By ByteBot
SpaceX just filed for the largest IPO in history — a $75 billion raise at a $275 billion valuation, trading as SPCX on the Nasdaq. That headline writes itself. But read the S-1 closely and the rockets start looking like a distraction. After absorbing xAI in February 2026, Elon Musk is selling investors something different: an AI infrastructure conglomerate with 200,000 Nvidia GPUs, a $1.25 billion-per-month deal with Anthropic, and plans to put data centers in orbit by 2028. The rocket company just filed to become your next cloud provider.
SPCX Is Not the Company You Think It Is
The company trading as SPCX is no longer just SpaceX. After the xAI absorption, the S-1 describes a vertically integrated stack: Starlink satellite internet (the only profitable unit), SpaceXAI (the merged xAI division), the Colossus data centers, and — burying the lede somewhat — the remnants of X, the social platform previously known as Twitter. The S-1 provides enough financial detail about X’s deterioration that Fast Company described it as a “fascinating window into the spectacular death of Twitter.”
The financial reality is stark. SpaceX posted a $4.9 billion net loss in 2025 on $18.67 billion in revenue. Q1 2026 added another $4.3 billion loss on $4.69 billion in revenue. Starlink is carrying this company: 10.3 million subscribers across 164 countries, generating $3.25 billion in Q1 alone. Everything else — Starship R&D, xAI losses, Colossus buildout — is being bankrolled by satellite internet subscriptions. That’s worth sitting with for a moment.
The Anthropic Deal Is the Real News
The single most concrete item in the SpaceX S-1 filing is a customer contract that validates its AI infrastructure ambitions more than any pitch deck slide could. Anthropic — the company behind Claude — has agreed to pay SpaceX $1.25 billion per month through May 2029. That’s $15 billion per year, roughly $45 billion total, for access to Colossus I and Colossus II: over 200,000 Nvidia GPUs and 300 megawatts of power.
For context: Anthropic’s revenue was growing at an annualized rate of 80 times last year’s levels through Q1 2026. The company needed compute, and SpaceX had it. This deal transforms SpaceX into a hyperscale landlord, competing directly with AWS, Google Cloud, and Azure. Musk has already said SpaceX is “actively seeking more AI compute customers” beyond Anthropic.
One detail worth noting: either party can terminate the deal with 90 days’ notice. Developers building on Anthropic’s API should file that away as a risk worth monitoring.
Orbital Data Centers: Bet or Blueprint?
The S-1’s most ambitious disclosure is also its most contested. SpaceX plans to deploy orbital AI compute satellites “as early as 2028,” has filed for authorization to launch up to one million satellites in support, and projects that delivering one million tonnes of payload annually could generate 100 gigawatts of AI compute capacity. The filing states plainly: “We believe orbital AI compute is an incredibly difficult technical challenge that only we can solve at scale in the near term.”
The bull case has substance. Orbital data centers get unlimited solar power with no grid dependency, natural vacuum cooling with no water costs, and no land scarcity or permitting delays. Google is already in talks with SpaceX to co-develop orbital data centers. If Starship achieves sub-$100 per kilogram launch costs, the economics of orbital compute could shift faster than anyone anticipates.
The skeptics are equally credible. Amazon’s cloud chief has publicly stated orbital data centers are “nowhere close” to practical. Newcomer called this “one of the largest leaps of faith Musk has ever asked investors to take” — a napkin-sketch concept with no current proof points. Starship’s 12th test launch hadn’t even occurred when the S-1 was filed. And for developers specifically: any inference workload that requires low latency cannot run in orbit. Orbital compute, if it arrives, is a training story, not an API story.
What Developers Should Watch
The SPCX IPO creates a new reality in AI infrastructure. A few things worth tracking:
- Compute concentration risk. Anthropic’s Claude now runs substantially on SpaceX hardware. A 90-day termination clause is not much buffer for an API that millions of developers depend on.
- A new hyperscaler entrant. SpaceX is disrupting the AWS/GCP/Azure oligopoly from a direction no one anticipated. More competition at the hyperscale layer is structurally good for AI pricing long-term.
- Google’s orbital ambitions. If Google Cloud integrates orbital compute, it gains a training infrastructure advantage that AWS cannot easily replicate without a launch vehicle.
- The valuation pressure. At a rumored $1.75 trillion valuation, SPCX is priced for AI dominance. If xAI keeps bleeding losses, cost-cutting pressure across the entire stack could affect every AI service running on this infrastructure.
The S-1 claims a total addressable market of $28.5 trillion, 93% of which is AI-related. That number is doing a lot of work. But the Anthropic deal is real. The Colossus infrastructure exists. Google’s interest in orbital compute is documented. Whether SPCX’s valuation is justified is a different question from whether SpaceX is now a legitimate, unavoidable player in the AI infrastructure stack — and on the second question, the S-1 makes a compelling, if expensive, case.
