Infinity Raised $15M From the Brains Behind AI. Here's Why Crypto Should Pay Attention.

We are told that AI and crypto are on a collision course. One centralizes power into the hands of a few trillion-dollar compute clusters; the other distributes trust across millions of nodes. The narrative is neat, binary, and wrong. Because the smartest money in AI—researchers from OpenAI and Anthropic—just poured $15 million into a startup called Infinity, and its pitch sounds suspiciously like the blockchain thesis I’ve been writing about for years. Infinity is not a crypto company. It’s an “AI infrastructure” play, a term so broad it could mean anything from a GPU leasing marketplace to a training orchestration tool. But the lack of detail is precisely the signal. The investors—Touring Capital, plus individual researchers who built the models that define this decade—are betting on a modular, software-defined layer that sits underneath the AI stack. They are betting on composability. On neutrality. On a platform that doesn’t pick winners. This is the same bet every L2 rollup makes. Let me step back. When I dropped out of my macroeconomics class in 2017 to organize crypto philosophy meetups in Capitol Hill, I learned that the most powerful technologies are invisible. The Ethereum I fell in love with wasn’t a coin—it was a substrate for coordination. The same is true for AI infrastructure today. The biggest bottleneck isn’t better models; it’s the plumbing between models, data, and users. Infinity, from the tiny amount of public information, is building a piece of that plumbing. And the fact that it’s raising from AI insiders rather than crypto-native VCs tells me something important: the crypto thesis is being validated from outside our echo chamber. Here’s the core of my argument. The $100 million valuation on a seed round with no product, no revenue, and no technical whitepaper isn’t just a bet on the team. It’s a bet on a specific architectural philosophy: infrastructure should be abstracted from the hardware, programmable, and trust-minimized. Sound familiar? That’s the exact philosophy behind Ethereum’s rollup-centric roadmap. Decentralization is a verb, not a noun. It is the process of breaking monolithic structures into modular, composable pieces. Infinity, whether they call it that or not, is applying that verb to AI compute. I’ve spent the last twelve years watching markets and seven years deep in protocol design. I’ve seen DeFi Summer turn smart contracts into slot machines. I’ve seen Bitcoin L2s that copy-paste Ethereum whitepapers and slap on a ticker. I’ve watched 90% of “crypto AI” projects raise millions for a tokenized GPU rental idea that offers zero technical innovation over AWS. But Infinity is different. It’s not raising from crypto degens; it’s raising from the people who wrote the code for GPT. That’s a signal with far more information density than any whitepaper. Now, the contrarian angle. The crypto community loves to claim that decentralized compute networks—Akash, Render, io.net—will eat the cloud. But here’s the uncomfortable truth: market makers won’t leave quotes on-chain because latency matters. The same logic applies to AI training. You cannot run a 10-day training run on a decentralized network of heterogeneous GPUs with unpredictable reliability. Infinity’s investors understand this. They are not building a decentralized compute market. They are building a software layer that makes centralized compute easier to use, cheaper, and more composable. That’s a smarter bet than any on-chain compute market. And it’s a bet that aligns with the values of sovereignty and flexibility that crypto champions, but without the dogma. Let me give you a concrete example from my own work. In early 2020, during DeFi Summer, I forked three yield strategies and lost 40% of my capital to impermanent loss. But I gained a lesson: infrastructure without abstraction is a trap. The protocols that won—Uniswap, Aave—were the ones that hid complexity. They made the underlying blockchain invisible. Infinity is trying to do the same for AI. If they succeed, they will enable a thousand experiments that today are too expensive or too complex. That’s a bigger unlock for decentralization than any token model. Of course, the risks are real. The AI infrastructure space is crowded. Lambda, Modal, Replicate, Anyscale, and every cloud hyperscaler are already fighting for the same developers. Infinity needs to ship a product that is 10x better at something, or it will disappear. But the fact that OpenAI and Anthropic researchers put their personal capital—and their personal reputation—on the line suggests they see a gap that existing players ignore. What does this mean for your portfolio? Nothing directly. Infinity is not going to launch a token tomorrow. But it means that the smartest minds in AI are thinking about infrastructure in terms that mirror the crypto ethos: modularity, composability, programmability. The same forces that drove the rise of Layer 2s will drive the next wave of AI adoption. And the winners will be the infrastructure layers that become invisible—just like the blockchain. I’ll leave you with this. Every bull market euphoria masks technical flaws. The current AI mania is no different. But beneath the hype, real infrastructure is being built. Infinity might succeed or fail, but the pattern is clear: the future of AI looks a lot like the decentralized web we’ve been preaching. Decentralization is a verb, not a noun. Infinity is just applying it to a different stack.