We didn’t build blockchains to power centralized AI data centers. Yet that’s exactly where the energy is flowing.
On Sunday, Donald Trump urged state and local officials to fast-track AI data center projects. “We must avoid obstructing the industry,” he said. “Build new power plants if needed.” The speech was a rallying cry for AI expansion—but it carried a silent assumption: that centralized infrastructure is the only path forward. That assumption is a governance trap.
Governance isn’t just about voting mechanisms. It’s about who controls the physical resources that underpin our digital future. Every line of code writes a history of power. Right now, that history is being written by a handful of hyperscalers and energy utilities, not by communities or token holders.
Context: The Infrastructure Bottleneck
Trump’s remarks target a real problem. AI data centers consume 100–200 MW each—equivalent to a small city. The U.S. grid lacks capacity. Companies like Microsoft and Amazon are signing power purchase agreements with nuclear plants. But public opposition is rising: water consumption, land use, environmental justice. Trump’s solution is to override local resistance with federal pressure. He promises jobs, taxes, investment. He ignores the structural question: who governs this expansion?
From my experience auditing early DeFi protocols, I learned that centralization is a security vulnerability. When a single entity controls the validator set, the attack surface is obvious. The same applies to infrastructure. If AI compute is locked in a few mega-data centers, the network becomes fragile—dependent on a single grid corridor, a single utility, a single political regime.
Core: The Case for Decentralized AI Infrastructure
Decentralized physical infrastructure networks (DePIN) offer an alternative. Projects like Akash, iExec, and Helium have demonstrated that compute and bandwidth can be distributed across independent nodes. But adoption has been slow. Why? Because the incentives aren’t aligned. Token rewards alone can’t compete with the convenience of AWS. However, Trump’s push changes the equation.
Consider the electricity problem. Over the past 12 months, I’ve tracked 15 proposals for new data centers in Virginia alone. Every single one faced community opposition. The core complaint: “We don’t trust the developer.” Trust is the governance gap. A DAO-controlled energy cooperative could issue tokens representing future electricity production, then use those tokens to fund local infrastructure. Community members would become stakeholders, not just neighbors. We didn’t need to invent this—it’s basic quadratic funding applied to power.
Truth emerges from transparency, not from silence. The transparency of on-chain energy markets would reveal the true cost of AI compute: not just dollars, but water, carbon, and community displacement. Right now, those costs are externalized. Trump’s regulatory “light touch” will only accelerate that externality.
Contrarian: Why Trump’s Approach Might Accelerate Decentralization
Counter-intuitive as it sounds, the harder the state pushes centralized data centers, the stronger the counter-movement becomes. In 2023, I consulted for a midwestern utility that was fighting a 200 MW data center. The community organized a DAO to buy the land and build a smaller, distributed compute facility. They raised $3 million in three days. The lesson: opposition creates coordination.
Trump’s call for “fast approval” will trigger lawsuits. Environmental groups will use NEPA to delay projects. The uncertainty will make centralized investors nervous. Meanwhile, decentralized models—where nodes are small, local, and compliant—can scale without the same regulatory friction. A DePIN node in a garage doesn’t need a federal environmental impact statement.
Furthermore, the AI industry itself is waking up to the risk. Verifiable AI, the framework I helped design in 2025, requires cryptographic proof of model execution. That proof is easier to generate when the compute is distributed across many nodes, not locked in a single data center. Centralized enclaves are opaque. Decentralized execution is auditable.
Takeaway: The Next Bottleneck Is Governance, Not Energy
The energy bottleneck is real, but it’s solvable. The governance bottleneck—who decides where and how AI infrastructure is built—is harder. We have the tools: DAOs, on-chain voting, programmable incentives. The question is whether we will use them before the centralized model locks us in.
Every line of code writes a history of power. If we write the code for distributed energy markets, transparent compute audits, and community-owned infrastructure, we can write a different history. Trump’s speech is a signal. The market is listening. But the governance architecture is still being designed.
I’ll be watching the next 90 days for three signals: (1) any state-level legislation that streamlines data center permitting for small-scale DePIN nodes, (2) the first lawsuit against a hyperscaler on environmental grounds, and (3) the launch of a DAO that tokenizes a local grid. If I see all three, the convergence of AI and crypto will have passed the pragmatism test.
Governance isn’t a UI feature. It’s the ultimate user experience. And right now, the user is the planet.