← Back to News

xAI Is No Longer a Standalone Company: It Is Now SpaceXAI - Elon Musk Folds His AI Empire Into SpaceX Ahead of a 2 Trillion Dollar IPO

Elon Musk in 2025, the founder of SpaceX and xAI, who announced today that xAI will be dissolved and fully integrated into SpaceX as SpaceXAI

Big news from Cupertino - actually, no. From Hawthorne, Memphis, and orbit.

Elon Musk announced today that xAI will be dissolved as a standalone company and fully integrated into SpaceX. The combined entity's AI products will operate under a new brand: SpaceXAI. Grok, the Colossus supercomputer in Memphis, and the entire xAI team are now part of SpaceX.

This is one of the largest corporate reorganizations in tech history. It comes weeks before SpaceX is expected to file for a 2 trillion dollar IPO. And it includes a strategic detail that has not gotten enough attention: Anthropic - typically positioned as one of Musk's chief AI rivals - just signed a deal to lease 300+ megawatts of compute capacity at SpaceX's Colossus 1 data center.

Below: the full breakdown.

Elon Musk in 2025, the founder and CEO of SpaceX and xAI, who today announced the merger of his two largest tech companies into a single entity

Elon Musk, 2025. Today he announced one of the most ambitious corporate consolidations in tech history. Photo: Wikimedia Commons, CC BY-SA.

1. What Was Announced

ElementDetail
xAI as a standalone companyBeing dissolved
New brandSpaceXAI - the AI division of SpaceX
Products transferringGrok chatbot, Colossus supercluster, all xAI IP and talent
BackdropBuilds on the SpaceX-xAI integration announced February 2026
Stated goalBuild orbital data centers in space, vertically integrate AI + space
TimingComes ahead of a planned mid-2026 SpaceX IPO at 1.75 - 2 trillion dollar valuation

2. The Colossus Supercomputer

Colossus is the centerpiece of what xAI - now SpaceXAI - is bringing to the merger.

Colossus SpecsDetail
LocationMemphis, Tennessee (former Electrolux factory)
Build time122 days for the initial 100,000-GPU cluster
Current GPUs~200,000 NVIDIA H100/H200 (with stated goal of 1 million)
Power draw~150 megawatts at full capacity
Cooling waterMillions of gallons daily
Primary purposeTraining Grok and serving X / SpaceX compute needs
StatusBelieved to be the largest AI supercomputer on Earth

Why it matters: Building Colossus in 122 days made xAI a credible compute player almost overnight. Now that compute - and the team that built it - belongs to SpaceX. Combine SpaceX's launch capacity, Colossus's compute capacity, and Grok's model stack, and you have the only fully vertically integrated AI infrastructure stack on the planet.

3. The Orbital Data Center Thesis

SpaceX Falcon Heavy rocket launch from Kennedy Space Center, the heavy-lift vehicle that paved the way for orbital infrastructure deployment

SpaceX Falcon Heavy launch. The vehicle that proved heavy-lift to orbit at radically reduced cost - the foundation of the orbital data center thesis. Photo: Wikimedia Commons, Public Domain.

The strategic case for orbital compute rests on three problems Earth-based AI is hitting hard:

Problem on EarthSolution in Space
Power supplyGigawatt-class clusters require small-city electricity. Grids cannot deliver fast.
Solar power in spaceUnlimited, 24/7 in geostationary orbit (no day/night cycle), no weather attenuation
Cooling on EarthMillions of gallons of water per day; air cooling no longer scales
Cooling in spaceRadiative cooling to vacuum is essentially free. The bottleneck is radiator surface area, not water supply.
Real-estate constraint on EarthMemphis, Phoenix, Northern Virginia - all approaching grid limits
In spaceEffectively unlimited orbital volume

The bottleneck for orbital compute has always been launch cost. SpaceX is the company that has driven launch cost down by approximately 10x over the past decade. Falcon Heavy made heavy-lift commercially viable. Starship - now flying repeatedly - is targeting another 10x reduction.

SpaceX Starship rocket ignition during the IFT-5 (Integrated Flight Test 5) launch, the heavy-lift vehicle that will deploy orbital data centers

SpaceX Starship ignition during IFT-5. Starship's payload-to-orbit cost is the linchpin of the orbital data center thesis. Photo: Wikimedia Commons, CC BY-SA.

4. The Anthropic Deal

The other announcement this week is in some ways more surprising than the merger itself.

Anthropic - the maker of Claude, founded in part by former OpenAI executives, and frequently positioned in the press as Elon Musk's AI competitor - signed a deal with SpaceX to use over 300 megawatts of compute at the Colossus 1 data center in Memphis.

Anthropic also publicly expressed interest in working with SpaceX to develop multi-gigawatt orbital compute capacity.

This is significant for several reasons:

  1. Anthropic is choosing to depend on Musk infrastructure for a substantial chunk of Claude's training and inference compute
  2. Musk personally has had a contentious public history with Anthropic - the deal happens regardless
  3. Compute supply is the new oil. Whoever has the largest, fastest-deployable capacity wins. SpaceXAI is now positioned as the dominant compute infrastructure provider
  4. If Anthropic is signing now, every other frontier lab is looking at the same supply curve

5. Starship and the Orbital Compute Pipeline

SpaceX Starship vehicle models on display at the SpaceX Design Center showing the heavy-lift architecture for orbital deployment of large payloads

Starship models at the SpaceX Design Center. The architecture that makes orbital data centers economically conceivable. Photo: Wikimedia Commons, CC BY-SA.

For orbital data centers to be economically viable, you need to launch tens of tonnes of GPU + radiator + solar hardware to orbit at radically lower cost than current launch vehicles allow. Starship's stated goal is to lift over 100 tonnes to LEO at a per-launch cost target that, fully amortized, brings cost-per-kg-to-orbit into the low hundreds of dollars range.

SpaceX Starship nosecone interior view showing the cargo bay where compute hardware would be deployed for orbital data centers

Starship nosecone interior - the cargo bay that will lift orbital compute hardware. Photo: Wikimedia Commons, CC BY-SA.

The honest reality check: Orbital data centers at scale are still 5-10 years away. The first deployments will be small (sub-megawatt). Radiator design at scale is unsolved. On-orbit servicing of GPUs is unsolved. Bandwidth back to Earth is a real constraint. The vision is real and SpaceX is uniquely positioned, but the timeline is decade-scale, not quarters.

6. The IPO Math

SpaceX is reportedly targeting a 2026 IPO with a valuation in the 1.75 to 2 trillion dollar range. That would make it one of the largest IPOs in history - and possibly the largest US IPO ever.

SpaceX IPO ComponentsDetail
Launch businessFalcon 9, Falcon Heavy, Starship
StarlinkLargest satellite constellation, 8,000+ satellites in orbit
Now: SpaceXAIGrok, Colossus, orbital compute thesis
Government / DoD contractsNational security launch, Starshield
Target valuation1.75 - 2 trillion dollars

The reason SpaceXAI matters for the IPO is not that AI is a hot tag. It is that orbital compute is a category investors have never been asked to price before. By folding xAI in pre-IPO, Musk gives prospective public-market investors a single asset that is simultaneously a launch monopoly, a satellite operator, and the only credible orbital AI infrastructure play. The investment thesis becomes vertically integrated in a way no other company can match.

7. What Could Go Wrong

  • Regulatory friction. Concentrating launch + AI + satellite + government contracts inside one entity invites antitrust scrutiny. The FCC, FAA, and DOJ all have legitimate questions.
  • Existing xAI investors. The mechanics of converting xAI shares into SpaceX shares are non-trivial. Some investors will resist.
  • Anthropic's choice. Anthropic is now compute-dependent on a Musk entity. The optics will be debated.
  • Orbital compute timeline. If the first orbital data center demonstration slips past 2030, the IPO thesis loses its most differentiated narrative.
  • The Grok product. Grok has been a compute and data play more than a frontier model lead. Whether SpaceXAI can ship a top-3 frontier model is still unproven.

8. The Big Picture

For two decades, Musk has run his companies as separate but adjacent bets - Tesla for energy and transport, SpaceX for launch, xAI for AI, Neuralink for brain interface, X for distribution. Today's announcement is the first time he has explicitly merged two of his largest companies. The integration says something specific about how he sees the next decade: AI compute and space launch are not adjacent businesses. They are the same business.

9. The Timeline at a Glance

DateEvent
2002SpaceX founded
March 2023xAI founded
November 2023Grok 1 launched on X
July 2024Colossus 1 operational (100K GPUs in 122 days)
February 2026SpaceX-xAI integration first announced
May 2026xAI dissolved. SpaceXAI established. Anthropic Colossus 1 deal announced.
Mid-2026 (target)SpaceX IPO at 1.75 - 2 trillion dollars

Sources

Source: Republic World, Stocktwits, CNBC ↗