

What Happens to Grok?

The Grok product will likely continue as the consumer-facing interface of SpaceXAI, but its role in the company's valuation has shifted. It is no longer the primary product; it is the data-gathering front-end for the broader ecosystem. The real value lies in the integration of AI with the physical world. This includes the automation of Starship manufacturing, the optimization of satellite beam-forming, and eventually, the software that will govern autonomous systems on other planets. By folding the xAI team into the existing SpaceX structure, Musk is forcing a merger of digital intelligence and mechanical hardware.

As the SEC prospectus (S-1) becomes public in the coming weeks, we expect to see a heavy emphasis on these synergies. The market will be asked to value a company that has successfully vertically integrated everything from the silicon to the rocket fuel. The dissolution of xAI isn’t an admission of failure in the AI space; it is an admission that the AI space, on its own, is not big enough for Musk’s ambitions. To reach $1.75 trillion, he needs the AI to be the nervous system of a global, and eventually interplanetary, industrial machine.