We audit the code, but who audits the conscience?
Last week, SemiAnalysis dropped a report that sent shockwaves through both the AI and crypto communities. SpaceX, according to the analysis, is on track to deploy over 10GW of computing power by the end of 2027. Elon Musk himself confirmed a conservative target of 6-8GW, with upside exceeding 10GW. At roughly $50 billion per GW, that means 2027 capital expenditures alone could hit $300-500 billion. The numbers are staggering. But as someone who has spent years auditing the ethical architecture of decentralized systems, I see a different story—one that questions not just the feasibility, but the soul of this infrastructure.
Context: The Compute Cartel
To understand the implications, we must first grasp the scale. SemiAnalysis models that when OpenAI and Anthropic run API inference on GB300 clusters, each GW can generate over $100 billion in annual revenue. At a rental price of $3 per GPU hour, the cost per GW is about $12 billion per year. The report also estimates that Microsoft's $250 billion infrastructure agreement with OpenAI in October 2025 corresponds to about 7GW. And now, Microsoft is allegedly signing a compute contract with SpaceX for roughly 3GW, valued at $150 billion. SemiAnalysis projects SpaceX's annual recurring revenue could reach $300 billion by end of 2027.
This is not just a power play. It is a concentration of computational resources that rivals the early days of the internet—except this time, the gatekeepers are not ISPs, but billionaires with rocket companies. For a blockchain evangelist like me, this raises a fundamental question: if compute becomes as centralized as hash power already is, what happens to the promise of permissionless innovation?
Core: The Hash Power Parallel
I have spent the last 14 years watching the evolution of decentralized networks. I've audited DAO governance models, reverse-engineered yield farming logic, and interviewed artists marginalized by the crypto bro culture. One pattern emerges repeatedly: the more compute concentrates, the more the network becomes a puppet of its largest stakeholders.
Take Bitcoin. After the fourth halving, miner revenue collapsed. Today, hash power is increasingly concentrated in three pools. The decentralization consensus is becoming hollow. The same dynamic is now unfolding in the AI compute layer. If SpaceX controls 10GW of compute, the cost to enter the high-performance computing market becomes prohibitive for anyone without a private rocket company.
I recall my experience auditing the 1Balance DAO in 2017. I found centralization risks in their smart contracts that would have allowed a single whale to control governance. The report was ignored until the project collapsed. The lesson? Centralization is not a bug in human systems—it is a feature of laziness and greed. SpaceX's compute empire is no different. It is a hardware-level version of the same problem.
Build not for the peak, but for the plain.
SemiAnalysis's revenue projections assume that demand for AI inference will continue to grow exponentially. But what if the demand is artificially inflated by the very entities building the infrastructure? The $300-$500 billion in capex is not just building machines; it is building a narrative that only the largest players can afford to play. For the rest of us, the plain—the everyday developer, the open-source researcher, the small-scale miner—we are left with scraps.
Contrarian: The Pragmatic Test
Here is the counter-intuitive angle: Musk's compute empire might actually accelerate the need for decentralized compute. If the cost of centralized inference becomes prohibitive, the market will naturally seek alternatives. I have seen this pattern before. During the 2022 bear market, when centralized exchanges collapsed, decentralized finance experienced a renaissance. Similarly, if SpaceX and Microsoft control the compute layer, the demand for verifiable, distributed compute networks—like those built on blockchain—will surge.
But there is a blind spot. The scalability of decentralized compute is still orders of magnitude behind. My own research on Layer 2 scaling solutions during the 2022 bear market showed that even the best optimistic rollups can only handle a fraction of the throughput needed for large-scale AI inference. The gap between 10GW and 10MW of decentralized compute is not just a power difference; it is a philosophical chasm.
Takeaway: The Conscience of Infrastructure
The numbers are seductive. $300 billion in annual revenue. 10GW of compute. A rocket company that builds the future. But as an open source evangelist, I ask: who audits the conscience of this infrastructure? We will continue to audit smart contracts, but the hardware layer is opaque. The code is not the law when the hardware is owned by a single entity.
I am not against progress. I am against progress that forgets the plain. Build not for the peak, but for the plain. The peak is where the billionaires sit. The plain is where the rest of us innovate. If SpaceX's 10GW becomes the new standard, we must ensure that a portion of that compute is accessible to the public, not locked behind a corporate API.
The final question is not whether SpaceX can deliver 10GW by 2027. It is whether we, as a community, will demand that the infrastructure serves the many, not the few. The answer will determine whether the next decade belongs to decentralization or to a new feudalism of silicon.
