Agentic Commerce Won't Kill Cards, But It'll Open A Gap
Original author: @nlevine19
Compiled by: Peggy, BlockBeats

Editor’s note: Whether stablecoins will replace Visa and Mastercard has long been a recurring topic of discussion in the crypto industry. Author Noah Levine argues that this debate may be missing the point: rather than challenging card networks, stablecoins are first poised to serve new types of merchants that traditional payment systems have yet to reach.

As AI programming tools lower the barriers to software development, an increasing number of “temporary” and “micro” services are emerging: entities without a formal company structure, no website, and no long-term operational history, yet capable of conducting high-frequency transactions between machines. Before traditional payment systems establish risk assessment and integration mechanisms, these new merchants often struggle to gain access to card payment capabilities.

In this institutional gap, stablecoins may first emerge as infrastructure—not replacing existing payment networks, but filling commercial scenarios that remain uncovered. Understanding this may bring us closer to the true logic of this payment transformation than debating “who will replace whom.”

Note: The author of this article, Levine, is currently an investment partner in a16z’s crypto division, with a long-standing focus on the intersection of crypto, payments, and financial infrastructure. Before joining a16z, he worked at Visa on on-chain strategy and data, and also held strategy and data roles at RTFKT.

The following is the original text:

Wrong battlefield

A few weeks ago, an article released by Citrini Research suggested that stablecoins would cause a "disintermediation" shock to Visa and Mastercard, a claim that briefly sent markets reeling and caused significant declines in the stocks of these card networks. The crypto community cheered on social media. The argument sounded logical: AI agents would optimize every transaction, and card fees are merely a "tax" that stablecoins could bypass. I work in the crypto industry every day and wish this assessment were true—but in reality, most of it is wrong.

The reason is not that stablecoins are unimportant, but that the real opportunity does not lie in replacing credit cards. The real opportunity lies with merchants who will find it difficult to access the credit card system in the future.

Bank cards will still win the majority of the battlefield.

Citrini's argument rests on the assumption that AI agents, unbound by human habits, will proactively eliminate credit card processing fees.

But bank card networks are not just tools for transferring funds. They also provide:

·Collateral-free credit

·Pre-authorization for uncertain transactions

· Fraud protection and chargeback mechanism

Stablecoins can indeed be transferred, but they are currently not able to fulfill these functions.

Here’s a simple example: If your AI agent books a hotel for you, but the actual experience is completely different from the listing, you can dispute the charge and request a chargeback with a credit card; however, if you pay with a stablecoin, the funds are essentially irreversibly transferred.

The reality is:

·82% of Americans hold credit cards with rewards points

·There are approximately 18 billion bank cards in circulation globally.

For most consumer scenarios, users will not voluntarily give up shopping protections and reward points to switch to a payment method that is irreversible and offers no additional benefits.

Fraud detection has further widened this gap. Card networks can run risk models in real time across billions of transactions worldwide, while stablecoins currently lack a similar network-level anti-fraud system.

Opponents often present more specific scenarios, but the conclusion usually still boils down to the same issue. Micropayments are often seen as a weakness of the card payment system. However, card networks have previously encountered numerous transactions deemed unsuitable for card use and have continuously adapted their products.

For example, Visa has processed over two billion public transportation payments by consolidating multiple card swipes into daily settlements. The card industry has never truly abandoned any transaction category; instead, it consistently develops new products to cover these scenarios.

Another common statement is: "AI agents cannot hold bank cards." But an AI agent is essentially just a new device. Your phone, watch, and computer can all hold different payment tokens pointing to the same card—this is exactly the technology used by Apple Pay. The phone itself does not complete KYC; it simply carries your payment token.

AI agents can be the same.

In fact:

Visa has issued over 16 billion payment tokens.

Visa's Intelligent Commerce framework has entered the pilot phase.

Mastercard's Agent Pay is now available to all U.S. cardholders.

Meanwhile, the Agentic Commerce Protocol, built by Stripe and OpenAI, has launched, Etsy has integrated it, and over one million Shopify merchants are preparing to connect.

For existing merchants and consumers, card networks will likely still dominate commercial payments in the AI agent era.

The real opportunity for stablecoins lies elsewhere.

Those "merchants that do not yet exist"

Each technological platform shift creates a new class of merchants that legacy payment systems cannot serve.

This pattern has occurred repeatedly throughout history:

When eBay enabled peer-to-peer trading, these sellers struggled to obtain merchant accounts, so PayPal stepped in to serve them and quickly grew into a platform with millions of users.

·Shopify grew from 42,000 merchants to 5.5 million merchants over 13 years

As investors Alex Rampell and James da Costa have pointed out: when Stripe was founded, many of the companies that would later become its customers didn’t even exist yet.

The pattern in the payments industry has always been simple: winners typically serve new merchants that traditional institutions are still unwilling to take risks on.

The AI wave may create these new merchants at an unprecedented speed.

Over the past year: 36 million developers joined GitHub; at Y Combinator’s Winter 2025 startup batch: over 95% of the code for a quarter of the companies was generated by AI; on the AI programming platform Bolt.new: 67% of 5 million users are not developers.

This means: millions of people who previously couldn’t write production code are now releasing software. They are simultaneously: buyers of development tools and sellers of new software services. And these transactions often occur via the command line, rather than sales meetings.

Imagine this scenario: A vibe coder uses an AI programming tool to build an API that displays financial data of publicly traded companies in four hours. This project might have: no website, no terms of service, no corporate entity. Yet, another developer’s AI agent calls it 40,000 times within a week, charging 0.1 cent per call, generating $40 in total revenue. No one ever visited a checkout page.

I see similar tools being created every week, and the first question these developers almost always ask is: "How do I get paid?"

And the answer now is often: they don’t receive the money.

Existing payment processors struggle to onboard these merchants not due to technical limitations, but because of the risk structure.

When a payment processor allows a merchant to onboard, it assumes the merchant's risk:

· If a merchant commits fraud

·If a large number of chargebacks occur

The processor must assume responsibility. Therefore, the processor will only approve merchants that can be underwritten.

An API service with no website, no physical company, and no operating history will find it difficult to pass this review.

The system did not malfunction; it was simply not designed for this scenario.

Payment processors may adapt to this change in the future. Historically, they have made similar adjustments, such as creating new risk tiers for platform merchants.

But this process is slow. It took 16 years from PayPal’s founding until the industry established Payment Facilitator risk rules, and these new merchants need to receive payments now.

For them, accepting stablecoins is like a street vendor accepting only cash—not because cash is better, but because their identity struggles to pass the bank card system’s verification.

For example:

The x402 protocol now allows stablecoin payments to be embedded directly in HTTP requests.

· No merchant account required

· No need for a payment processor

·No review required

· There is also no chargeback risk

These do not require people to believe that stablecoins are better than bank cards. Only one fact is needed: the traditional payment system has not yet adapted to these merchants.

Stablecoins are not meant to replace bank cards, but rather to replace "nothing at all".

These new merchants won't choose between stablecoins and bank cards. Their choice is: stablecoins, or no payment method at all.

What will happen?

Historically, every wave of new merchants has eventually been absorbed by traditional payment systems. This time is likely to be no different—just a matter of time.

But the pattern is always the same:

Merchants appear first

The risk review system follows closely behind.

Between this time difference, stablecoins become the infrastructure.

Bank card services are available to all merchants approved by payment processors.

Stablecoins serve all merchants who cannot pass the approval process.

The next wave of business models is likely to emerge from the gap between these two.

[Original link]

Click to learn about the open positions at BlockBeats

Welcome to the official BlockBeats community:

Telegram subscription group: https://t.me/theblockbeats

Telegram group: https://t.me/BlockBeats_App

Official Twitter account: https://twitter.com/BlockBeatsAsia