On April 1, SpaceX filed a confidential draft registration statement with the Securities and Exchange Commission. Bloomberg broke the story. Reuters, CNBC, and the Wall Street Journal confirmed it independently within hours. The numbers were staggering even by Musk standards: a target raise of $75 billion at a valuation of 1.75 trillion dollars, with a Nasdaq listing expected by late summer 2026.
Goldman Sachs, Morgan Stanley, and JPMorgan Chase are leading the deal. If it closes at the stated valuation, SpaceX will become the most valuable company to ever go public, eclipsing Saudi Aramco's 2019 record of $29.4 billion raised. By a factor of 2.5.
But the filing wasn't really about SpaceX the rocket company. It was about what SpaceX became two months earlier.
The xAI Merger Changed Everything
In February 2026, SpaceX acquired xAI, Musk's artificial intelligence venture. The deal valued xAI at roughly $250 billion and created a combined entity worth approximately 1.25 trillion dollars. A corporate restructuring nicknamed "K2" integrated xAI's Grok large language models directly into SpaceX's operational architecture.
The strategic logic was radical. Instead of running AI workloads in data centers on the ground, SpaceX would push compute into orbit. The Starlink satellite network, already the world's largest commercial constellation, would become the backbone of a distributed computing platform. xAI's models would run on GPU clusters attached to satellites, powered by constant solar energy and cooled by the vacuum of space.
Musk had been talking about orbital data centers for months. At the World Economic Forum in Davos in January, he made the pitch in characteristically blunt terms:
"The lowest-cost place to put AI will be in space, and that will be true within two years, maybe three at the latest. You're power constrained on Earth. Space has the advantage that it's always sunny." — Elon Musk, CEO of SpaceX (World Economic Forum, Davos, January 2026)
Terrestrial data centers consume massive amounts of electricity and water. They require land, permits, and grid connections that take years to secure. Space eliminates all three constraints.
The IPO filing signaled that Musk was done talking. He was asking the public markets to fund it.
The Scale of the Bet
| Detail | Value |
|---|---|
| SEC Filing Date | April 1, 2026 |
| Target Raise | $75 billion |
| Valuation | $1.75 trillion |
| Exchange | Nasdaq |
| Expected Listing | Late summer 2026 |
| Lead Underwriters | Goldman Sachs, Morgan Stanley, JPMorgan Chase |
| Retail Investor Allocation | 30% of shares |
| Previous IPO Record | Saudi Aramco, $29.4 billion (2019) |
The capital would fund three priorities. First, Starship commercialization at industrial scale. Second, uncrewed Mars missions targeting a 2028 launch window. Third, and most important to institutional investors, the construction of orbital AI data centers.
SpaceX plans to launch up to 1 million data-center satellites into low Earth orbit. Each satellite would carry GPU clusters networked together through the existing Starlink mesh. The pitch to Wall Street: AI training and inference workloads would migrate from power-hungry ground facilities to solar-powered orbital nodes, with Starlink's 10+ million subscribers providing a built-in customer base.
SpaceX generated roughly $15 billion in total revenue during 2025, with Starlink as the primary driver. At over 10 million subscribers paying an average of roughly 90 dollars per month, the satellite internet business is the engine beneath everything else.
That revenue stream gives the IPO something most speculative infrastructure bets lack: a profitable existing business underneath the moonshot.
Eight Launches a Day, Every Day
The orbital data center vision demands a launch cadence that doesn't exist yet.
To deploy and maintain a constellation of 1 million satellites, SpaceX would need approximately 3,000 Starship launches per year. That works out to eight launches every single day. Current Starship flight rates are nowhere close. The vehicle has flown a handful of test missions. Scaling from test flights to daily operations requires manufacturing breakthroughs, launch pad expansion, and FAA regulatory approvals that could take years.
When NVIDIA unveiled its AI infrastructure roadmap at GTC, the company acknowledged that compute demand was outstripping the planet's ability to build data centers. Musk is betting that the answer is to leave the planet entirely. NVIDIA's Jensen Huang hasn't commented on the SpaceX filing, but the GPU maker's chips would almost certainly power whatever SpaceX puts in orbit.
Deutsche Bank's infrastructure analysts estimated that orbital data centers won't reach cost parity with terrestrial facilities until "well into the 2030s." The analysis assumed aggressive reductions in launch costs, from today's low thousands of dollars per kilogram to the low hundreds. That's the kind of cost curve that took Falcon 9 a decade to achieve.
Bezos Is Already Building the Rival
SpaceX isn't the only company chasing orbital compute.
Jeff Bezos' Blue Origin launched a competing initiative called Project Sunrise, planning to deploy 51,600 satellites equipped with compute capacity. The ambition is similar. The funding is not. Blue Origin remains private, operating within the constraints of Bezos' personal capital and whatever Amazon's board is willing to allocate.
SpaceX's IPO changes the competitive math. If the offering succeeds, Musk will have $75 billion in fresh capital earmarked for orbital infrastructure. Bezos will need to decide whether to match that investment or cede the market.
The AI infrastructure spending race has already driven companies like Mistral to borrow hundreds of millions for GPU clusters on the ground. SpaceX and Blue Origin are proposing to move that arms race into orbit. The price tag just went from billions to tens of billions.
The Skeptics Have Strong Arguments
The counterarguments against orbital data centers are serious and well-documented.
The economics don't close yet. Microsoft's Project Natick, an underwater data center initiative, successfully met every technical target it set. The company abandoned it anyway. The economics of building, maintaining, and cooling computing infrastructure in hostile environments couldn't justify the cost premium over conventional facilities. Orbital data centers face the same fundamental challenge at far greater distances.
Latency only helps some workloads. AI inference benefits from low-latency global access, which orbital data centers could provide. But AI training, the most compute-intensive workload, relies on massive data batching and doesn't need sub-second response times. Putting training workloads in space makes no economic sense with current architectures.
Regulation is an open question. The FCC, ITU, and international space governance bodies have not established how orbital data centers would be classified. Are they telecommunications infrastructure? Computing assets? A new category entirely? A single adverse ruling could undermine the business model.
Space debris compounds the risk. A constellation of 1 million satellites increases collision probability in low Earth orbit. The Kessler syndrome, a cascade of collisions that renders orbital altitudes unusable, becomes a real concern at that scale. Blue Origin's Project Sunrise has already drawn warnings from scientists about its 51,600-satellite plan. SpaceX's proposal is 20 times larger.
The Retail Investor Play
One detail in the filing stood out to market watchers. SpaceX designated 30% of total shares for retail investors. That allocation is unusually large for an IPO of this size and represents roughly $22 billion flowing through retail brokerage channels.
The messaging writes itself. A generation of retail investors tracked SpaceX landings on social media, followed Starlink expansion in real time, and debated Mars timelines on Reddit. Now they can own a piece of the company. Musk has always understood the power of retail conviction, from Tesla's shareholder base to Dogecoin's community. The 30% allocation is a direct play for that same energy.
Musk's personal net worth sits at approximately $340 billion as of early 2026. A successful IPO at the stated valuation could push it toward 500 billion, given his estimated 40% ownership stake. That would make him wealthier than the annual GDP of most nations.
The Bottom Line
SpaceX's filing is the largest bet ever placed on the idea that artificial intelligence infrastructure belongs in orbit. The $75 billion raise, the 1.75-trillion-dollar valuation, and the 1-million-satellite plan represent ambition at a scale the public markets have never been asked to fund.
The filing itself is real. Bloomberg, Reuters, CNBC, and the Wall Street Journal all confirmed it. The technology foundation is real. Starlink works. Starship flies. xAI's Grok models run on production hardware.
What remains unproven is everything in between: the launch cadence, the orbital economics, the regulatory framework, and the decade-long execution required to make data centers in space cheaper than data centers in Virginia. Microsoft tried a simpler version of this idea with underwater servers and walked away. Deutsche Bank says parity is a decade out. Engineers who've reviewed the numbers are more blunt.
Musk's response to that skepticism has always been the same: build it, launch it, iterate in public, and let the doubters update their models. That strategy turned SpaceX from a startup that blew up its first three rockets into the most prolific launch provider on Earth. Whether it works for orbital data centers is the $75 billion question.
Sources
- Bloomberg: SpaceX Files for Historic IPO — Bloomberg, April 1, 2026
- SpaceX reportedly files plans for massive IPO — CNN Business, April 1, 2026
- Bloomberg, CNBC Report SpaceX Submitted Confidential Filing for IPO — Via Satellite, April 1, 2026
- SpaceX IPO 2026: Musk Targets 1.75 Trillion in Record Public Debut — The USA Leaders, April 1, 2026
- The SpaceX IPO Could Raise Over 75 Billion — Motley Fool, April 1, 2026
- SpaceX acquires xAI in bid to develop orbital data centers — SpaceNews, February 2, 2026
- Analysis: SpaceX's Orbital Data Centers Could Face Same Hurdles as Microsoft's Abandoned Undersea Project — Reuters via U.S. News, April 1, 2026
- Will data centers in space work? Elon Musk says yes — NPR, April 3, 2026
- Jeff Bezos joins Elon Musk in fueling Kessler syndrome fears — UNILAD, March 20, 2026