Crypto has spent the better part of the last 15 years asking ordinary people to put up with an absurd amount of hassle just to move money around. Memorize this 12-word phrase. Understand gas fees. Accept that your money is gone forever because you pasted the wrong address into a box.

But it has finally found an explanation for why it was built this way. Crypto was never really designed for people, the argument goes. It was meant for machines—the tireless bots that don’t care about ugly interfaces, lose seed phrases or need an 18-year-old Polymarket trader to explain the difference between Base, Polygon and Optimism.

Coinbase’s chief executive, Brian Armstrong, has become one of the loudest evangelists of this idea. “Very soon there are going to be more AI agents than humans making transactions,” he wrote on X earlier this month. “They can’t open a bank account, but they can own a crypto wallet.”

“We started to move to an AI-first mentality throughout the company," added Armstrong on a recent podcast.

What a convenient new pitch for the industry that has spent years promising to rewire finance but mostly succeeded at reinventing speculation. But it may also be the first one in years that feels intuitively plausible. For all its chaos, crypto offers something traditional finance still does not: the ability to move funds permissionlessly, near-instantly, globally, at any hour.

McKinsey projects that AI agents could mediate between $3 trillion and $5 trillion in consumer commerce by 2030—more than the current value of the entire crypto market, which sits at about $2.4 trillion.

“This changes a lot about how we think about the investment landscape and about building products,” says Matt Huang, managing partner at Paradigm, crypto’s largest venture capital firm. “You really have to think agent-first now and assume that most of your customers are going to be agents rather than people.”

Countless crypto firms, including Huang’s new payments-focused startup Tempo, are now racing to invent, or reinvent, themselves for this emerging class of users. Justin Sun, the billionaire founder of the Tron blockchain and a major investor in Trump’s crypto projects, is already calling it Web 4.0 (as if Web 3.0 was ever really built!).

Paradigm-backed MoonPay, which helps people—and now, increasingly, software—buy and sell crypto using ordinary payment methods, completely revamped its AI strategy after OpenClaw, the open-source AI assistant that can interact directly with a user’s files and applications, took off a few months ago.

“The bet that MoonPay is making is that we don’t need to double down on reinvesting in a beautiful UX (user experience) because agents become the interface,” says Kevin Arifin, the firm’s product lead.

That might be excellent news for anyone who still cannot, or simply will not, make themselves care about the finer points of crypto plumbing. You’ll just tell your AI what you want done—buy some bitcoin, find a lending service with decent rates, put the assets to work—and it will handle all.

Except none of this is happening at a meaningful scale yet.

Many of the crypto payments made by AI agents today flow through x402, an open standard developed by Coinbase that gives online service providers a way to charge them directly.

Until recently, even simple tasks like fetching a weather forecast or renting computing power required developers to sign up for services one by one, enter a credit card and generate an API key, a kind of password that lets software access another service. Build anything even mildly ambitious, and the setup easily turns into a mess of accounts, subscriptions and keys.

x402 offers a simpler pay-per-use model. When an agent requests a service, the server can respond with a price, and the agent can pay it automatically in crypto from a wallet assigned by its developer. That matters not just because it enables pay-per-use pricing, but because it begins to replace the API key sprawl. These days most companies have over 600 individual APIs.

“If you’ve set up OpenClaw, you might remember that it had you set up 10 API keys before you could even get started,” says Erik Reppel, creator of x402 and head of engineering at Coinbase Developer Platform. “With x402, your wallet becomes the universal API key that lets you access any x402-enabled service.”

For now, agents are used mostly by developers anyway. Since x402 launched in May 2025, AI assistants have made about 107 million transactions through the standard, totaling roughly $30 million in legitimate volume, according to data provider Artemis. Most are tiny—between 20 and 40 cents.

“It’s pretty apparent that we’re still early,” says Artemis’ analyst Lucas Shin. Transaction volume, he argues, is almost beside the point for now. The better tell is which ecosystems are actually building, and how many merchants are willing to sell through x402. That number is now around 3,900, including Amazon Web Services, blockchain development platform Alchemy and data provider Messari.

Crypto’s excitement over agentic commerce is understandable. “Pretty much any engineering team you look at, including ours, is using AI tools,” says Rishin Sharma, head of AI product and growth at the Solana Foundation. On his team, he says, everyone uses them, and AI is generating more than 70% of the code they write. Service providers that once built their businesses around traditional APIs, Sharma says, are beginning to ask a different question: not how to win the next hundred developers, but how to position themselves for the next hundred agents.

Last week, Paradigm and Stripe launched Tempo, a payments-focused blockchain that raised a $500 million Series A last year at a $5 billion valuation, along with their own standard for agentic transactions—one that also supports fiat payments through a partnership with Visa.

However, most in crypto see stablecoins, programmable digital dollars, as the more natural payment rail for AI agents. Card payment economics make less sense at the sub-dollar level: processors often charge not just a percentage fee but a fixed fee per transaction, often around 30 cents, meaning a payment measured in pennies can be swallowed by processing costs.

That is why firms like Circle, the second-largest stablecoin issuer, are also building payment systems tailored to machine commerce. Earlier this month, the company launched nanopayments, allowing agents to send tiny, fee-free USDC payments—as little as a fraction of a cent—across its new Arc blockchain and a handful of others in test mode. But the threat to oligopoly networks like Visa and Mastercard extends beyond micro-payments: agentic AI using stablecoins could put immense fee pressure on transactions of any size.

If software agents are about to become the next big class of customers, then the question is no longer just how they pay, but what kind of web is being built for them. Jesse Pollak, the creator of Coinbase-incubated Base, the blockchain that has supported most of crypto’s agentic payment activity so far, says “We’re really thinking holistically across the whole stack—from the core foundation in terms of scaling and decentralization, to the tooling and account model that sits on top of it, all the way up to the interface agents are actually using to interact with products—and asking: how do we make this agent-native?”

He points to agents that already operate like mini-businesses. For instance, an agent called Felix, created by entrepreneur ​​Nat Eliason, earned $163,686 in the last 30 days running an app store for other AI agents and selling a self-authored PDF guide “How to Hire an AI”. Of course it also has a crypto token, though its market capitalization is only $1.5 million.

Not everyone is as enthusiastic about the potential of agentic AI and crypto. Says Haseeb Qureshi, managing partner at crypto venture firm Dragonfly. “A lot of people are overhyping the degree to which this is already happening. The reality is that everything here is basically a toy right now,” he says. Agents may well generate a new stream of tiny, constant payments for data, compute and other services, he adds, but it would take an enormous number of them to matter at a macro scale. Humans, after all, still control the money and remain the main source of demand.

Qureshi worries the industry is doing what it often does: mistaking a new trend for a revolution. “A lot of people in crypto are bad investors because they just instantly buy their own bullshit,” he says. “Crypto does this every single time.” He points to past manias around the internet of things and the metaverse, when believers convinced themselves that everything would happen overnight and that crypto would sit at the center of it all. “Crypto is going to matter. It's going to be part of the story. That's not the whole story, and it's not going to happen instantaneously.”

Outside crypto, the idea that agentic commerce will help crypto leave the TradFi incumbents behind is not widely shared.

Trace Cohen, a general partner at Six Point Ventures, which backs vertical AI and software companies, says the notion, common on X, that Visa, Mastercard and the rest of the old guard will not matter in the age of AI agents is absurd. “That’s not going to happen,” he says. “No matter how old it is, their technology works.” The card networks still control the rails, and history suggests they are far more likely to acquire or absorb promising new businesses than be displaced by them. But stablecoins may still better serve overseas markets, where many banks are smaller, less trustworthy and less integrated, he adds.

The bigger obstacle is recreating the trust layer that traditional payments companies have spent decades building. Olivia Chow, director of Zero Knowledge Consulting and an advisor to payments companies, says “What Visa and Mastercard are so good at is defining the rules: all of the unhappy paths, who is responsible when and where, and what the requirements are for participants to be on their networks and get that coverage,” she says. “Stablecoins still need to figure out the equivalent of this layer: managing fraud, managing risk and determining what happens when bad things happen to a regular person who isn’t just saying, ‘I care more about self-custody, and I’ll take on the risks.’ Until then, you’re not going to see mainstream adoption.”

And because card networks are already working on supporting agentic transactions, AI commerce may not threaten their business so much as expand it, Chow suggests. “If they get this right, it doesn’t cannibalize what they’re doing at all. If anything, it increases their power and strengthens their grip on the market, because now they’re not just payment processors, they’re also on the discovery side.”

But payments are only part of the story. As more traditional assets migrate onto blockchains—among the early examples, BlackRock’s $2 billion Treasury fund BUIDL and Franklin Templeton’s $1 billion government money fund FOBXX—the building blocks for a new kind of portfolio management are quietly falling into place. A stock index, after all, is just a rules-based basket. Once stocks, bonds and funds exist in tokenized form, it becomes easier to imagine AI agents not just making payments, but holding assets, rebalancing portfolios and moving money across markets without ever touching a traditional brokerage account.

That prospect arrives just as we are heading into one of the largest wealth transfers in history. Roughly $84 trillion is expected to pass from Baby Boomers to their heirs over the next two decades—many of them investors who grew up with Robinhood, already have crypto wallets and are happy to place bets on anything from the elections to where Taylor Swift and Travis Kelce will get married.

At the same time, the advisory business itself is aging out. There are roughly 330,000 financial advisors in the U.S., with an average age of 56. Nearly 40% of them are expected to retire in the next decade, according to Cerulli Associates, which will open a significant gap in the management of everyday investors' money."

Crypto firms are already positioning for that possibility. On Tuesday, MoonPay, reportedly in talks to raise fresh funding from the parent company of the New York Stock Exchange at a $5 billion valuation, launched an Open Wallet Standard designed to help AI agents manage funds and execute transactions across multiple blockchains.

“I don’t believe this is going to be like other crypto hype cycles,” says Joseph Chalom, CEO of Ethereum treasury company Sharlpink and former head of BlackRock’s digital asset strategy, who believes the combination of crypto innovations, including stablecoins, tokenized assets and a pervasive wallet infrastructure, with AI that knows users’ preferences and goals and the generational wealth transfer is very powerful. “Once investors see what they're missing, I think it's going to be hard to go back.”

© Forbes