When Microsoft announced its latest $50 billion AI infrastructure fund last week, the market cheered. I saw something else: a geometry of silence. Geometry remembers what markets forget. A report surfaced, predicting that Big Tech will pour $735 billion into AI data centers by 2026. The numbers are dizzying. But as I traced the lines of this investment map, I felt a familiar pull—the same quiet I felt during the 2022 bear market, when the loudest narratives were the emptiest. This is not a story about AI. It is a story about the geometry of trust, and how we are about to make the same mistake again.
Context: The Infrastructure Mirage
Let me step back. The report—which I will not name, because the name is irrelevant—claims that the world’s largest technology companies will collectively spend $735 billion on AI data centers by 2026. This is not a small number. It is roughly the GDP of Switzerland. The narrative is seductive: AI is the future, data centers are the factories, and the crypto industry—especially DePIN—will be carried along by this wave. The logic seems sound: more AI compute demand means more need for decentralized compute networks like Akash, Render, and Filecoin. But I have been here before.
In 2017, I watched the ICO frenzy consume Ethereum’s block space. The code was beautiful—Golem’s Sybil resistance mechanisms, the mathematical elegance of early smart contracts. But the narrative was hollow. Everyone talked about “decentralized compute” as if it were a foregone conclusion. The reality was that most projects never shipped. I published a series of visual essays on Zhihu, trying to explain the geometry of trust—how decentralization is not a toggle switch but a delicate balance of incentives and participation. Those essays gathered 50,000 followers among math and philosophy enthusiasts. But the market didn’t care. It cared about the narrative.
Now, in 2026, I am 38 years old. I have spent the last decade watching the industry evolve. I have audited DAO governance tokens and found 12 critical centralization flaws. I have written a report on the ethical price of stability. And I have come to believe that the greatest threat to decentralization is not regulation or adoption—it is the illusion of progress. The $735 billion AI data center investment is the latest illusion.
Core: The Geometry of Fragmentation
Let me be clear: I am not against AI. I am not against data centers. What I am against is the silent fragmentation of our digital commons. The report claims that these investments will “change the digital asset landscape.” But what does that mean? From my perspective, it means one thing: liquidity is being sliced, not scaled. We have seen this before in Layer 2s. Dozens of rollups, all competing for the same small user base. This is not scaling—it is slicing already-scarce liquidity into fragments. The same is happening with AI infrastructure.
Consider the geometry. A massive data center is a point source of compute power. It is centralized by design. It relies on a single owner, a single power grid, a single location. The geometry of resilience, on the other hand, is a distributed network—many small nodes, each contributing a fraction of the total. DePIN projects like Akash and Render are trying to build this. But the $735 billion narrative is pulling the opposite direction. It is telling the market that the future belongs to big, centralized, capital-intensive infrastructure. And the market is listening.
I remember DeFi Summer in 2020. Compound and Uniswap felt like organic ecosystems. The protocols stacked like LEGO bricks, creating liquidity pools that breathed like natural systems. I co-authored a whitepaper on “Liquidity as a Public Good,” arguing that DeFi was not just finance but a new social contract. That social contract was built on permissionless composability. Now, in the AI era, the composability is being replaced by vertical integration. The same Big Tech companies building these data centers are also building their own AI models, their own cloud services, and their own tokenization platforms. They are not building for the commons. They are building for their own walls.
And the market is cheering. I see the FOMO everywhere. New projects launch every week, claiming to be the “AI-native blockchain” or the “decentralized GPU marketplace.” But when I audit their code, I find the same patterns: over-reliance on a single sequencer, admin keys that can freeze funds, and tokenomics that reward insiders more than users. The geometry of trust is broken. DeFi breathes; don't let it suffocate.
Contrarian: The Silent Warning
Here is the contrarian angle that no one wants to hear: the $735 billion investment is not a tailwind for DePIN—it is a headwind. The massive capital deployment will create a new kind of centralization risk: infrastructure dependency. If Web3 projects rely on AWS or Azure for their compute, they are not decentralized. They are tenants. And when the rent is due, the landlord can change the terms. I have seen this happen in the 2022 bear market, when many projects quietly migrated to centralized cloud providers to cut costs, only to face censorship requests later.
Silence is the loudest warning. The report does not mention the word “decentralization.” It does not discuss the risk of a single point of failure. It does not ask who controls the data centers, or what happens when a government decides to shut one down. The market is too busy projecting a $10 trillion market cap for AI tokens to notice that the emperor has no clothes.
Let me give you a specific example. I recently audited a governance token for a DAO that claimed to be building a decentralized AI training network. The token had a staking mechanism that required users to delegate to a “validator” who was actually a single entity running nodes on Google Cloud. The code was clean, but the architecture was a house of cards. When I pointed this out to the team, they said, “But it’s fine—Google Cloud is reliable.” That is the geometry of trust collapsing. They had forgotten that the whole point of blockchain is to eliminate reliance on third parties.
This is the same pattern I saw in 2024, when I collaborated with a Beijing-based fintech lab on a report titled “The Ethical Price of Stability.” We used game theory to show that institutional entry, while stabilizing prices, also eroded the core values of decentralization. The same is happening now. The $735 billion is the institutional entry of AI infrastructure. It will make the system more efficient, but less resilient. It will create more liquidity, but for fewer players. It will produce more compute, but under more centralized control.
Takeaway: Prune the Dead Branches, Save the Tree
So where do we go from here? I am not a pessimist. I still believe that blockchain can verify human authenticity in an age of synthetic media. I am exploring the convergence of AI and zero-knowledge proofs, teaching users how to protect their digital identity against algorithmic overreach. But I am also a realist. The path forward requires pruning the dead branches.
Prune the dead branches, save the tree. That means being skeptical of narratives that benefit the incumbents more than the community. It means investing in projects that are truly permissionless, not just in name. It means recognizing that the $735 billion is not a signal to buy every AI token, but a warning to look deeper at the geometry of control.
I will end with a question that I have been asking myself for a decade: When the next bear market comes, and the AI data centers are half-empty, who will still be building? The answer, I suspect, will be the same as always: the quiet ones, the ones who care about the geometry of trust, the ones who remember that code is cold, but community is warm. And that is the only investment that matters.