Musk isn’t just talking about the moon – he’s talking about power
Elon Musk’s latest idea – a lunar factory catapulting AI satellites into orbit – sounds like classic sci‑fi showmanship. But the timing is anything but random: half of xAI’s founders have just left, SpaceX and xAI are merging, and a blockbuster IPO is reportedly on the way. Behind the spectacle is a much colder play: controlling the physical infrastructure that tomorrow’s AI systems will depend on. In this piece, we’ll look past the lunar hype to what this means for compute, regulation, corporate governance – and why Europe should be paying close attention.
The news in brief
According to TechCrunch, citing reporting from The New York Times, Elon Musk held an all‑hands meeting at xAI this week where he outlined an audacious plan: build a manufacturing facility on the moon to produce AI satellites and launch them into space using a giant electromagnetic-style launcher. The stated goal is to secure vastly more computing power than any rival AI company.
The discussion comes amid major internal churn. TechCrunch notes that six of xAI’s 12 original founding members have now left, including two co‑founders who announced their departures within roughly 24 hours of each other this week. At the same time, SpaceX and xAI have been combined into a single entity that is reportedly preparing for a SpaceX IPO as soon as this summer, with a targeted valuation around $1.5 trillion.
Musk has also publicly shifted emphasis from Mars to the moon, suggesting a self‑expanding lunar city could be built in about half the time of a Martian colony. TechCrunch further points out that his lunar ideas lean on a legal framework in which states cannot own the moon, but companies can own resources they extract, under a 2015 US law.
Why this matters: AI supremacy is becoming an infrastructure war
Strip away the lunar romance and Musk’s plan is about three things: compute, data, and control.
Whoever controls the most scalable compute stack for AI – chips, power, cooling, networking, and in this case orbital infrastructure – controls the pace of AI progress. Today that stack is largely terrestrial and centered on US hyperscalers like Microsoft, Google, and Amazon, plus Nvidia’s hardware monopoly. Musk is signaling that xAI and SpaceX intend to build an orthogonal stack that reaches beyond Earth itself.
From a strategic standpoint, a lunar factory is not just a bold engineering challenge; it’s a moat. If xAI can manufacture, launch and operate a dense mesh of AI‑optimized satellites from the moon, it could lock in:
- Dedicated compute that competitors cannot easily match
- Proprietary real‑world data streams (Earth observation, communications, sensor data)
- A vertically integrated system that sits outside normal terrestrial constraints like land, grid capacity and in some cases national regulation
The potential winners are clear: Musk, who further consolidates influence across Tesla, SpaceX, xAI and other ventures; US national interests, which gain another lever in the space and AI race; and investors who buy into the IPO story of an end‑to‑end, space‑to‑cloud AI infrastructure company.
The losers are less obvious but just as real. Open ecosystems and smaller AI labs will find it even harder to compete if the frontier is set by actors that own their own off‑planet infrastructure. Regulators will struggle to apply terrestrial frameworks to a stack that increasingly lives in orbit or beyond. And inside xAI, the rapid departure of half the founding team raises hard questions about governance, culture, and the level of internal alignment around this strategy.
In other words, this is not a sideshow; it’s a declaration about where the next layer of AI power will reside.
The bigger picture: from cloud wars to orbital moats
Musk’s lunar talk plugs into several converging trends.
First, the “compute crunch.” Over the past two years, AI labs have been constrained less by ideas and more by GPUs, energy and data center capacity. Big Tech is racing to lock in long‑term chip supply, build massive new data centers, and experiment with nuclear and alternative energy for AI workloads. Owning a differentiated orbital compute and communications layer is a logical – if extreme – extension of that arms race.
Second, the quiet shift from "AI models" to "AI infrastructure." OpenAI is tied tightly to Microsoft’s cloud. Anthropic is leaning on Amazon and Google. Meta is betting on open models but still rides on traditional hyperscale infrastructure. Musk seems to be aiming for something rarer: an AI company whose core advantage is not just models or data, but a physical, hard‑to‑copy infrastructure stack that spans rockets, satellites, and potentially lunar facilities.
We have seen this movie before. In the 19th century, control over railroads determined who could move goods and at what price. In the 20th century, oil majors and telecom giants dictated the rules of energy and communication. In the 21st, cloud platforms became the new utilities. A lunar‑enabled satellite megaconstellation dedicated to AI would be the latest incarnation of that pattern: own the pipes, and you shape the market.
Third, this aligns with Musk’s long‑running “world model” ambition, described in TechCrunch’s piece via a venture investor: combining data from Tesla (roads and energy), Neuralink (brains), SpaceX (space), The Boring Company (subsurface) and now the moon. Whether that level of integration is technically or politically feasible is another matter, but the direction of travel is unmistakable: richer, more proprietary data feeding ever larger models.
Against competitors, the messaging is blunt. While OpenAI argues about product roadmaps and safety charters, and Google DeepMind refines its research stack, Musk is essentially saying: we’re going to change the substrate everything runs on.
The European angle: strategic autonomy meets lunar capitalism
For Europe, this is not just an entertaining Silicon Valley subplot. It cuts straight into debates about digital sovereignty, space policy and the EU’s ability to shape AI’s next phase.
Europe already worries about depending on US cloud giants for critical infrastructure. Now imagine a future in which the most powerful AI systems also depend on an orbital and possibly lunar infrastructure controlled by a single US‑centric private entity. That is the opposite of "strategic autonomy".
Regulators in Brussels have been aggressive on data, platforms and AI – with GDPR, the Digital Markets Act, Digital Services Act and now the EU AI Act. But these frameworks largely assume that infrastructure is either terrestrial or at least governed through traditional telecom and satellite regimes. A lunar factory building AI satellites sits awkwardly outside existing EU toolkits.
There is also the space policy angle. The EU and ESA have their own lunar concepts, such as the long‑discussed "Moon Village" idea, and member states like Luxembourg have passed national laws around space resources. But Europe is far from fielding an integrated AI‑plus‑space stack at Musk’s scale. European launch providers are only just regaining momentum after delays in Ariane 6, while US and Chinese actors sprint ahead.
For European AI companies – from frontier players like Mistral and Aleph Alpha to regional startups – the risk is that the game board shifts again. They may find themselves building sophisticated models that ultimately run on someone else’s off‑planet infrastructure, with limited bargaining power.
The EU still has levers: competition law, procurement, space collaboration, and the ability to set norms around resource extraction and off‑planet data governance. But it will need to start treating orbital and lunar AI infrastructure as a strategic domain, not a distant curiosity.
Looking ahead: hype, hardware and hard politics
Several things are likely to happen next.
In the near term, the lunar factory will function primarily as narrative fuel for the combined SpaceX‑xAI IPO story. It allows bankers to pitch not just rockets and satellites, but a vertically integrated AI infrastructure play with almost limitless upside. Expect glossy decks, concept art and a lot of talk about synergies between launch, constellations and AI.
Whether hardware follows the hype is a tougher question. Building any form of industrial footprint on the moon is a multi‑decade, ultra‑capital‑intensive project involving NASA or other national agencies, complex logistics, and unproven technologies. Even by Musk’s standards, timelines are likely to slip. A more realistic path is incremental: more advanced Starlink‑style constellations optimized for AI workloads, experiments in in‑orbit compute, and small‑scale lunar precursor missions.
Internally, the departure of six out of 12 founders before an IPO is a red flag. It may simply reflect the classic transition from early‑stage chaos to late‑stage structure, as Musk suggested. But it can also signal deeper misalignment about risk tolerance, governance, or the balance between bold vision and near‑term execution. Watch who replaces them, what kind of board is assembled, and how much independence xAI leadership really has inside the merged entity.
On the regulatory front, the next few years will likely feature:
- Growing debate over space resource rights, especially between US‑aligned Artemis Accords members and countries like China and Russia
- Questions about how export controls, AI safety rules and data protection apply to off‑planet infrastructure
- Potential attempts by the EU and other blocs to shape multilateral norms around lunar activities and AI in space
The main risk is concentration: a single private actor controlling key layers of both the space and AI stacks. The opportunity, if managed well, is accelerated innovation in communications, Earth observation, and even sustainability – for example via more efficient monitoring of climate and infrastructure from orbit.
The bottom line
Musk’s lunar factory idea is less about science fiction and more about owning the next layer of AI infrastructure. If even a fraction of this vision materialises, it will concentrate unprecedented technical and geopolitical power in very few hands. The question is no longer just "Can he build it?" but "Who gets to set the rules when AI’s nervous system extends beyond Earth?" That’s a conversation regulators, investors and citizens need to join now, not after the first lunar catapult goes live.
