I don’t believe in SpaceX’s AI narrative until I see a single customer who isn’t Elon Musk.
Let’s start with the number that stopped me cold: 90% of SpaceX’s future growth will come from AI. Not rockets. Not Starlink. Not Mars. AI. That’s what ARK Invest’s recent analysis claims the company is telling its investors. A rocket company that built its reputation on reusability and cost disruption is now pitching itself as an AI infrastructure play. The audacity is impressive. The economics? Less so.
Context: The Narrative Shift
SpaceX just completed its largest IPO in history. That’s the context you need to understand why this story is being pushed. The company is no longer just a launch provider. It’s now a vertically integrated conglomerate: rockets, satellites, ground stations, data centers, and AI models (via the acquisition of xAI and Grok). The pitch is simple: by controlling the entire stack from Earth to orbit, SpaceX can offer compute at a fraction of the cost of traditional cloud providers. ARK’s report estimates launch costs could drop to under $100 per kilogram at scale. Today, Falcon 9 runs around $1,500/kg. That’s a 15x improvement — if it happens.
The article also claims that orbital data centers cost 25% less to build than ground-based ones and have near-zero energy costs thanks to solar power. Customers like Anthropic and Google are already leasing compute resources. Sounds like a classic disruption story. But when I dig into the assumptions, the cracks start to show.
Core: Deconstructing the Economics
The launch cost variable is everything
If Starship never reaches sub-$100/kg, the entire AI narrative collapses. Period. We’ve seen this movie before with fusion energy and flying cars. The math only works if the key enabler hits an aggressive trajectory. I’ve spent years analyzing infrastructure trade-offs in crypto — Layer2 rollups, sharding, DAGs — and the lesson is always the same: a single unvalidated assumption can wipe out years of work. Right now, that assumption is Starship’s cost curve. Next orbital test is critical. If it slips by two years, the valuation story becomes a liability.
Orbital data centers: physics is not optional
The claim that space-based data centers are cheaper to build and operate is seductive. Solar energy is free in space — yes. But the hardware to survive the environment isn’t. GPUs generate massive heat. In a vacuum, you can’t use air cooling. Radiative cooling requires heavy radiator panels. Add radiation hardening, vibration damping, and the cost of launching the whole package. The 25% savings ARK cites might be true for a prototype, but at scale? Unlikely. From my time auditing smart contracts, I’ve learned that the easiest cost to underestimate is operational complexity. Space operations are the definition of complexity.
Customer validation is thin
Anthropic and Google are mentioned as compute clients. But how much? At what price? For what duration? Without public contracts or revenue disclosure, these are likely trial commitments or strategic partnerships. Google is both a potential customer and a direct competitor (GCP). That’s a conflict of interest that makes me skeptical. In crypto, we see this all the time — a protocol announces a “partnership” with a major exchange, but the integration is a single wallet connection. Real adoption is measured in locked value, not press releases.
The 90% growth claim is a math trick
If 90% of future growth comes from AI, then Starlink and launch services together account for only 10%. Starlink already generates over $10 billion in annual revenue and is growing. For AI to dominate, its absolute revenue must be astronomical. That implies either Starlink growth is about to hit a wall, or AI infrastructure is going to explode like AWS did. AWS took over a decade to reach $10B. SpaceX is asking investors to believe it can do that in half the time, from a standing start, in a sector with entrenched incumbents. The math doesn’t add up.
Vertical integration: sword or shield?
ARK makes a strong case for vertical integration as a moat. SpaceX controls rockets, satellites, ground stations, data centers, and AI software. That’s powerful on paper. But it also means SpaceX carries all the risk. One Starship explosion that destroys a batch of orbital compute nodes? Capital loss plus delayed revenue. No one else to blame. In crypto, we call this “single point of failure.” Decentralization exists precisely to mitigate this. SpaceX is the most centralized compute provider you could imagine.
The competition isn’t standing still
AWS, Azure, and GCP have decades of experience, ecosystems, and software stacks. They can offer not just raw compute but managed services, security certifications, and global availability. SpaceX’s orbital data centers, even if operational, will have latency issues due to orbital mechanics. AI training requires low-latency interconnects. Inference for real-time applications? Forget it. The niche for space-based compute is probably high-security, batch-processing workloads. That’s a fraction of the total addressable market.
And let’s not forget the crypto-native compute networks. Akash, Render, Filecoin — they’re already aggregating idle GPU resources from around the world. They don’t need rockets. They just need a smart contract and a token incentive. Their cost advantage doesn’t depend on engineering miracles; it’s already baked into the utilization of existing hardware. Technology without distribution is just a hobby. SpaceX has distribution — but crypto networks have alignment.
Contrarian: The Angle Nobody’s Talking About
What if SpaceX is actually setting itself up as a wholesale compute provider for the crypto ecosystem? Imagine a network of orbital data centers that offer verifiable, tamper-proof compute — perfect for blockchain validators, oracle nodes, or privacy-preserving computation. Zero trust required. The physical security of space plus the transparency of on-chain verification. That would be a killer use case. But it’s not the one ARK is pitching. ARK wants a $500B cloud market story. The crypto angle is a side note.
The real contrarian take: SpaceX’s AI pivot is a distraction from the fact that Starlink’s growth is plateauing. The broadband market has a ceiling. AI infrastructure has a much higher ceiling. By rebranding, SpaceX can justify a higher valuation in its next funding round. It’s a financial narrative, not a technical roadmap. This isn’t about being right—it’s about not being wrong. And for ARK, which holds SpaceX in its funds, publishing optimistic analysis is a form of price support.
Takeaway: What to Watch
I’m not dismissing the vision. In five to ten years, space-based compute could be a meaningful supplement to terrestrial clouds. But the path is narrow and the risks are high. Here’s my watchlist:
- Starship next test flight: If it fails, the timeline slips two years.
- Customer contracts: Real revenue numbers, not names.
- Technical whitepapers: How they plan to cool GPUs in space. Without a credible solution, it’s vapor.
Crypto markets should pay attention because the same capital that flows into SpaceX’s AI narrative could flow into decentralized compute projects if SpaceX stumbles. Decentralized compute doesn’t need to solve orbital mechanics. It just needs to solve coordination. And that’s a problem crypto is uniquely suited for.
If you’re not embarrassed by your first version, you launched too late. SpaceX’s first version of orbital compute looks ambitious. But I’d rather back a network that can iterate on software than one that must iterate on hardware in space.
