There’s a number from a 2004 NBER paper that keeps showing up in how I think about crypto investing. The economist William Nordhaus studied US productivity data from 1948 to 2001 and found that innovators captured roughly 2.2% of the total surplus from their innovations. The other 97.8% passed to everyone else: consumers, competitors, and whoever sat closest to the friction point that didn’t commoditise.
The steam engine compressed manufacturing costs. The factory system kept the margin. Railroads reduced US freight costs by 95% over a century. Most railroad companies destroyed each other through competition and overbuilding. What actually got monetised was the land around the stations, the steel, and Standard Oil riding the rails better than anyone else. The internet gave us TCP/IP, HTTP, and SMTP as near public goods almost immediately. Google and Amazon didn’t build the internet. They built the interfaces that organised attention and commerce on top of it.
Crypto is running the same playbook. Just faster.
Aave keeps a 0.2% lending spread. Banks keep 3%. Solana and Monad move stablecoins for fractions of a penny. Visa charges 2%. The value creation is enormous and it’s real. But if the historical pattern holds, most of the infrastructure layer is becoming the new public good. The question isn’t whether the rails are useful. The question is who builds the factory on top.
This week gave us three signals that are hard to dismiss together.
Mastercard launched a Crypto Partner Program with more than 85 companies including Circle, Binance, Ripple, Gemini, PayPal, Solana, Monad, and MetaMask — focused specifically on cross border transfers, B2B settlement, and global payouts. Not tokenisation experiments. Payment pipes. Separately, they agreed to acquire BVNK, stablecoin payment infrastructure operating across 130+ countries, for up to $1.8 billion. A company that processes $350 billion in digital currency payment use cases annually doesn’t spend $1.8B on a bet. It spends it on infrastructure it can’t afford not to own.
Stablecoin market cap is now at $315 billion, up roughly 50% in a year. Adjusted transfer volume in 2025 ran to approximately $11 trillion. For context, Visa processed around $13 trillion in the same period. And exchange inflows have been net negative since January. The money is not going into speculation. It’s going into these pipes.
The third signal is the strangest and probably the most important. AI agents cannot open bank accounts. Banks require KYC verification. KYC requires a human. Agents don’t have one. When an agent needs to pay for cloud compute, data, or another service, it’s structurally blocked by TradFi infrastructure. A crypto wallet requires only a private key. No identity. Coinbase launched Agentic Wallets on its x402 protocol in February and had already processed over 50 million transactions by March. Software paying software. This is a demand vector that traditional finance literally cannot serve, and it has only just started.
The asset universe is quietly bifurcating in response. There are tokens that trade on fundamentals like actual cash flows, protocol earnings, defensible revenue multiples. Think Hyperliquid, Aave, Jupiter, Morpho. There are assets that trade like money (BTC and ETH) valued on scarcity, network effects, and macro positioning. And then there’s everything else, which is increasingly being repriced against the verticals that are demonstrably working. If stablecoins, payments infrastructure, and exchanges represent the categories with proven product market fit, then marginal capital in the rest of the space faces a harder question: why does this deserve allocation relative to what’s already generating real economic activity?
The optimistic read is that the technology is working. Monad has processed over 222 million transactions and attracted almost 2.5 million active users since its mainnet launch, with no reported downtime and Korean Treasury Bonds now live onchain via Etherfuse and Shinhan Securities. The SEC and CFTC jointly classified 16 cryptocurrencies as digital commodities. Indonesia’s registered crypto investor base now exceeds its entire stock market investor base. The infrastructure is not a whitepaper. It’s in production.
The bear market in prices and the bull market in infrastructure are happening simultaneously. That has happened before in every major technology cycle. It’s usually when the most interesting positions get taken.
Who captures the margin? The stablecoin issuer sitting at the fiat-onchain gateway. The frontend that controls order routing and user attention. The protocol with embedded distribution, not just the matching engine. The 97.8% rule doesn’t stop applying because the rails are programmable. But the people who figure out where the friction point is earliest tend to build the most durable things.