SpaceX’s $184.8 Billion AI Bet Is Really a Test of Who Owns Compute

A rocket company just asked Wall Street to believe it is a cloud company. On August 7, 2025, Bank of America kept its Buy rating on SpaceX, set a target price of $235 against a reference price near $125.33, and added a claim that changes the entire investment narrative. SpaceX’s AI infrastructure business, not launch, is now the growth engine. The bank projects 2026 AI revenue near $24.5 billion — 52 percent of the $46.9 billion total revenue it expects for the same year. Anthropic has been a paying customer since May. Google’s compute partnership begins in October. I have audited enough token contracts to know that a multi-billion dollar revenue line without an open architecture is a promise, not a proof. That sentence is not a criticism of SpaceX. It is a discipline I learned in 2017, when I spent four months auditing ERC-20 standards for three young Cape Town projects. I found reentrancy vulnerabilities in two of them. Both later collapsed, and I watched $45,000 in potential losses evaporate before the community even understood what had happened. The lesson was simple: technical precision is a form of social protection. A balance sheet can be elegant and completely disconnected from the code that is supposed to back it. When I look at Bank of America’s SpaceX numbers, I feel the same need to trace the revenue back to the physical architecture behind it. BofA’s full forecast is breathtaking. Total revenue grows from $46.9 billion in 2026 to $100.7 billion in 2027 and $184.8 billion in 2028. That is roughly a fourfold increase in two years. Free cash flow is negative across all three years: minus $43.6 billion, minus $45.4 billion, and minus $37.4 billion. Add those, and SpaceX would burn about $126.4 billion while building whatever “AI infrastructure” actually means. For a company that has historically disciplined itself around reusable rockets and vertical integration, this is an enormous philosophical shift. It is no longer a launch company. According to BofA, it is a capital-hungry compute utility. But what is the product? The report does not say. There is no mention of GPU clusters, chip sourcing, data center locations, energy contracts, or network latency. Is SpaceX building orbital data centers powered by solar panels, cooled by the void, and connected through Starlink? Or is it building ground-based supercomputers near its launch sites? Or is it simply an intermediary that resells cloud capacity to AI labs under long-term contracts? The difference matters more than the revenue figure. A satellite data center faces radiation, collision risk, and high bandwidth costs. A ground data center faces power procurement and heat management. A reseller faces nothing but a paper margin. Yet the bank’s valuation treats all three as the same business. The honest response is to stress-test the financial assumptions, because there is no public engineering blueprint to verify. Let me be precise about what the $24.5 billion number implies. In 2026, 52 cents of every dollar SpaceX earns would have to come from AI contracts. With only two named customers — Anthropic and Google — that means a handful of signatories would be providing the majority of the company’s revenue. In token markets, we call this oracle centralization. If one oracle fails, the entire protocol can be drained. If one AI lab renegotiates or delays payment, the equity story of a $235 target price breaks. The bank’s model probably assumes signed multi-year agreements, but signed agreements are not a substitute for verifiable delivery. In 2020, I ran a “DeFi for Everyone” workshop in Cape Town to explain impermanent loss. I saw what happens when retail users trust marketing instead of mechanics. They confuse a contract with a commitment. The same confusion is happening now, except the contract is a sell-side price target and the commitment is a rocket company’s capex plan. The capital intensity deserves a second look. A cumulative free cash flow deficit of $126.4 billion does not appear in a vacuum. It means SpaceX will need equity, debt, pre-payments, or some combination of all three. Every new financing round will dilute existing believers or increase fixed charges on future revenue. Bank of America’s $235 target price is a clean number, but it is sitting on top of a messy pile of future borrowings. When I look at infrastructure projects, I ask one question: who gets paid first? In a protocol, the answer should be the users who provide security. In SpaceX’s case, the answer is probably the banks underwriting the debt. That is not evil; it is just not decentralization. After the 2022 crash, I sat with developers in Code & Conversation sessions. We converted despair into structural lessons. The same resilience is needed now, except the risk is a single point of computational failure rather than a token price. Now, let’s talk about what open source has to do with any of this. Every line of code is a hand extended in trust. When I helped ten South African digital artists build a royalty enforcement toolkit in 2021, we found that 60 percent of secondary sales on major NFT platforms did not automatically pay creators. The marketplace did not want to fix it. So we wrote open source smart contract modules that did the job. The artists did not need permission; they needed a protocol that respected their ownership. Artists own their pixels; we just hold the keys. That is the principle I want to carry into the AI infrastructure debate. If AI compute becomes as concentrated as launch capability, then the artists of the next generation will not even own the process that creates their work. They will rent it from a vertically integrated provider. The bank’s report is not asking whether that future is desirable. It is simply pricing it. The contrarian angle is not to short SpaceX. The contrarian angle is to question whether centralized AI infrastructure should be the default answer at all. BofA may be right about the revenue. SpaceX may indeed become the world’s most important compute contractor. That is exactly what worries me. We have spent the last decade trying to decentralize money, identity, and content. If we wake up in 2028 with three giant corporations controlling all meaningful AI compute, the blockchain experiment will have been reduced to a ledger for a niche decentralized finance culture. Open source is not a license; it is a promise. That promise is broken when the most important computational resource in the world is sold as a service with no transparency, no governance layer, and no community audit path. This is where decentralized physical infrastructure networks, or DePIN, enter the story. Token-incentivized compute networks are still young, but they represent a different model. Instead of putting GPUs in a satellite launch stack, they put them in the hands of thousands of independent operators. Instead of negotiating with one hyperscaler, they allow any user to verify where computation is happening. I am not saying DePIN will automatically outperform a vertically integrated SpaceX. I am saying the design question is political before it is technical. Do we want a world where the largest AI infrastructure provider is a private company with launch capabilities, or a world where compute is a shared, auditable substrate? The market is going to chase the $184.8 billion number. I would rather chase the protocols that let ordinary people verify whether that number is real. In 2025 I worked with a global team of fifteen researchers on decentralized identity and AI verification. We built a framework that let five thousand users prove the origin of digital content without exposing personal data. We prevented two thousand identity fraud attempts. The project succeeded because we refused to separate the technical from the human. Decentralization was not a fashion; it was a defense mechanism against a world where AI can imitate anyone. That experience taught me that the next frontier is not faster chips. It is verifiable trust. We build bridges, not just blocks, between people. So what should a thoughtful investor or builder take from this report? Bank of America has, perhaps unintentionally, provided the clearest signal yet that AI compute is becoming the new sovereign currency. Education is the only true decentralized currency, yet it cannot be monetized by an exchange or a rocket. Those of us who care about open systems have a choice. We can accept the SpaceX story as an inevitability and buy tickets to a future we do not control. Or we can use the next two years to build infrastructure that is owned by its users, audited by its community, and resilient enough to survive a bank’s spreadsheet error. Tracing the code back to the conscience behind it is the work. The bull case for SpaceX is not the point. The point is who gets to build the critical infrastructure of the age — and who is left outsourcing that responsibility to a launch provider with a price target. We have seen what happens when code becomes law without equity. Let’s not repeat it with compute.