The neocloud is already SpaceX’s economic center of gravity
The company’s artificial intelligence division billed more than three times its previous year’s total in its first quarter as a publicly traded company, and now surpasses traditional space business in revenue.
SpaceX went public, and Elon Musk became the planet’s first trillionaire. Less than two months later, the first quarterly report has revealed an uncomfortable truth for romantics of space exploration: the company that symbolizes private Mars exploration is today, in terms of its bottom line, essentially a company renting out computing power for artificial intelligence. The neocloud—that massive processing infrastructure that feeds large models—has displaced the rocket as the economic center of gravity. This is no accident: it is confirmation that the value chain of artificial intelligence is no longer controlled only by model labs, but by those who possess the energy and machines to train them.
The silent shift toward computing
The data, published by TechCrunch and The Verge, is stark. SpaceX doubled its total revenue year over year, driven by three vectors: computing agreements with Anthropic and Google, and Starlink’s growth. But the detail that redefines the reading is another: according to the IPO documents cited by The Verge, the artificial intelligence division was the source of most of the company’s revenue. The U.S. outlet put it this way: SpaceX earned more revenue as an artificial intelligence company than as a space company.
It is worth pausing on the magnitude of the shift. SpaceX’s AI division went from a modest annualized turnover to bringing in far more in a single quarter. That growth does not come from launching satellites or manned missions: it comes from renting out computing capacity to other AI companies that need to train and operate ever-larger models. In industry jargon, this is known as a neocloud: cloud computing infrastructure specialized in AI workloads, where the scarcity of high-end chips and electrical power turns the owners of those facilities into the new gatekeepers of access to technology.
The irony is striking. SpaceX built its legend sending rockets into space; its financial future, however, is being written in hangars full of GPUs and liquid cooling systems. The company has not abandoned its space mission—Starlink continues to grow and the launch business continues—but quarterly evidence suggests AI has become its true economic engine.
The neocloud frenzy: Anthropic and the war for computing
SpaceX’s move is not an isolated case. The same day its results came out, TechCrunch reported that Anthropic—one of the world’s most valuable AI labs—had signed a multibillion-dollar deal with Volta, a startup specializing in cloud services for AI. The pact is part of what the outlet describes as a wave of cloud alliances for Anthropic in recent months.
The strategic reading is clear: Anthropic is aggressively diversifying its computing supply. The company cannot afford to depend on a single infrastructure provider. Its deal with SpaceX—part of the revenue that has driven up the AI division’s turnover—is now complemented by this alliance with Volta. The message to rivals and investors is unambiguous: access to computing is as strategic as model quality, and those who control that infrastructure will hold non-negotiable power.
This phenomenon has a name in the sector: the computing war. AI labs spend enormous sums on processing capacity, and neoclouds—companies building data centers specifically designed to train large-scale models—have become the indispensable intermediaries of that economy. SpaceX, with its access to energy and massive facilities, has capitalized on that demand extraordinarily quickly.
The price of the bubble: Musk and the volatility of the new order
But there is another side to this story worth examining with cold eyes. The same week SpaceX posted record results, Elon Musk’s fortune suffered a historic setback. According to data compiled by Chinese outlet 36Kr, Musk lost an unprecedented amount in July—the largest monthly decline ever recorded—leaving his net worth at its lowest level since December. The figure remains inconceivable: it exceeds the combined fortunes of the billionaires ranking second through tenth, and is more than double the wealth of Jensen Huang, Nvidia’s CEO. Larry Page, second on the list, holds a fortune far below Musk’s.
The paradox is fascinating. SpaceX, Musk’s crown jewel, just proved that its AI business is more profitable than its space business. And yet the market punishes his fortune with unprecedented violence. What explains this contradiction? The answer likely lies in the speculative nature of the current cycle. Investors reward computing growth, but they also fear a bubble: if neocloud revenues depend on AI labs’ ability to monetize their models, any sign of slowing AI demand could trigger a devastating domino effect.
Musk’s decline, moreover, is not just a color detail about the volatility of the ultra-rich. It is an indicator that the market is reassessing the fundamentals of the AI economy. Infrastructure is worth a great deal, but its long-term profitability remains an open question. SpaceX has shown it can bill like a neocloud; what it has not yet shown is that this business model is sustainable when the AI investment cycle cools.
The new map of technological power
What we are witnessing is a reordering of the global industrial map. For years, the dominant discourse in AI centered on model labs, from Anthropic to major Chinese labs like Qwen, DeepSeek, or GLM. The narrative was that whoever developed the best model would control the future. But recent months have revealed a more prosaic and more powerful truth: whoever controls infrastructure—chips, energy, data centers—controls access to AI.
Neoclouds have become first-tier strategic players. Companies like Volta now sign multibillion-dollar deals with the world’s most important labs. SpaceX, which no one would have imagined as a computing provider, has found in this economy a vein that surpasses its original business. The question for governments is whether the infrastructure underpinning AI requires its own rules.
The question left hanging is uncomfortable: are we facing a bubble or a structural shift? Neocloud advocates argue that computing demand is insatiable—each new model generation requires more capacity than the last—and that the scarcity of energy and high-end chips will keep margins healthy for years. Skeptics, for their part, point out that neocloud revenues depend on labs’ ability to generate returns on their own models, something not yet demonstrated at scale.
What seems clear is that AI has ceased to be an exclusively algorithmic battle. The new frontier is physical: it is fought in data centers, on power grids, and in multibillion-dollar agreements for processing capacity. SpaceX, the company that promised to take humanity to Mars, has found its most immediate future on Earth, selling computing to the machines that are rewriting the world. The conquest of space can wait; the conquest of artificial intelligence cannot.