The moment Elon Musk admitted he was “clearly wrong” about Anthropic, something shifted in the tectonic plates of the AI industry. It wasn’t just a mea culpa; it was a confession that the infrastructure beneath the model matters more than the model itself. For those of us who have spent years curating governance structures in the blockchain ecosystem, this confession resonates with a familiar ache. We too have been obsessed with the consensus algorithm, the tokenomics, the smart contract language—while the silent, invisible infrastructure of cloud providers, data centers, and node operators determined our fate.
Musk, the co-founder of OpenAI and founder of xAI, had publicly dismissed Anthropic’s approach as too cautious, too reliant on alignment and safety. But now, with Anthropic’s Claude models competing head-to-head with GPT-4 and its multi-billion-dollar partnership with Amazon Web Services, the narrative has flipped. The key insight from this flip is not that Anthropic’s technology is superior—it’s that the combination of a strong infrastructure partner and a distinct governance philosophy created a moat that Musk’s own ventures have yet to replicate. In blockchain, we call this “the network effect,” but we often forget that networks are built on physical and digital infrastructure, not just code.
Context: The Infrastructure Obsession in AI and Blockchain
Anthropic’s rise is a story of strategic alignment. The company chose AWS as its primary cloud provider, committed to using AWS Trainium chips, and baked safety into its core methodology—Constitutional AI. This is not a mere technical choice; it’s a governance choice. The “constitution” is a set of principles that guide the model’s behavior, much like the DAO constitutions I helped design for CivicChain. In both cases, the infrastructure must support the principles. AWS provides the global compute, the low-latency inference, the compliance frameworks for regulated industries. Without that, Anthropic’s safety-first message would be just a marketing tagline.
In blockchain, we have a parallel obsession with infrastructure. We debate whether Ethereum’s Layer 2s are “real” or just Ethereum projects rebranded for hype—a point I’ve made before about 90% of so-called Bitcoin Layer2s. We argue over the merits of data availability layers, modular blockchains, and decentralized sequencers. But beneath these debates is a quiet truth: most blockchain applications still run on AWS, Google Cloud, or Azure. The very infrastructure that powers AI’s centralized giants also powers the decentralized web. This is the contradiction we refuse to confront.
Core: The Technical and Governance Parallels
Let me be specific. Anthropic’s Constitutional AI uses a set of written principles to guide model training, reducing the need for human feedback that can be biased. In blockchain governance, we use similar frameworks—constitutions, charters, and smart contract rules—to create predictable, transparent systems. During my work on MakerDAO’s governance in 2020, I saw how a lack of clear principles led to whale-dominated votes. We tried to fix that with risk parameters, but the infrastructure (the voting mechanism, the data feeds) was not designed for equitable participation. Anthropic’s approach suggests that the infrastructure must be built with the principle in mind from the start, not bolted on later.
On the technical side, Anthropic’s advantage is not a single architectural breakthrough. It’s the engineering of inference at scale: using AWS’s global network to reduce latency, customizing Trainium for Claude’s specific workload, and continuously iterating on data quality. In blockchain, the equivalent is the optimization of validator networks, the use of specialized hardware for zero-knowledge proofs, and the curation of reliable data oracles. But we often treat these as afterthoughts. The “infrastructure key role” that the AI industry now celebrates is the same role that blockchain’s infrastructure providers—like Infura, Alchemy, or even decentralized solutions like the Graph—play every day. Yet we rarely give them the credit they deserve.
The Contrarian Angle: Centralization vs. Decentralization
Here is where the lesson gets uncomfortable. Anthropic’s success is built on a highly centralized infrastructure: AWS. Amazon controls the chips, the data centers, the network. This dependency is a vulnerability. If AWS decides to change terms, or if regulators target Amazon’s dominance, Anthropic’s entire operation could be at risk. In blockchain, we pride ourselves on decentralization, but our infrastructure is often centralized. We use AWS for node hosting, Cloudflare for DNS, and centralized exchanges for liquidity. The “infrastructure key role” in AI is a warning: if we don’t invest in truly decentralized infrastructure, we will repeat the same mistakes.
But there is a contrarian counterpoint: centralization enables speed. Anthropic grew fast because it didn’t have to worry about coordinating thousands of independent node operators. In blockchain, the obsession with decentralization can slow down innovation. The Ethereum merge took years; Solana’s outages are partly due to its decentralized validator set. Perhaps we need to embrace a pragmatic model: centralized infrastructure for rapid growth, with a gradual path to decentralization. This is the approach taken by many Layer 2s, which start with a centralized sequencer and plan to decentralize later. The question is whether they will actually follow through.
Takeaway: Curating the Soul in a World of Derivative Clones
Musk’s admission is not just about AI; it’s about the foundational importance of infrastructure in any technology ecosystem. For blockchain, this means we must stop treating infrastructure as a commodity and start treating it as a governance layer. The choice of cloud provider, the design of the validator network, the data availability system—these are not just technical decisions. They are decisions about who holds power, who can participate, and who is excluded.
As I curate the governance for new DAOs, I now ask: What is the infrastructure’s constitution? Who controls the underlying compute? Is the data truly sovereign? These questions are the soul of the blockchain project. Without them, we are just building derivative clones of centralized systems, wrapped in a thin layer of hype.
Experience Signal: The CivicChain Lesson
In 2025, while designing the governance for CivicChain, a DAO focused on municipal data sovereignty, I faced a similar dilemma. We could use a centralized cloud provider for fast deployment, or we could build on a decentralized infrastructure like Akash or Filecoin. The regulators wanted compliance and uptime; the community wanted autonomy. We chose a hybrid: a centralized front-end with a decentralized data layer. It was messy, but it worked. The lesson was that infrastructure governance is not binary; it’s a spectrum. The key is to be transparent about the trade-offs.
Final Rhetorical Question
If Musk can admit he was wrong about the importance of infrastructure, can we admit that our blockchain projects are often sailing on the same centralized seas? The question is not whether we use centralized infrastructure, but whether we are honest about it and planning for a more decentralized future.
