Elon Musk dropped a bold claim this week. Speaking to SpaceX employees at an all-hands meeting, the chief executive declared that artificial intelligence revenue would eclipse every other part of the business as soon as September. Not probably. Definitely.
“Probably our AI revenue — not probably, definitely — our AI revenue will exceed all other SpaceX revenue probably in September, like next month,” Musk said. “And will significantly exceed all other SpaceX revenue in the fourth quarter.”
The remarks, captured in video and widely reported, mark a striking pivot for a company long defined by rockets, satellites and human spaceflight. Yet the numbers behind the shift have been building for months. In the second quarter of 2026, SpaceX posted $7.8 billion in total revenue, according to its earnings report. AI contributed nearly $2.6 billion of that total. Connectivity, powered largely by Starlink, brought in about $4.3 billion. The traditional space launch business lagged far behind.
How AI Compute Contracts Are Reshaping SpaceX’s Financials
Those AI dollars come primarily from renting out massive data-center capacity. SpaceX has struck lucrative cloud-services agreements with companies including Anthropic and Alphabet’s Google. The deals together generate roughly $2.15 billion in monthly contracted revenue, The Motley Fool reported. One agreement with Google is set to begin in October. The Anthropic pact is already active.
But Musk isn’t stopping at terrestrial data centers. He envisions a future where compute moves to orbit. SpaceX plans to launch its first AI data centers in space in 2027, The Wall Street Journal detailed following the company’s quarterly results. Solar power and the natural vacuum of space could solve the crushing demands for electricity and cooling that hobble ground-based facilities. Musk has spoken of scaling to terawatts of orbital compute. The vision ties directly to Starship, the company’s super-heavy rocket still undergoing rapid testing and iteration.
Recent deals show the momentum. Tesla is co-funding a $16.8 billion compute facility in Texas. SpaceX also agreed to acquire the AI coding platform Cursor for $60 billion, a move meant to accelerate enterprise AI capabilities. And the company released Grok 4.5 in July, its most powerful model yet. Training that model draws on the sum total of SpaceX data, with employees positioned as the “parents” of the system, Musk explained in the same all-hands session.
The scale is eye-watering. Musk told employees he expects 2 gigawatts of compute capacity online by the end of 2026. That figure could reach 10 gigawatts by the close of 2027. At $30 to $50 of value per watt, the latter target implies $300 billion to $500 billion in annual revenue. “Big numbers,” he said simply.
Longer term, Musk sees AI accounting for 99% of SpaceX’s value within four or five years. The company’s total addressable market now sits at $28.5 trillion. AI represents $26.5 trillion of it. Starlink? Just $1.6 trillion by comparison. These figures come straight from SpaceX’s own IPO-related disclosures.
Yet the capital demands are equally enormous. SpaceX spent more than $18 billion on capital expenditures in the second quarter alone. The majority went toward AI infrastructure. Full-year spending could approach $65 billion if the current pace holds. Losses narrowed in the quarter to $541 million from $1 billion a year earlier, but the company remains unprofitable as it pours money into growth.
Investors have taken notice. SpaceX trades at a market capitalization near $1.85 trillion following its public debut earlier this year. The stock has been volatile. Some analysts warn that delivering on orbital data centers requires Starship to achieve full and rapid reusability, something that remains unproven at scale. Others question whether the AI revenue ramp can outpace the spending spree.
But the second-quarter results told a story of acceleration. Revenue jumped 92% year-over-year. Adjusted EBITDA more than doubled to $3.5 billion. Starlink subscribers doubled, driving connectivity growth. Government contracts for Starshield added more than $6 billion in multi-year commitments. The company closed cloud-services deals worth $14.1 billion in contracted sales.
And. The launch business still matters. Falcon 9 and the emerging Starship architecture remain the enablers. Without cheap, reliable access to orbit, neither the Starlink constellation nor future AI satellites get off the ground. Musk has repeatedly stressed vertical integration as SpaceX’s core advantage. Rockets, satellites, software, data centers, even the AI models themselves all feed one another.
Recent coverage reinforces the shift. Yahoo Finance highlighted how Musk reframed the company as an AI hyperscaler that happens to control the world’s dominant launch provider. The future, he told the room, “is fundamentally AI and robots.” AI isn’t just another product line. It has become vital to the entire enterprise, in both hardware and software.
The Orbital Compute Gamble and What Comes Next
Space-based data centers represent the biggest leap. Early concepts call for solar-powered satellites with liquid cooling, operating at roughly 600 kilometers in altitude. One early design, dubbed AI1, features a 70-meter wingspan and 150 kilowatts of peak compute. Scaling to a constellation of thousands or even millions of such platforms could bypass Earth’s power-grid constraints entirely.
Critics point to the technical hurdles. Radiation in orbit can disrupt electronics. Data transmission back to Earth introduces latency. Maintaining and upgrading hardware far from technicians poses its own headaches. Yet Musk’s track record suggests he will iterate quickly. Starlink itself was once dismissed as unrealistic. Today it serves millions and generates the bulk of SpaceX’s current profits.
The acquisition of xAI, completed earlier in 2026, fused the AI talent and technology directly into the rocket company. The combined entity rebranded elements as SpaceXAI in some contexts. It also brought Grok into the fold, with plans to train future versions on vast troves of real-time SpaceX telemetry and internal knowledge.
Wall Street’s reaction to the latest earnings was mixed. Shares dipped even as revenue beat expectations, reflecting concern over the sustained capital burn. One research firm issued a sell rating shortly after the IPO, citing elevated valuations and the need to prove out multiple growth vectors simultaneously.
Still, the trajectory looks clear in Musk’s telling. By the fourth quarter, AI should significantly outpace everything else. Next year the gap widens further. And within half a decade the company could generate $1 trillion in annual revenue, Musk has projected. Much of that hinges on AI infrastructure, both on the ground and, eventually, in space.
So the question isn’t whether AI will dominate SpaceX’s revenue mix. Musk has made that outcome sound inevitable. The real test lies in execution. Can the company build and power those gigawatts without collapsing under the capital weight? Will Starship deliver the orbital infrastructure at the required cadence and cost? And how will customers respond to AI compute that lives off-planet?
Answers will come quickly. September is, after all, just weeks away. If Musk’s forecast holds, SpaceX will enter the final months of 2026 looking less like a rocket company with an AI sideline and more like one of the world’s largest compute providers that also flies to Mars. The transformation is already underway. The financials are starting to prove it.