Hook: When a single revenue projection exceeds the entire global GDP of today, you are no longer analyzing a company—you are witnessing a narrative weapon.
In July 2026, Morgan Stanley analyst Adam Jonas published a note projecting SpaceX’s revenue to hit $33 trillion by 2040. That number is larger than the combined GDP of the United States, China, and the European Union. The primary driver? Starmind—a yet-unbuilt constellation of AI satellites that would serve as orbital data centers. Before you dismiss this as a typo, let me be clear: the target is deliberate. It is designed to create a psychological ceiling so high that any smaller number seems achievable. But as someone who spent 2022 dissecting DeFi protocols with similar math, I can tell you: when the model depends on a physical impossibility, the only thing growing is the gap between expectation and reality.
Context: The Starmind Narrative
SpaceX, valued at around $270 billion in private markets, has long been the darling of space investors. Its Starlink constellation provides global broadband, and Starship promises to slash launch costs. Morgan Stanley’s 2025 research, relying on SpaceX’s S-1 filing and internal interviews, positions Starmind as the next exponential leap: a network of AI-optimized satellites that can process machine learning workloads in orbit, bypassing the latency and bandwidth constraints of ground-based cloud computing.
The report’s key numbers: 2025 revenue of $18.7 billion (mostly from launch and Starlink), ballooning to $3.19 trillion by 2030, and then to $33 trillion by 2040. The implied total addressable market is $28.5 trillion, of which $26.5 trillion is tied to AI. Jonas assigned a price target of $300 per share, implying a doubling from the then-current $125.
This is not a financial analysis. This is a science fiction pitch dressed in Excel formulas.
Core: A Systematic Teardown
Let me walk through the three pillars that make this projection structurally unsound.
1. Technology: Zero Architecture, Infinite Hubris
The Starmind vision is built on the assumption that Starship can cheaply deploy thousands of satellites, each carrying high-performance AI compute. But the report provides zero detail on chip architecture, inter-satellite networking, power generation, or thermal management. In space, there is no water cooling, no easy power scaling. A single H100 GPU consumes 700W. A constellation of 2,000 such satellites would need over 1.4 GW of continuous power—equivalent to a nuclear reactor—and would radiate heat equivalent to a small city. No solar array at scale can handle that. The report mentions “orbital data centers” as if they are just data centers in a different zip code. They are not. The physical engineering challenges are so acute that no company—SpaceX included—has demonstrated even a laboratory prototype.
Based on my audit experience with DeFi protocols that promised “decentralized compute” but ran on AWS, I see the same pattern: a narrative gap between what is said and what is buildable. The report’s technology section is essentially blank.
2. Commercialization: A $33 Trillion Revenue Company?
Here the math disconnects from reality. In 2023, global GDP was about $105 trillion. For SpaceX to generate $33 trillion in revenue by 2040, it would need to capture roughly 30% of the entire world’s economic output. Even if AI becomes the dominant sector, a single company owning a third of all economic activity is absurd. The report conflates the total addressable market for “AI” with SpaceX’s serviceable obtainable market (SOM). That is like claiming a restaurant in a city of $1 billion total food spending can capture the full $1 billion by selling a single dish.
Moreover, who are the customers? The report does not name a single potential client for orbital AI compute. Ground-based cloud providers like AWS, Azure, and GCP already serve most enterprise needs at lower cost, with infinite scalability. The only plausible niche is latency-sensitive applications—high-frequency trading, remote telemedicine, military drones—but that market is measured in tens of billions, not trillions.
3. Infrastructure: Starship Is Not a Silver Bullet
Starship’s low-cost launch is a necessary condition for Starmind, but far from sufficient. Each AI satellite would be a custom-built, radiation-hardened, self-sustaining computing node. The cost to design, test, and deploy such a satellite is orders of magnitude higher than a Starlink node. The report assumes that volume launch will solve the cost problem, ignoring that the satellite itself is the expensive part. Furthermore, maintenance in orbit is virtually impossible. A single malfunction could render a $50 million asset useless.
From my days auditing reentrancy vulnerabilities, I learned that elegance in design does not equal safety. Here, the design is not even defined.
Contrarian: What the Bulls Got Right
To be fair, there is a kernel of truth. The Starmind concept—if realized—could provide unique value for applications requiring ultra-low latency (under 10 ms) across global distances. High-frequency trading between exchanges in New York and Tokyo, for example, could benefit. Additionally, military use cases are real: orbital AI for battlefield intelligence would be a game-changer. The Pentagon has already invested in similar concepts.
But these niche applications represent at most $50–100 billion in annual revenue by 2040, not $33 trillion. The bulls are right that space-based compute has some demand, but they are wrong about the magnitude. The report’s error is not in direction—it is in scale.
Your alpha is someone else: the true alpha here is the narrative itself, which inflates SpaceX’s valuation to the benefit of late-stage investors and insiders looking for a liquidity event.
Takeaway: A Narrative Trap for the Unwary
Morgan Stanley’s analysis is not a forecast; it is a fundraising pitch. The $33 trillion number is not meant to be believed—it is meant to be repeated. For investors, the takeaway is clear: ignore the orbital AI fantasy and revert to first principles. SpaceX’s real business—rocket launches and Starlink subscriptions—is worth perhaps $150 billion at a generous multiple. The rest is speculative froth.
If Starmind never launches a single operational satellite, the stock will still have been sold on a dream. The question is: who will wake up first? Based on my experience watching ICOs promise world computers in 2017, the answer is always the same—the retail bagholder, holding a narrative that someone else got paid for.