We’ve spent the last few years teaching computers how to think. Now we’re teaching them how to do things.
Book my flight. Find me a hotel. Order dinner. Buy whatever information you need. Pay for it. Let me know when you’re done.
These are called AI agents. Instead of just answering a question like ChatGPT does today, the idea is that you give an AI a job and it goes out and completes the job for you.
This creates an interesting problem that I hadn’t really thought much about until I read a fascinating new research note from BlackRock.
Their basic argument is that AI is creating machines that can think and act, while digital assets are creating money and financial infrastructure those machines can actually use.
Which got me thinking:
Robots need money, too.
I don’t mean actual robots walking around with wallets in their pockets. I’m talking about software that can make decisions and then spend money without asking you to pull out your credit card every five minutes.
Imagine telling your AI, “Book me a trip to Miami next weekend for less than $2,500.”
The AI starts working. It checks your calendar. It searches for flights and hotels.
Maybe it pays a few pennies to access better flight information. Maybe it pays another service for hotel prices.
It compares everything, books the trip, and sends you the itinerary.
You never see the hundreds of little things happening behind the scenes. You just know you’re going to Miami.
But here’s the question: How does the robot pay for all that stuff?
BlackRock (BLK) has a great way of describing what may be coming. They call artificial intelligence “machine-native intelligence” and digital assets “machine-native money.”
In other words, we’re building computers that can think for themselves at the same time we’re building money that computers can use by themselves.
Maybe those two things were always going to find each other.
Who Does a Robot Bank With?
Think about money from a computer’s point of view.
Banks close at night. They close on weekends. Credit cards charge fees. Some payments take time to settle. Opening a bank account can require paperwork and identification.
That’s fine for us. We’ve been building the financial system around humans for hundreds of years.
But computers don’t sleep. They don’t know it’s Sunday. They don’t care that Christmas is a bank holiday.
And they might want to make millions of transactions so small no human would ever bother making them.
Maybe an AI needs to pay one penny for a piece of information. Another computer might charge half a cent to use some data. Another might charge a few cents for a little extra computing power.
This is where digital money starts getting interesting.
Take stablecoins. The easiest way to think about a stablecoin is as a digital dollar that lives on the internet. One U.S. dollar stablecoin is generally designed to be worth about one U.S. dollar.
There are already more than $300 billion of stablecoins in circulation. According to BlackRock, adjusted stablecoin transaction volume exceeded $11 trillion in 2025.
That’s a lot of digital dollars moving around before the robots have even really shown up.
And here’s where it gets fun.
You’ve probably clicked on a broken website and seen “404 — Page Not Found.”
Well, there’s another internet code called “402 — Payment Required.”
Coinbase is developing a system called “x402” that puts that old code to work. The basic idea is incredibly simple.
One computer asks another computer for something. The other computer says, “That’ll be a penny.” The first computer pays the penny, and the second computer hands over the information.
Done.
No checkout page. No credit card number. No person sitting there clicking “buy.”
Machines paying machines.
And it’s not just crypto companies working on this.
Companies such as Visa (V), privately held Stripe, and Google parent Alphabet (GOOGL) are building different ways for AI agents to make payments using both the financial system we already have and newer digital systems.
That’s the part I find so interesting.
For years, AI and crypto have been treated like two completely different things. AI was ChatGPT and Nvidia (NVDA). Crypto was Bitcoin and Ethereum.
We’ve spent plenty of time around here talking about Bitcoin, including recently asking why Nobody’s Selling Bitcoin.
But maybe they’re not so different after all.
AI gives machines the ability to think and make decisions. Digital money gives machines a way to pay for those decisions.
What If Computers Buy Computers?
There’s one more piece to this that might be my favorite.
AI requires an enormous amount of computing power. That’s why you’ve heard so much about Nvidia chips and giant data centers being built all over the world.
We’ve talked before about how wild the economics of the AI buildout are getting, with everybody seemingly financing everybody else to build more data centers, buy more chips, and create more computing power.
Think of computing power like electricity for artificial intelligence. The more work AI does, the more computing power it needs.
And we’re talking about a potentially enormous market. BlackRock points to analyst estimates suggesting that the big cloud businesses from Amazon (AMZN), Microsoft (MSFT), and Google could generate about $1.1 trillion per year by 2030.
That’s why I keep coming back to the same question when I look at all the money pouring into AI infrastructure: What If This Is Just the Beginning?
Whenever humans find something valuable, we eventually figure out how to trade it.
We trade oil, natural gas, and electricity. Farmers can even trade corn that hasn’t been grown yet.
So why wouldn’t we eventually trade computing power?
BlackRock actually discusses the possibility of compute futures.
Imagine your AI is working on a huge project and realizes it needs more computing power. Instead of asking you what to do, it goes shopping.
One company has computers available in Texas. Another has them in Virginia. One is cheaper. Another is faster. Another has newer chips.
Your AI compares the choices, picks the best one, rents the computing power, pays for it and gets back to work.
You didn’t do anything.
Eventually, you could have computers buying computing power from other computers and paying them with digital money.
It sounds like science fiction until you realize that pieces of this are already being built.
So what does any of this mean for us as investors?
I don’t think the answer is to run out and buy every stock with “AI” or “crypto” in the presentation.
The more interesting question is what these machines are going to need.
They need computing power. They need chips. They need electricity. They need data centers and networks.
And if BlackRock is right, they may also need a new kind of financial infrastructure that lets machines pay other machines.
That gives us a pretty good list of things to watch.
We can watch the companies building the infrastructure. We can watch the blockchains and payment networks where the money is actually moving.
We can watch stablecoin adoption. We can watch demand for computing power.
And, most importantly, we can watch the prices of the assets tied to all of it.
We don’t have to predict exactly what the robot economy will look like in 2030.
That’s the beauty of being an investor.
If this is really happening, the market will tell us.
For the last 15 years, one of the biggest questions about crypto has been pretty simple: What are people actually going to use this stuff for?
Maybe that’s been the wrong question.
We’re creating an entirely new group of participants in the economy.
They don’t sleep. They don’t take weekends off. They can talk directly to other machines. They can make thousands of tiny decisions in seconds.
And, increasingly, they’ll be able to spend money based on those decisions.
BlackRock calls AI machine-native intelligence and digital assets machine-native money.
I think that’s the whole story.
We’ve spent years waiting for people to start using digital money.
Maybe we should’ve been waiting for the robots.