For about two years, nearly every crypto conversation has circled the same question: is the winter finally over? Lately a different idea has been creeping into the headlines. The thing that pushes crypto into the mainstream might not be banks, ETFs or new laws. It might be AI.
That’s the case BlackRock makes in The Machine-Native Economy, a white paper it put out in September 2026. The world’s biggest asset manager reckons that as AI agents start doing real work online, they’ll need a way to pay for stuff, and blockchains may handle that better than the banking system we’ve got.
Headlines like that send plenty of people off to look up how to buy cryptocurrency. Fair enough, but buying coins is only one way to get involved, and it helps to know what’s actually being claimed first. So here’s what the paper says, what it doesn’t, and where the doubters have a point.
Banks weren’t built for robots
Let’s make it concrete. Leo, a made-up founder, runs a small data startup. Overnight, an AI agent keeps his systems ticking along. At 3 a.m. it decides it needs more cloud computing, a few thousand API calls and a fresh dataset from a supplier. Each of those costs a few cents.
Now try doing that with a normal bank. Opening an account needs a person. So do identity checks, passwords and payment approvals. Card networks can be automated a bit, but fees swallow tiny payments, and lots of steps still assume a human is awake and watching. The agent can’t stroll into a branch, and it can’t sit around for three days waiting for a wire.
That’s the gap BlackRock thinks crypto can plug: money that software can hold, send and receive by itself, whenever it likes.
Stablecoins get there first
The paper expects stablecoins to be the early winners. They’re crypto tokens built to track a real-world asset, usually the US dollar, so they don’t bounce around like Bitcoin. They move almost instantly, cross borders easily and never close for the night.
For Leo’s agent, that’s pretty much perfect. One word of caution, though: “stable” doesn’t mean “can’t go wrong.” A stablecoin is only as good as the reserves behind it, and even the big ones have wobbled. USDC briefly slipped well below a dollar in March 2023, when Silicon Valley Bank collapsed.
The paper highlights a couple of new tools that let agents pay without a human clicking “approve” every time. One is Coinbase’s x402, which lets an agent pay for things like API calls straight over the web. Another is the Machine Payments Protocol, built by Stripe and Tempo, which does a similar job and can settle in stablecoins or regular money. Coinbase boss Brian Armstrong has gone even further in public, saying AI agents will one day make more transactions than all humans put together. Whether or not he’s right, the basic logic holds: agents need money they can program, and crypto has it.
So, how big is this already?
Bigger than you might think. BlackRock says adjusted stablecoin transaction volume passed $11 trillion in 2025. That’s in the same ballpark as what Visa and Mastercard each handle in a year, though still a long way short of the roughly $93 trillion that flowed through the US ACH system. It’s growing fast, too. By the paper’s numbers, adjusted volume grew about 80% a year from 2020 to 2025, while ACH grew around 8.5%.
Unsurprisingly, regulators are paying attention. Stablecoin rules have tightened in several places, Europe’s MiCA framework among them, and supervisors are beginning to take legal initiative on the bigger tech shift as well. In September 2026, ESMA, the EU’s markets watchdog, said AI and tokenisation will be supervisory priorities from 2027. It plans to map where firms use them with customers and check their governance and data.
What a world full of agents might look like
If agents really do become everyday economic players, blockchains offer more than just a way to pay. They’re good at microtransactions, the kind of tiny payments that make no sense on a card network. They can also follow rules automatically: release this payment only when the job’s done, stop spending once the budget’s gone.
And it’s not just stablecoins. Tokenised real-world assets, meaning shares, bonds or funds that live on a blockchain, could let agents hold and move value at any hour. Stablecoins can even be used as collateral in automated deals, with no bank sign-off required.
We’re already seeing a taste of this in retail, where AI agents are taking over holiday shopping chores people used to do themselves. The twist here is that the agent would pay as well as shop.
Computing power could become a market of its own
This is the paper’s boldest idea. Today, someone at a company spends hours comparing cloud providers, GPUs and server prices. An agent could do it non-stop: watch the market, compare deals, rent exactly the capacity a job needs, then hand it back.
BlackRock suggests that rights to computing power could one day become standardised tokens that agents trade, finance or even borrow against. To its credit, the paper admits this is a long-term bet. Liquid markets for standardised compute contracts don’t exist yet.
Hang on, is anyone actually using this?
Not much, yet. The roads are being built faster than the traffic’s arriving. Back in March 2026, CoinDesk reported that x402 was handling about $28,000 a day, spread over roughly 131,000 transactions. Analytics firm Artemis reckoned around half of that was self-dealing or wash trading. One of its analysts called the whole agent-payments boom “still mostly a mirage.” Hardly a gold rush.
There are other loose ends too. Banks are trialling tokenised deposits that could go head to head with stablecoins. Any business letting an agent spend real money will need spending caps, permissions, fraud checks and a human sign-off for big amounts, the kind of guardrails covered in AI agent security. And the rules change a lot depending on where you are.
Where that leaves us
BlackRock makes a serious argument that AI agents could become one of crypto’s biggest sources of demand, with stablecoins out in front and tokenised computing power as the long shot. The stablecoin numbers are real and climbing. But machine-to-machine payments are still tiny, and a lot of this story is a prediction about customers who haven’t really shown up yet.
If you’re tempted to dip a toe in, keep your head. Crypto prices can lurch wildly, the rules depend on where you live, and exchanges aren’t available or regulated the same way everywhere. Do your homework, never put in more than you can afford to lose, and treat big forecasts, this one included, as food for thought rather than a promise. This article is for information only and isn’t financial advice.
Related: AI Economy 2026: How AI Is Reshaping Factories and Power
