Hook

In a recent note from Morgan Stanley, analyst Adam Jonas painted a breathtaking picture: SpaceX, the rocket and satellite company, was no longer just a transportation or communications firm—it was an AI infrastructure giant. The headline number was staggering—2040 revenue of $33 trillion. That is not a typo. That figure, roughly 30% of projected global GDP at the time, was built almost entirely on a single unproven concept: Starmind, a network of AI-powered satellites running orbital data centers.

As a digital asset fund manager who has spent eight years watching capital chase narratives with far less grounding in physical reality, I felt an immediate, almost visceral skepticism. This was not analysis; it was fiction dressed in the language of finance. And for the crypto markets, which are currently awash in AI-themed tokens and decentralized compute projects, this report is a case study in the dangers of narrative-driven valuation. We have seen this movie before—in 2017 ICO whitepapers, in 2021 DeFi yield farms, and in the Terra ecosystem promises. The script is always the same: a beautiful story, a massive addressable market, and a complete absence of engineering feasibility.

Context: The Report That Defied Gravity

The Morgan Stanley report, published in early 2025, used SpaceX’s confidential S-1 filing and the firm’s own projections to argue that the company’s market cap could reach $300 per share (implying a valuation north of $400 billion) from a then-trading price of around $125. The core thesis hinged on Starmind, a project that envisions placing hundreds to thousands of AI-capable satellites in low Earth orbit, connected by Starlink’s laser links, to offer cloud computing services that are globally distributed, low-latency, and—crucially—seemingly immune to terrestrial constraints.

Jonas estimated SpaceX’s 2025 revenue at $18.7 billion, driven by launch services and Starlink subscriptions. By 2030, the number jumped to $319 billion. By 2040, $33 trillion. The report claimed a total addressable market of $28.5 trillion, of which $26.5 trillion was related to AI. The implication was clear: Starmind would capture nearly all of the AI market growth over the next 15 years.

Harvesting the liquidity that others overlook

I have audited dozens of token whitepapers that made similar leaps. In 2017, I evaluated a project called "EtherGem" that claimed to be a decentralized cloud provider using satellites—it failed because the team had no satellite experience and the tokenomics collapsed within six months. This feels hauntingly similar, but with a much larger stage and a few real assets: SpaceX does have a functioning rocket and a satellite constellation. But having a truck does not make you a logistics company, and having Starlink does not make Starmind viable.

Core: Forensic Deconstruction of the Narrative

Let me dissect the three pillars of the Morgan Stanley thesis—technology, commercialization, and infrastructure—and show why each crumbles under the weight of physics and market fundamentals.

Technology: The Code Is Missing

The report contained zero technical details on Starmind. What chip architecture? (GPU? ASIC? CPU?) How are the satellites interconnected for distributed training? What is the latency and bandwidth between nodes? How do they handle the extreme thermal and radiation environment of space? These are not minor details; they are the entire product. Without them, Starmind is a concept, not a roadmap.

From my experience writing DeFi liquidity mining scripts in 2020, I learned that the difference between a working system and a whiteboard sketch is the engineering grind. In space, that grind is orders of magnitude harder. The power draw for a single H100 GPU is 700 watts. A satellite currently has at most a few kilowatts of solar power. Running a cluster of 1,000 GPUs would require a megawatt of power—implying a solar array the size of a football field. And the heat? In vacuum, you can only radiate. The radiator area for 1 MW of thermal output would be enormous, adding mass and cost.

Watching the silence between the candlesticks

The authors of the Morgan Stanley report know this. But they also know that markets do not price physics; they price narratives. The 33 trillion revenue projection is not a forecast; it is a lure for capital. In crypto, we call this a "vapor node."

Commercialization: The TAM Fallacy

The report claimed a $28.5 trillion addressable market, with $26.5 trillion from AI. That is an egregious error. The total global AI market by 2030 is estimated at around $1.5-2 trillion by most credible analysts. The report appears to have taken "all software and cloud services that could use AI" and labeled it as Starmind’s potential. That would be like a single coffee shop claiming the entire global beverage industry as its addressable market.

Even if Starmind worked, who would pay for orbital compute? The cost per FLOP would likely be 10-100x higher than a ground-based data center, given the launch costs (even with Starship) and the energy budget. Unless there is a specific use case that demands global low latency (e.g., military command and control, high-frequency trading across continents), customers will choose cheaper and more reliable terrestrial options.

Diving for pearls in the deep web of value

I have seen this TAM inflation before. In the 2021 crypto bull market, projects claimed to capture the entire global payments market ($100 trillion) or the entire art market ($60 trillion). The reality was that even the best projects captured less than 0.1% of those markets. SpaceX’s current revenue is $18.7 billion; assuming Starmind captures 1% of the global AI market by 2030 (a generous assumption), that would be around $20 billion, not $319 billion. The rest is fantasy.

Infrastructure: The Physical Impossibility

The fundamental constraint is energy. A single Starship can lift about 100 tons to low Earth orbit. A data center-grade AI cluster (say 10,000 GPUs) would weigh several hundred tons and require megawatts of power. To deploy a cluster of 100,000 GPUs (the scale of a medium cloud data center) would require dozens of Starship launches, each costing tens of millions. The capital expenditure would be in the hundreds of billions, with ongoing maintenance and replenishment costs.

Moreover, the satellite-to-ground bandwidth is limited. Even with laser links, each satellite can only beam down a few gigabits per second. For a planet-scale AI inference service, that is a severe bottleneck. The entire architecture conflicts with the fundamental economics of computing: it is cheaper to bring data to compute than to bring compute to data, especially when that compute is in orbit.

The pattern emerges from the chaos of noise

This is not a critique of SpaceX. The company may succeed in building a small-scale orbital compute capacity for very specific applications (disaster response, military, etc.). But the idea that it will displace terrestrial cloud computing and generate trillions in revenue is a dangerous illusion.

Contrarian Angle: Why This Matters for Crypto Investors

You might wonder why a digital asset fund manager is writing about a SpaceX report. The answer is that this report is a symptom of a broader capital flow pattern that directly affects crypto markets. When mega-cap firms like SpaceX—or for that matter, BlackRock, Microsoft, or Nvidia—promise paradigm-shifting narratives that capture investor imagination, they divert massive amounts of risk capital away from other emerging technologies, including decentralized compute networks like Render, Akash, and Bittensor.

These projects offer a fundamentally different vision: a peer-to-peer network of distributed GPUs, powered by token incentives, that is already operational and serving real customers. They do not promise trillions in revenue; they promise a more efficient allocation of compute resources. And they are running right now, today, on the ground.

The Morgan Stanley report implicitly argues that centralized, capital-intensive, top-down infrastructure is the only way to achieve global AI compute. That is the opposite of the crypto ethos. But more importantly, it is likely wrong. The history of the internet shows that decentralized, permissionless networks often defeat centralized alternatives (e.g., TCP/IP vs. private networks, Bitcoin vs. sovereign currencies). The same could happen for compute.

Solitude reveals the truth the crowd ignores

In 2022, after the LUNA collapse, I retreated to a cabin in the Blue Mountains. I read Stoic philosophy and realized that markets crash not just because of bad fundamentals, but because narratives break. The SpaceX narrative is one of hubris. It assumes that the future of AI compute is a single, massive, managed platform. I believe the future is more diverse, more atomic, and more decentralized.

Takeaway: Cycle Positioning

This report is a classic top-of-cycle signal. When analysts start predicting 33 trillion in revenue from an unproven technology, it is time to rebalance your portfolio away from narrative-heavy assets and toward basics: Bitcoin as a monetary asset, Ethereum as a settlement layer, and maybe a few utility tokens that generate real yield from actual users. Do not buy the SpaceX IPO at a $400 billion valuation. And do not sell your decentralized compute tokens because a satellite story looks shinier.

Patience is the leverage that never depreciates

The stars are beautiful, but they are not data centers. Watch the silence between the candlesticks.

— Emma Thomas, Digital Asset Fund Manager, Sydney