The news broke quietly on Crypto Briefing, but its implications are seismic: Anthropic, the AI frontier lab, has signed 70 to 80 Letters of Intent for data center capacity. On the surface, this is a bullish signal—a startup scaling to meet demand. But for those of us who have lived through the ICO mania, the DeFi summer, and the NFT gold rush, this number triggers a familiar discomfort. It’s the same pattern: a central entity securing massive physical infrastructure, pre-empting competition, and shifting the balance of power.
I’ve seen this playbook before. In 2017, during MakerDAO’s early days in Cape Town, I watched as ICO projects pre-ordered server capacity and promised the moon. The difference then was that we were building decentralized finance. Now, the AI industry is mirroring the same centralized infrastructure play, but with a crucial twist: it’s happening without the guardrails of community governance or transparent on-chain accountability.
Let’s unpack the context. Anthropic is a private company backed by Google, Salesforce, and others. Its 70-80 LOIs represent a bid for approximately 700-1,600 MW of capacity—enough to power a small city. This is not just about training next-generation models; it’s about building a global inference network that can serve enterprise clients with low latency. The philosophy behind this is pure centralization: control the compute, control the narrative. But as a community that champions decentralization, we must ask: what happens when the gatekeepers of AI become as powerful as the banks we sought to replace?
Code is law, but ethics is conscience. The core of my analysis lies in the values embedded in this infrastructure play. Anthropic’s LOIs are a bet on proprietary, closed-stack compute. They lock in relationships with data center operators like Equinix and Digital Realty, and likely with NVIDIA for GPUs. This is reminiscent of the early days of Bitcoin mining, when Bitmain controlled both the hashrate and the hardware. We saw how that centralization led to censorship fears and cartel-like behavior. Now, AI compute is following the same path. The irony is that the crypto ecosystem has been building decentralized alternatives—Bittensor for model training, Render Network for GPU sharing, and a host of DAO-governed compute marketplaces. Yet, the big money is flowing the other way.
Based on my experience running community education for MakerDAO, I know that the moment a project starts buying physical infrastructure in bulk, the power dynamic shifts. The community becomes a customer, not a participant. The 70-80 LOIs are not just about capacity; they are about creating a moat. Anthropic is signaling to investors: “We have the compute, so we have the edge.” But this edge comes at a cost. It accelerates the winner-take-all dynamics of AI, where the largest models are trained in secret, and the public loses access to the base layer.
Solidarity over speculation. Here is the contrarian angle: perhaps this is necessary. Perhaps Anthropic needs this scale to ensure alignment research and safety testing are done responsibly. I’ve worked on AI governance frameworks myself—in 2025, I co-authored a whitepaper on human-centric AI for the Ethereum Foundation. I understand the argument that controlled environments reduce risk. But the pragmatism test fails when we examine the history of centralized infrastructure. The same logic was used to justify centralized exchanges, proprietary blockchains, and walled-garden DeFi platforms. Each time, the result was opacity, single points of failure, and regulatory capture. The LOIs, if signed, will create a permanent advantage for Anthropic, making it harder for decentralized alternatives to compete.
Moreover, the sheer number of LOIs—70-80—is a PR tactic. I’ve seen this in the crypto world: projects announce a dozen partnership LOIs to create FOMO, but only a fraction convert. The high number here is designed to signal market dominance, not necessarily to secure capacity. It’s a move to intimidate competitors and reassure investors. But for the crypto community, this should be a call to action. If we believe in decentralized AI, we need to fund the infrastructure that supports it. We need to support projects that are building open, community-owned compute networks.
Culture on-chain, heart on-screen. The takeaway is not a summary; it’s a vision. The current AI infrastructure race is a mirror of the early crypto narrative: a few powerful entities building silos. But the blockchain community has the tools to build a different path. We can use token incentives to fund distributed GPU networks, create DAOs that own data centers collectively, and implement transparent governance for AI compute allocation. The question is: will we? Or will we watch as the same centralization patterns that plagued traditional finance and social media repeat in AI?
The market is sideways, and patience is thin. But this is the moment to position. Not for a quick trade, but for a long-term bet on the values we hold. The next time you see a headline about a company buying 70-80 data centers, ask yourself: who controls the compute, and who controls the future? The answer will determine whether the blockchain revolution becomes a footnote in the history of centralized AI, or the foundation of a truly decentralized intelligence.
