Translation: Baihua Blockchain

Visa's entire business revolves aroundbehavior bets. It involves human consumption habits and psychology. The reward points you collect, the human protection you rely on, the coveted Centurion black card, and the "zero liability policy" that makes you feel safe when withdrawing cash from ATMs abroad—all of these exist not because "money movement" is so difficult, but becausehumans are anxious, status-driven, and not good at reading terms and conditions. Visa built a $500 billion company in this tower.

However,AI agents do not possess these traits at all.

They do not collect points, are not made safer by fraud protection, and do not aspire to black cards. They have only one directive:complete any task. When tasks involve payments, intelligent interfaces perform calculations that people are too lazy to do:the cheapest path, the fastest settlement, the least fees. Each time, it is completed automatically, without emotion, and with pinpoint accuracy.

2028 Global AI Crisis

Last month, an article titled "2028 Global AI Crisis" on SubStack led toVisa's stock dropping 4%, Mastercard down 6%, and American Express plummeting 12%. Although the report referred to this as a "scenario imagination" rather than a prediction, the market did not buy it. The technical statement itself is not important; the core issue is:by 2027, AI agents will reroute the Tokyo system (swaps) and instead use stablecoins for settlement. Visa took fifty years to create the perfect product for a customer base that is now being replaced.

In "machine-to-machine" (M2M) commerce,2-3% swipe fees (swap rates) are an extremely obvious target for attack. As Citrini Research states: this does not mean that AI will destroy Visa tomorrow, butthe rate structure on which Visa built its empire is essentially a tax on "human irrationality," while agents are perfectly rational beings.That's their significance.

What is Visa Selling?

To understand why this is crucial, you must grasp what transaction fees are. When you buy something with a credit card, the merchant pays 2-3% to the card organization and issuing bank. The previous money funded yourreward points, gift protection, shopping insurance, and dispute resolution services. The entire consumer value of credit cards is universally believed to be financed by the purchase of items, and merchants then shift the costs onto ownership by slightly raising prices. A beautiful and stable system that has been running for fifty years,is paid for by humans in transactions, merely willing but not directly.

AI agents do not need these things.They do not initiate disputes and do not require cashback. The protective measures that support transaction fees are essentially defenses against human error, human fraud, and alerts.Once humans are removed from the transaction, these fees completely lose value.

American Express (Amex) epitomizes this issue. Its customers are high-income, high-spending elite groups seeking status. Its fees are higher than Visa or Mastercard because customers are willing to pay for the privilege of status and exclusivity. The entire model assumes aconscious purchase, choosing Amex over Visa for lounge access. Butagents will not choose Amex; agents contact the high-end population looking for the cheapest option to complete tasks.In the world of software cardholders, so-called "tiers" do not exist at all.

Agent-driven commerce bypassing transaction fees poses a huge threat to those dependent on affiliated banks and single-business issuing institutions. A significant portion of these institutions' profits comes from these 2-3% fees, and they have built an entire business segment around possessing their own rewards programs. Visa and Mastercard may have transformed network businesses, but thosewhose profit-loss models (P&L) rely entirely on transaction fees and rewards points will have nowhere to go.

Weekly Cumulative Shipments

Citrini's report coincided with the dense infrastructure launch within the same three-week window.

• Tempo was launched on the mainnet on Wednesday. This is a payment blockchain launched in collaboration with Stripe and Paradigm, designed for high-frequency stablecoin settlements.

• Also launched wasMachine Payments Protocol, an open standard allowing AI agents to autonomously make payments while consuming every confirmation step by humans. This protocol introduces a "Sessions" model: humans authorize a one-time spending limit, and agents continuously makestreaming micropayments as they consume data, compute, or API calls. This is akin to "money's version of OAuth." Authorization flows in place, agent expenditures, each step incurs a swipe.

Anthropic, DoorDash, Mastercard, Nubank, OpenAI, Ramp, Revolut, Shopify, Standard Chartered Bank, and Visa have all been default partners of Tempo.The entire payment business stack acknowledges this structural change.

On the same day Tempo launched,Visa's crypto department released a command line interface (CLI) tool targeted at AI agents. Agents can make payments directly in terminal without needing an API switch, an account, or human authorization for individual transactions. Visa calls it "Command Line Commerce"—machines transact without human intervention.

Additionally:

• Mastercard agreed to acquire stablecoin infrastructure companyBVNK for $1.8 billion.

• Circle has launched a test forNano payment online. Sub-framble, zero Gas fee USDC transactions designed for paying AI agents per API call.

• Sam Altman'sWorld (formerly Worldcoin) project has launched AgentKit, allowing agents to showcase their representation of real humans' cryptographic technology, directly integrating into Coinbase's payment rails.

  

In my view, what happened this week isthat companies are racing to become the "new Visa" because Visa realizes it has lost everything.

The Compliance Theory of Origins

Now, there has yet to be a clear indication:Visa is not sitting idle.

It is involved in the Machine Payments Protocol of Tempo, established the Visa Crypto Lab, and its crypto department head has explained in Fortune how agents use synchronized rails for payments under new standards. Mastercard is laying out $1.8 billion for stablecoins, and Stripe acquired Bridge and Privy.Legacy organizations understand this transition and are embedding themselves into the new infrastructure just before technology fully arrives.

Visa's argument is:before it can build new rails that make Visa connections irrelevant in agent businesses, it can first expand its own rails to cover agent transactions.

The argument is not currently flawed. Stripe's figure reached $19 trillion in 2025 (a 34% year-on-year growth). The card organizations' channel advantages are hard to replicate. But I must admit, I am reluctant to state this viewpoint loudly because historically, whenever this point is made, new products emerge and make you look foolish.

The flaw in this argument lies in that Visa's channel advantages are built on merchant relationships and consumer trust. Merchants accept Visa because consumers hold Visa; consumers hold Visa because merchants accept it. The core of the entire flywheel is “humans in transactions” . Once agents become the primary buyers in certain significant commercial categories, the flywheel will terminate.Agents have no brand loyalty, no wallets; what they have are directives and instructions. Whichever rail is the cheapest and fastest wins their business, and switching costs are quite clear.

The Gap Between Data and Narrative

I want to accurately describe our phased approach because narratives often run ahead of the data. Although around thex402 protocol (note: the original refers to a type of agent payment protocol) the ecological valuation is approximately $7 billion, on-chain data shows that the protocol processed $28,000 per day last week, mostly from testing. This figure is not even in the same league as Visa's daily processing volume.

However,x402's transaction volume has surpassed 50 million transactions. Although the single amounts are negligible, the transaction count indicates thatthe foundation is being utilized, and developers are building on it.Payment merchants accepting agent payments are growing.This is how payment networks are born.

McKinsey predicts that by 2030, AI agents may intervene in$3 trillion to $5 trillion of global consumer commerce. This valuation could be accurate or overly optimistic. But the undeniable controversy is:agent-driven commerce has not yet scaled. Those businesses building agent services, deploying agents as primary buyers, and making large-scale transactions that will genuinely challenge the fee economics are still under construction.

Citrini's report creating market disruption is due to a series of reliable chain reactions. As of 2027’s first thousand-quarter reporting may not make "agent-driven price optimization" a frightening reason.It is not time yet.

Shock will first occur in the field of micropayments for AI infrastructure, in the mass non-consumption sector. An agent executing research tasks may call hundreds of specialized data APIs in one session. Each call costs only a few dollars. Over a week, it may generate $40 in revenue.Traditional card organization networks cannot handle such transactions. The minimum economic model for these kinds of transactions won't work; merchant onboarding processes won't work, fee structures won't work.This category of commerce design will not operate on Visa's rails. It requires something entirely new, and x402, Nano payments, and Tempo are building it.

As for the innovations simulated by Citrini pertaining to consumption, that will come later. It requires agents to handle a considerable proportion of autonomous spending, which depends onwhether humans are willing to trust and authorize their current purchasing decision in consumption operations to agents.

Visa is being disrupted by a "better customer"—this customer is related to those who once made Visa great in non-ethnic national interests. 2-3% transaction fees are not a transaction tax, but a tax onhuman rationality. Agents are perfect rational beings.

How do I know this is important?Because Visa spent $1.8 billion this week to ensure it wouldn't be excluded from the answer.

Article link: https://www.hellobtc.com/kp/du/03/6262.html

Source: https://www.thetokendispatch.com/p/ai-agents-are-coming-for-visas-lunch