Cathie Wood Shifts Focus: Watch AI Agents’ Spending, Not Developers

Cathie Wood changed her advice. The ARK Invest CEO, speaking at Robinhood’s Summit in Houston, told investors to stop watching developers and start watching where AI agents spend money. That’s a pretty significant shift from someone who built her reputation betting on the builders.

Wood’s argument is straightforward: AI agents don’t just answer questions or generate text anymore. They execute tasks. They make purchases. They interact with financial systems on behalf of users — and they’re doing it without a human approving every single step. So if you want to know which payment rails are winning, which stablecoins are actually getting used, and which blockchain networks are seeing real adoption, watch the agents. Their spending patterns won’t lie.

Why Blockchain Could Win the AI Agent Race

Joseph Chalom, co-CEO of SharpLink, sees a real danger in what happens if this AI-agent economy gets captured by a handful of large companies. His worry is basically that a few dominant platforms end up controlling how agents spend, where they transact, and which financial providers they’re allowed to use. That’s a walled garden problem, and it’s not hypothetical — it’s already how a lot of digital ecosystems work.

Chalom wants something different. He wants users to control their agents’ spending limits directly. He wants people to be able to switch financial providers without being locked in. And he thinks open blockchains — Ethereum specifically — are probably the best infrastructure for that kind of universal financial network. The idea is that agents could transact across providers seamlessly, without centralized gatekeepers deciding what’s allowed. Think of it like porting a mobile phone number from one carrier to another. The agent keeps its financial identity, its permissions, its transaction history — and moves freely.

That’s the vision, anyway. Whether it plays out that cleanly is unclear.

BlackRock, Coinbase, and the Stablecoin Angle

BlackRock put out a paper on this. The asset manager sees AI agents as potentially needing to pay for services independently — API calls, data access, computing power — without a human approving each transaction. Round-the-clock, micro-payment-scale, machine-to-machine. Traditional payment systems weren’t built for that. Bank transfers don’t settle at 3 a.m. on a Sunday. Credit card rails weren’t designed for transactions that cost fractions of a cent.

Stablecoins and blockchains fit that use case pretty well, at least on paper. They run continuously. They can process tiny payments. They don’t require a human in the loop for every step. BlackRock’s research basically frames AI agents as a new class of economic actor — one that needs financial infrastructure built specifically for autonomous operation.

See also: Trump Coin Investors Could Lose $3.2 Billion as Controversial Event Approaches

Coinbase is already moving on this. CEO Brian Armstrong mentioned “Grok” as a notable client for agentic traders on the platform. No further details were given on the specifics of that relationship, but the fact that Armstrong named it publicly says something about how seriously Coinbase is taking the agentic trading space.

And Coinbase isn’t alone. Stripe, Visa, Google, and OpenAI are all reportedly developing ways for agents to conduct purchases. That’s a crowded field fast. The question for crypto investors is whether stablecoins and blockchain rails win that race or whether traditional payment infrastructure adapts fast enough to stay relevant.

Control, Transparency, and the Accountability Problem

Chalom keeps coming back to user control. It’s not enough for AI agents to operate autonomously — they need to operate within limits that humans set. Spending caps. Access to transaction histories. The ability to revoke permissions at any time. Without that, you don’t have agentic finance, you just have financial activity that no one can audit or stop.

That’s a harder problem than it sounds. The more autonomous these agents become, the more difficult it gets to maintain meaningful oversight. And the interoperability piece adds another layer of complexity — if agents are supposed to move between providers freely, carrying their financial identities with them, someone has to build the standards that make that possible across different platforms. That’s not a solved problem.

For now, the competitive pressure is real. Tech giants are in this. Payment networks are in this. Crypto infrastructure companies are in this. Wood’s point is that you don’t need to predict who wins — you just need to watch the money move. AI agents will show you where adoption is actually happening versus where it’s just being announced.

Armstrong put Grok on the record. BlackRock published the research. Chalom laid out the architecture he thinks should win. Wood told a room full of investors in Houston to pay attention.

The agents are already spending.