We didn't see $33 trillion. We saw a slide deck with no testnet, no radiation-hardened chip, and no orbital data center blueprint. Morgan Stanley's latest projection—SpaceX hitting a $3.3 trillion revenue run rate by 2040, powered by "AI orbital infrastructure"—is not analysis. It's narrative engineering.
Let me be clear: The source article, scraped from a blockchain/Web3 aggregator, quotes an analyst at a bulge bracket bank who claims SpaceX can become the backbone of global AI compute. The math: from $187 million in 2025 to $3.3 trillion in 2040. That's a 17,000x increase in 15 years. For context, that exceeds the combined market cap of every publicly traded company on Earth today. This is not a forecast. It is a liquidity magnet.
Context: The Narrative Factory We are in a bull market. Euphoria masks structural flaws. Every week, a new story emerges: AI agents, DePIN, RWAs, now orbital compute. The common thread? A massive TAM (total addressable market) projected out to an unrealistically distant horizon, with zero intermediate milestones. The Morgan Stanley piece is no different. It originates from a blockchain news site republishing a non-public research note. No link to the original report, no model methodology, no sensitivity analysis. Just a headline and a number.
I've seen this pattern before. During the 2021 NFT mania, floor prices were projected based on "community value" with no on-chain activity backing them. In 2020, DeFi protocols were valued at multiples of their TVL despite unaudited contracts. This is the same playbook: sell the vision, ignore the engineering reality.
Core: The Technical Impossibility Let me deconstruct the "AI orbital infrastructure" claim. As someone who has audited smart contracts and modeled computational bottlenecks in decentralized systems, I can tell you: the physics doesn't work.
First, power. A single H100 GPU consumes 700W. A Starlink satellite generates roughly 1-2kW from solar panels. To run even one H100 in orbit, you'd need to dedicate an entire satellite's power budget to just the GPU. No communication, no station-keeping, no thermal management. For a 100,000-GPU cluster—modest by terrestrial standards—you'd need 100,000 dedicated satellites, each costing $250,000 to build and launch. That's $25 billion before you run a single inference.
Second, radiation. Cosmic rays flip bits. Single-event upsets (SEUs) corrupt memory. Terrestrial chips fail within months in low Earth orbit unless hardened—typically using older, slower process nodes (65nm or above). The latest H100, built on 4nm, is extremely vulnerable. SpaceX would need custom radiation-tolerant AI accelerators. No such chip exists. Developing one is a 5-10 year program, $500 million minimum.
Third, latency. The analyst claims "low-latency AI" from space. Light speed in fiber is 200,000 km/s. In atmosphere, it's slower. A satellite 1,000 km away adds 7ms round-trip. That's fine for voice, terrible for real-time inference. Autonomous driving, high-frequency trading, interactive AI agents—they need sub-millisecond response. Space cannot provide it.
Contrarian: This Is VC Narrative, Not Engineering The contrarian angle: this report benefits Morgan Stanley's relationship with SpaceX. A boutique bank like Morgan Stanley leads private placements for SpaceX. Publishing a wildly optimistic valuation supports secondary market pricing for insiders seeking liquidity. The "AI infrastructure" label is a hook to attract deep-pocketed institutional investors—pension funds, sovereign wealth—who crave a story that justifies irrational multiples.
The article mentions "vertical integration" as a moat. Yes, SpaceX controls launch, satellite manufacturing, and ground stations. But vertical integration does not solve physics. It does not create a radiation-tolerant GPU. It does not generate 700MW of orbital power. The moat is in rocketry, not computing.
The real story: SpaceX is excellent at launching mass to orbit. That's its core competency. The AI narrative is a distraction. Smart money knows this. The contrarian trade is simple: sell the hype, buy the actual infrastructure assets—ground-based AI compute providers, satellite communications (non-SpaceX), and radiation-hardened chip makers.
Takeaway: Actionable Price Levels Do not treat this report as a valuation anchor. Treat it as a sentiment indicator. When such absurd projections surface in bull markets, it signals peak narrative saturation. The next leg is downward re-pricing as reality sets in.
For SpaceX: The private secondary market may see a temporary spike, but without a tangible milestone—a space-qualified AI chip, a demo of orbital inference—the premium will evaporate. For public markets: short any "space AI" concept stocks. For builders: focus on terrestrial compute. The orbital infrastructure is a decade away, minimum.
We didn't see $33 trillion. We saw a liquidity trap dressed in analyst prose. Exit before the narrative shifts.