And most people are still calling it a bubble.
Think back to 1997.
The internet existed. People were using it. But if you told your uncle that this thing — this slow, jittery web of text and dial-up screeches — would eventually replace his bank, his bookstore, his entire way of getting news, he would have laughed you out of the kitchen.
It wasn’t that he was stupid. It’s that the infrastructure wasn’t there yet. The pipes were being laid underground while everyone argued about the wallpaper.
That’s exactly where we are right now with AI and crypto.
The Pipes Are Going In
Here’s the thing about transformative technology: it doesn’t announce itself. It shows up looking messy, overhyped, and half-finished. Then one day the pieces connect, and suddenly it’s everywhere and everyone acts like they saw it coming.
The convergence happening between artificial intelligence and decentralized finance right now is one of those moments. It’s not obvious yet. It’s still in the “arguing about the wallpaper” phase. But underneath the noise, something structural is shifting.
Let me explain what I mean.
Money, at its core, is information. It’s a shared story we tell each other about value. The financial system we have today — banks, clearinghouses, brokerages, wire transfers that take three days — was built to move that information around safely in a world of fax machines and paper ledgers.
The internet made information nearly free to transmit. But it didn’t fundamentally rewire money. You can email a photo to someone in Tokyo instantly. You still can’t send them $47 without a bank deciding if that’s okay.
Crypto was the first serious attempt to fix that. And AI is the engine that could finally make it usable at scale.
What’s Actually Happening
There’s a useful frame for what’s unfolding: two worlds in tension. Call them the Tower and the Square.
The Tower is what’s happening with AI right now. Microsoft, Google, Amazon, Meta — the hyperscalers — are collectively spending north of $467 billion on AI infrastructure this year. They’re building centralized, vertically integrated systems. Fast. Powerful. Closed. If you want access to the best AI, you go through them.
The Square is what crypto has always been about. Open networks. No gatekeepers. Value flowing directly between participants without anyone in the middle deciding who’s allowed in.
The question isn’t which side wins. It’s what gets built in the seam between them.
AI agents — software that can reason, plan, and take action on your behalf — are getting good fast, and they’re starting to interact with financial systems. Right now this looks modest: chatbots that analyze portfolios, tools that draft trade theses. But zoom out a year or two, and the trajectory points somewhere more significant — autonomous agents that can move money, execute contracts, and navigate financial markets without a human clicking “confirm” at every step.
That only works if the financial rails underneath are programmable. Which is exactly what blockchain-based finance is designed to be.
Think of it like this: AI is the driver, crypto is the road. Neither works great alone. AI without programmable money is a very smart assistant who can’t do your banking. Crypto without AI is a highway with no cars — impressive infrastructure, weirdly empty.
There’s also a physics to it that matters. AI is stochastic — probabilistic, generative, messy at the edges. Crypto is deterministic — precise, trustless, rules-based. They don’t just coexist. They correct for each other’s weaknesses in ways that start to look like something genuinely new.
Why This Is Actually Optimistic
The cynical read of this is: “great, now robots will move money around even faster and make rich people richer.”
But here’s the more interesting version.
The traditional financial system has enormous gatekeeping built in. Getting a loan, accessing capital markets, sending money internationally — these things require institutions to trust you, and institutions are slow, biased, and expensive to deal with.
A system where AI can assess creditworthiness from real behavior, and crypto can execute and settle a loan without a bank intermediary, isn’t just faster. It’s more accessible. The billion people globally who are “unbanked” aren’t unbanked because they’re bad with money. They’re unbanked because the infrastructure was never built for them.
That’s changing. Slowly. Then all at once.
What To Do With This
You don’t need to go all-in on anything. But you do need to be paying attention. Here’s where to start:
Read & Listen
- For the AI × finance × macro picture: Kyla Scanlon (newsletter and videos — gold standard for making economic shifts feel human), Dwarkesh Patel for long-form conversations with frontier AI builders, and Bg2 Pod for the investor and operator lens. Good Work by Dan Toomey for connecting tech narratives to the real economy.
- For crypto: Four Pillars and Delphi Digital are where serious research lives. Arthur Hayes’ Crypto Trader Digestis essential — even when you disagree, the macro framing is worth it. Ryan Watkins and Blockworks/0xResearchfor on-chain analysis without the noise.
Watch (Projects)
- Bittensor (TAO) — the most serious attempt to build a decentralized marketplace for machine intelligence. With dTAO live and the first halving done, subnets now compete on real utility. Follow Teng Yan’s Chain of Thought for the best ongoing coverage.
- Virtuals Protocol — not just AI agents as novelties, but as on-chain economic entities that can own assets, earn revenue, and operate independently. The question isn’t “is this agent cool?” — it’s “can this agent survive economically on its own?”
- Aethir and io.net on the compute layer. AI needs GPUs the way the internet needed servers. Aethir targets enterprise-grade deployments; io.net has aggregated 294,000+ GPUs from underutilized sources globally. Whoever wins the decentralized compute supply chain wins something important.
- Watch which L1s and L2s are positioning as the settlement layer for AI agent transactions. That race is underway.
Consider
- If you hold crypto, you need a thesis — not a price target. “Programmable money for the agentic economy” is a thesis. “It went up last cycle” is a prayer.
- If you’re not in crypto yet, the entry point isn’t a token. It’s understanding why programmable money matters when AI agents start transacting at scale. Get the mental model first.
- Seriously engage with the macro layer. Ray Dalio’s framework on debt cycles and Arthur Hayes’ view on dollar debasement aren’t just abstract — they’re the reason Bitcoin keeps getting a second look.
Invest
- Think in layers. The Delphi framework breaks the AI × crypto stack into three tiers: compute (where AI runs), protocols (how AI agents coordinate and transact), and applications (what they actually do). Roughly 88% of value in generative AI is still sitting at the infrastructure layer. That won’t hold indefinitely — but it tells you where the current moat is deepest.
- Compute bets: Aethir and io.net are early and volatile, but they’re betting on a bottleneck — GPU access — that doesn’t go away regardless of which models win. Gensyn and Prime Intellect are longer-horizon plays on decentralized training.
- Protocol bets: Bittensor (TAO) is the clearest expression of the thesis — machine intelligence as a market. Allora sits at the model-selection layer, governed by on-chain incentives.
- Application bets: Where most retail attention lands, and where most value hasn’t accrued yet. Be careful of tokens that are “AI-themed” without a real answer to: what does this agent do, and who pays it?
- Honest framing: most of these projects are pre-revenue and priced on narrative. That’s not a reason to avoid them — it’s a reason to be precise about position sizing. Small, deliberate exposure to infrastructure and protocol is how you stay in the game long enough to be right.
The money internet is being built right now. It looks a little chaotic. It looks a little early.
That’s usually when it’s worth paying attention.