Elon Musk’s SpaceX has spent two decades perfecting reusable rockets and blanketing Earth with Starlink satellites. Now the company wants to put AI itself into orbit. Its recent S-1 filing ahead of a massive IPO paints a picture far larger than launches or broadband. Of a projected $28.5 trillion total addressable market, $26.5 trillion sits in artificial intelligence. Investors still call it a space stock. They may soon call it something else.
The shift didn’t happen overnight. SpaceX acquired Musk’s xAI in a February 2026 all-stock deal that valued the combined entity near $1.25 trillion. The move pulled Grok, vast data centers and ambitious orbital plans under one roof. The Motley Fool laid out the thesis days ago. SpaceX already rents its Colossus 1 cluster to Anthropic for $1.2 billion a month. Alphabet pays nearly $920 million monthly for additional capacity. Those contracts turn idle GPUs into immediate cash flow.
But ground-based compute faces hard limits. U.S. data centers burned 177 to 192 terawatt-hours of electricity in 2024. That equals 4 to 5 percent of national power supply. Projections from the Electric Power Research Institute suggest the share could hit 9 to 17 percent by 2030. Water shortages, noise complaints and grid strain follow. Musk sees the bottleneck clearly. Solar energy arrives free in space. The vacuum provides natural cooling. No fuel burned on Earth. No communities displaced.
SpaceX applied in January to launch up to one million satellites dedicated to AI workloads. Early designs, dubbed AI1 or Starmind in recent discussions, feature solar arrays stretching 230 feet. Musk described the hardware in technical updates as building on existing Starlink V3 technology. “A lot of this is technology we’ve already made for the Starlink V3 satellites,” he said according to Reuters. “We don’t think this is a super hard problem compared to the things we already do.”
The numbers get staggering fast. Oppenheimer analyst Timothy Horan projects SpaceX revenue reaching $31.3 billion by 2026, driven heavily by Starlink and AI infrastructure. Fox Business covered his outlook last week. Goldman Sachs goes further in longer forecasts cited by The Motley Fool, seeing the AI segment alone climbing from $3.2 billion in 2025 revenue to $322 billion by 2030. Total company revenue could hit $474 billion. Those figures assume orbital data centers come online around 2028 and scale aggressively with Starship flights.
Starship itself sits at the center. Reusable, high-payload and rapidly iterated, it makes launching massive compute nodes economically feasible. Recent technical sessions outlined satellites that could deliver 120 kilowatts average power and peak near 150 kilowatts. Thermal systems borrow from Starlink. Power comes from expansive solar wings. The vision extends beyond low Earth orbit. Some engineers talk lunar manufacturing. Optimus robots could build satellites from regolith. Electromagnetic mass drivers on the Moon might fling hardware into space without chemical rockets. One X post from an xAI-affiliated account called it the pathway to raising humanity’s energy ceiling.
Yet execution carries real friction. SpaceX reported a $4.9 billion net loss in 2025 after a brief profit the year before. AI capital expenditures hit $12.7 billion last year, dwarfing spending on rockets or connectivity. This week the company launched a $25 billion notes offering to cover debt and fuel further AI expansion, Reuters reported Tuesday. Regulatory questions multiply. National security reviews, spectrum allocation and orbital debris rules all loom. The FCC continues examining the orbital data center filing.
Competition intensifies on the ground. Hyperscalers race to secure power purchase agreements and chip supplies. SpaceX struck a separate $6 billion deal this week with Reflection AI for data-center capacity, adding $150 million monthly starting July. The Wall Street Journal detailed the arrangement. Similar pacts with Anthropic and Google already exist. Meanwhile SpaceX acquired the AI coding startup behind Cursor for $60 billion in stock. The move bolsters Grok’s capabilities and gives access to developer data that can refine models.
Integration across Musk’s companies adds complexity and opportunity. Starlink provides the bandwidth backbone for orbital nodes to communicate with Earth. X supplies real-time data streams that can train Grok. A Starlink veteran now leads xAI’s human data team for Grok training, Bloomberg revealed earlier this month. Shared engineering talent flows between ventures. A joint Terafab chip venture with Tesla targets two-nanometer processes to feed the orbital clusters.
Critics point to past overpromises. Grok has lagged rivals in some benchmarks. Colossus 1 reportedly used a mismatched mix of Nvidia GPUs that proved inefficient for training, leading to its rental for inference instead. Ars Technica examined the IPO filing and orbital pitch in May. The company counters that Colossus 2 will standardize on newer Blackwell chips for frontier training. Orbital systems promise to sidestep terrestrial power wars entirely.
Financial markets prepare for volatility. SpaceX’s IPO could value it near $1.75 trillion, potentially landing it among the largest U.S. companies. Index providers debate fast-track inclusion rules. Nasdaq leans toward quick addition. S&P Global takes a more cautious stance. Either way, public trading will shine a harsher light on quarterly AI spending and launch cadence.
The long game looks different. If even a fraction of those one million satellites function as compute nodes, available AI training capacity could expand dramatically. Energy arrives 24 hours a day in orbit. Heat dissipates into the void. Latency for certain inference tasks might prove acceptable for non-real-time workloads. Military and enterprise users could tap sovereign orbital capacity. And the flywheel spins. More satellites mean more Starlink-like bandwidth demand. More launches refine Starship. Revenue from compute rentals funds the next wave.
Not every analyst buys the timeline. Building, launching and operating a million-satellite AI constellation demands manufacturing breakthroughs, autonomous servicing and novel cooling solutions at scale. Radiation hardening for GPUs in orbit adds cost. Data transmission back to Earth requires enormous antenna arrays or laser links. But SpaceX has lowered launch prices before. It scaled Starlink despite early skepticism.
Musk described the merger as creating the most ambitious vertically integrated innovation engine on and off Earth. The phrase appears across multiple reports including Futurum Group coverage from February. Whether that ambition delivers the largest AI company by infrastructure rather than chatbots remains the open question. For now the hardware flies. The models train. The contracts pay. And the satellites on the drawing board carry more than just internet signals. They may carry the future of compute itself.
Recent market moves reinforce the focus. SpaceX’s $25 billion debt raise targets AI expansion directly. The Reflection AI deal, covered yesterday by both Bloomberg and Reuters, adds another recurring revenue stream through 2029. Public investors will soon judge whether the orbital bet justifies the capital burn. The data centers of tomorrow might not sit in Virginia or Texas. They might circle silently overhead, drinking sunlight and returning answers at the speed of light.
WebProNews is an iEntry Publication