Why did the five giants collectively go out of business within a week to open bank accounts for AI?

By David, Deep Wave TechFlow

Original title: This week, everyone is helping AI open bank accounts

On March 18, another blockchain mainnet was launched.

It's called Tempo, and Stripe and Paradigm are behind it. Stripe, one of the world's largest online payments companies, handled $1.9 trillion in transactions last year; Paradigm is one of the biggest ventures in the crypto industry. The two companies teamed up to invest $500 million in Tempo last year. The project is estimated at:

50 billion.

The $5 billion blockchain doesn't trade coins, doesn't do DeFi, and doesn't send memes. On the day of the main network launch, Tempo's highest-profile products were:

Have the machine pay the machine.

This sounds a bit abstract; you can understand that AI is now costing money every step of the way. It costs money to adjust the API once, it costs money to buy a piece of computing power, and it costs money to pull a batch of data from a database...

However, all existing payment systems are designed for humans. Bank accounts require ID cards, credit cards require face recognition, and Alipay requires mobile phone verification codes.

AI alone can't beat it.

It can help you complete the entire workflow, but when you get to the point where you pay, you have to stop and wait for a human to click “confirm.”

As a result, an open protocol called MPP (Machine Payments Protocol) was launched along with the mainnet, co-authored by Stripe.

Simply put, it sets a set of rules for transactions between machines, including how to request payment, how to authorize, and how to settle.

The expected scenario is that AI can independently spend money within a pre-set amount, and there is no need to find a human signature for every transaction. On the day of launch, more than 100 service providers, including OpenAI, Anthropic, and Shopify, were connected.

But Tempo isn't the only one doing this week.

Within five days, Visa set up a new division to release AI payment tools, Coinbase's payment agreement underwent a major upgrade, Mastercard spent $1.8 billion to acquire a stablecoin company, and Sam Altman's World released a toolkit dedicated to AI authentication.

Five giants crammed into the same door for a week, rushing to open bank accounts for AI.

Two roads, one door

What Tempo does is help AI make payments. However, settlement is only one part of the payment system. To truly spend money independently, an AI agent also needs payment tools, funding channels, and identity authentication.

Here, traditional payment companies and crypto companies are using methods they are good at stealing the cake.

On March 18, the same day that Tempo's main network went live, payment giant Visa also moved. The newly formed Crypto Labs division has released its first product: Visa CLI, a tool that allows AI agents to initiate credit card payments directly from the terminal.

There is no need for an API key, and there is no need to register in advance. If AI needs to buy any service while running a task, you can pay by pressing a single command. Visa calls this “command-line commerce.”

Visa's global card network is connected to billions of cards and tens of millions of merchants. If AI payments can run on this ready-made network, it doesn't need to wait for any new infrastructure to mature.

Visa is an extension of the old path. Its opponent, Mastercard, chose another option: buy the road directly.

On March 17, Mastercard announced a $1.8 billion acquisition of London stablecoin infrastructure company BVNK. This is the biggest stablecoin acquisition in the crypto industry's history.

The purpose of this acquisition is also very straightforward. If the money paid by AI has to go through stablecoins, then stablecoins will go through my channel.

On the side of cryptographic native companies, the action is just as intense.

Coinbase's x402 protocol has undergone a major upgrade, expanding the scope of payments from several stablecoins to all ERC-20 tokens, and also released an MCP toolkit. Developers can connect AI tools to the payment network with one click.

The two sides seem to have different starting points, but they are doing things in the same direction: traditional payment companies are embracing crypto, and crypto companies are embracing AI. Ultimately, crypto infrastructure is becoming the underlying conduit for AI payments.

There's one more step left. AI can spend money, but how do merchants know if anyone is responsible for the AI that spent the money?

On March 17, World, co-founded by Sam Altman, released AgentKit, connected to Coinbase's x402. It only does one thing: let the AI pay while proving that a real, verified person stands behind it. The merchant can confirm that someone is responsible for this transaction, but they can't see who this person is.

In five days, every step of the five companies, settlement, channels, tools, agreements, and identity, was placed in a card slot.

AI finished dividing the cake, only the cashier was left

Over the past three years, the AI industry chain has basically taken over the position it can occupy.

The model layer is the table of OpenAI, Anthropic, Google, and many Chinese companies. The computing power is locked to death by Nvidia, and the application layer is a red sea from programming assistants to search engines...

Every floor is full of people, and the barriers to competition are getting higher and higher on every floor.

However, the payment level is still relatively empty.

It's not that nobody thought of it; it's just that the time hasn't come. There is a prerequisite for AI Agent payment: AI must first be able to independently complete an entire chain of tasks. If it can only chat, don't need to adjust the API, buy computing power, or hire another agent to do the work, then payment isn't just what you need.

Over the past year, this premise slowly began to take hold.

OpenClaw allows AI to directly operate computers, the MCP protocol allows AI to connect to external services, and the agent capabilities of major models will focus on breaking through in the second half of 2025. AI has changed from a “conversation tool” to a “tool for work”, and it costs money to do work...

The need to spend money is here, but the infrastructure to spend money doesn't exist yet.

That's why Stripe, Visa, Mastercard, and Coinbase launched at the same time. For traditional payment companies, this is the first time in the entire AI wave that they have gained a home field advantage. They can't make models, they can't make chips, but payment is something they've been doing for decades.

Visa's global card network is connected to billions of cards and tens of millions of merchants. Mastercard covers more than 200 countries, and Stripe processed $1.9 trillion in transactions last year. If every AI spend goes through these channels, then the more capable the AI, the more profitable it is.

For crypto companies, the logic isn't the same.

Coinbase CEO Brian Armstrong previously said a very straightforward statement: “AI can have a crypto wallet, but it can't open a bank account.”

Every step of the traditional financial system confirms “who you are”. Opening a bank account requires an ID card, applying for a credit card requires face recognition, and an SMS verification code for every transaction. AI is software, not people; it can't pass any of these levels.

But crypto wallets don't need these. A private key is an account. For AI Agents, on-chain payments are the least resistant path.

Whether encrypted or not, AI payments will be a new infrastructure-level market. The only difference is whose pipe is better suited to the machine.

The road was repaired, but the car didn't come

At this point in the story, it seems like everything is ready; the five giants each have their own.

But there's one number worth checking out.

Coinbase's x402 protocol is currently the earliest implemented AI payment protocol with the widest ecosystem. According to x402scan, the entire ecosystem's transaction volume over the past 24 hours was $654 million. 150,000 transactions, with an average of less than 5 cents each.

What infrastructure matches this figure? Tempo was valued at $5 billion, Mastercard spent $1.8 billion to buy BVNK, Visa set up a new division, and Stripe personally wrote the agreement.

The infrastructure, valued at several billion dollars, serves a market with a daily trading volume similar to that of a street milk tea shop.

This seems to be the norm for all infrastructure businesses.

On the eve of the Internet bubble in 2000, telecommunications companies laid millions of kilometers of fiber under the sea. After it was completed, it was discovered that only 5% of global internet traffic was used. Most of those companies have gone bankrupt, but fiber is still there.

Ten years later, video streaming and mobile internet have filled those pipelines. The people who paved the road didn't make any money, but the road was real.

AI payments are now at this stage. The logic of demand holds true: AI agents are indeed becoming more and more capable, they really need to spend money independently, and they really need a new financial infrastructure.

Everyone reached the starting line, but after firing the gun, they discovered that they were alone on the runway for the time being.

As to whose path eventually runs out, when did the AI Agent's first truly autonomous transaction happen in your life, either faster than anyone expected, or slower than everyone expected.

The only thing that's certain is that this battle has already started, and you and my wallet are probably the last to know.