Crypto infrastructure builders are increasingly looking beyond Web3 and toward AI. In “Why crypto is not that fun anymore,” Kydo argues that unless a team is working on stablecoins or is deeply committed to financial markets, much of the excitement has drained out of day-to-day work in crypto.
The piece says that in the last cycle, building products for crypto developers was still workable. Metrics, brand partnerships, and market energy often mattered more than revenue. That model, according to the article, began to break down roughly 18 months ago, as a wider share of the industry came to accept the idea that crypto is mainly useful for financial applications.
Outside DeFi, the addressable market is described as narrow
Kydo says the crypto-native applications that consistently make money on-chain still center on trading and lending. Many builders, though, are not focused on DeFi. They have spent years building infrastructure intended to support use cases beyond finance.
That is where the frustration sets in. The article estimates that the total addressable market for this segment is only about $200 million to $300 million per year, spread across hundreds or thousands of institutions. Even the strongest businesses in that category, it says, have reached only tens of millions of dollars in revenue. In the author’s view, that is close to the ceiling after years of market development.
With that constraint, teams are left with a short list of options: sell infrastructure into traditional financial institutions, or move into AI. The article says many people choosing AI are doing so from a clear reading of their own strengths, not because they have lost touch with crypto, but because they do not want to spend years in slow and complex enterprise sales aimed at traditional finance.
The crypto-AI overlap is active, but the routes are limited
The article outlines several paths already on the table. One is the familiar formula of a conventional product, AI features, and a token on top. Kydo dismisses that route as another version of financial wrapping, where the token has no real function. Another is decentralized AI infrastructure, usually built around privacy, security, and verifiability. The author treats that as a long campaign with weak short-term feedback and little immediate revenue.
A third option is stablecoin infrastructure for AI agents. Kydo says that path is commercially interesting, but also brutally competitive. Circle, Stripe, and major stablecoin players are all targeting the same space, leaving startups with a hard fight unless they have a clear wedge.
Why the author is focused on agent companies
Kydo’s core argument is that crypto works best as a mechanism for instant, global, programmable capital formation, while AI has sharply reduced the cost of building products that can actually grow. Work that once required 50 people, the article says, can now sometimes be done by one person.
From that premise, the author frames the real opportunity as using crypto’s capital engine to accelerate products that are already showing growth, revenue, users, and feedback loops. The proposed answer is “Agent Companies.” To make that work, Kydo says tokens first need to be able to “own” things, adding that this has been the team’s focus for the past five years and is now being turned into a product.