Report Finds 90.8% of AI Agent Payments Are Under a Cent

The newest wave of AI hype is nearly impossible to escape at every major financial conference this year.

The same refrain keeps coming up: autonomous AI agents are poised to take over global finance, moving serious capital with ease and running the economy for their human owners.

Does any of that reflect what is actually happening? Are AI agents moving money at real scale? Who is genuinely using them? And if the broader AI bubble collapses, would agentic payments go down with all that venture capital?

BeInCrypto’s new report, The State of AI Agent Payments 2026, sets out to answer those questions by looking at real transactions on x402, a payment protocol built for machines.

Produced in partnership with AWS, Cloudflare, Chainalysis, and Zoomex, the research finds that the rails for machine payments technically work, but real commercial demand is almost nonexistent.

Can AI Handle Your Payments Yet?

The data reveal a deep split between promotional narratives and on-chain reality. Even as protocols like Coinbase’s x402 and Stripe/Tempo’s Machine Payments Protocol process millions of transactions, 90.8% of all agent payments on record are valued at less than a penny.

In a 35-day study covering 6.4 million transactions, the total value processed was just $119,947.

The volume numbers are not evidence of an autonomous shopping or B2B boom; they simply mask sub-cent automated pings for API access, compute time, and basic data checks.

Francesco Gi, founder and CEO of Castle Labs, argues the numbers show an industry that put payment infrastructure in place before real market demand existed:

“BeInCrypto’s report accurately captures the current state of agentic finance: the payment infrastructure works, but demand is lacking. Most activity in the machine economy is small, repetitive, and concentrated among a handful of users, indicating that the market is still early. Solutions already exist for some problems that haven’t even arisen yet, but technology ultimately depends on adoption and use cases,” Francesco Gi said.

Institutional adoption reveals even more caution.

Micropayments earned a solid 60.7 out of 100 on the report’s readiness framework, while Institutional Financial Workflows came in at just 12.0 out of 100.

Among the 19 trading tools and financial systems the report examined, 14 operate only within spending limits that humans approved in advance. Setups at major exchanges such as Zoomex still require a person to approve every trade order.

AI-labeled tools accounted for less than 0.55% of all trading volume on platforms such as Hyperliquid.

Albert Castellana, CEO of GenLayer Labs, says sending a few cents across a blockchain is simple, but trusting software to make complex economic decisions is a very different matter:

“Most real agent transactions today are relatively small: paying for APIs, compute, data, maybe simple digital services. An agent can execute a transaction exactly as instructed and still make a terrible decision because the information was wrong, manipulated, or the other side didn’t actually deliver what was agreed. Moving money is one problem. Deciding what should happen when the transaction goes wrong is another,” Castellana said.

Castellana says current safety debates put too much emphasis on wallet permissions and not enough on whether decisions are accurate:

“You can give an agent a $100,000 limit, perfect wallet permissions, and every normal security control, and it can still spend $100,000 on something that clearly wasn’t what you intended. Getting money to move at machine speed is useful. Eventually we also need a way to resolve what happens when something goes wrong at machine speed.”

Security Dangers Are Real

The security issues go beyond spending caps. The report flagged $281,200 lost in exploits where properly budgeted funds were diverted to attacker-controlled wallets.

Robin Nordnes, founder and CEO of Raiku, says the biggest danger in trading is not simply malicious prompts; he also warns of what he calls the “execution gap”:

“The risk I worry about more sits between the decision and the confirmed trade. Say an agent makes a sound, compliant trade within budget, and the transaction gets delayed or outbid on the way to the block. If the agent can’t tell whether it went through, it tries again. Then again. Every retry was individually approved, and together they can do real damage.”

Once dependable execution is in place, Nordnes expects agents to take over trading tasks directly and human approvals to disappear — especially on high-speed chains such as Solana, which handled 94.5% of all transactions in the report.

Venture capital is already putting significant money behind that vision.

The report says 92% of VC funding, or $3.91 billion, went to underlying payment and cash management infrastructure such as Tempo and Mastercard/BVNK, rather than to pure trading bots.

Castle Labs’ Francesco Gi concludes that once agents build verifiable track records, the move from sub-cent API calls to true autonomous finance will inevitably happen:

“Today, a penny-worth credit transaction costs $0.30, making it economically unviable to use traditional finance infrastructure for micro-payments in the machine economy. As agents build verifiable track records, we expect investors to start backing those with verifiable reputations. Stablecoin and agentic finance adoption is inevitable,” said Francesco Gi.