Crypto has been founded on a single premise for years:
There is someone behind a human who is sitting behind the wallet.
Someone creates an exchange, purchases some crypto, links a wallet, signs a transaction and proceeds with life.
But, that's beginning to sound like a thing of the past.
AI agents can perform more than just answering questions. They are able to search, make decisions, call APIs, call applications, and — increasingly — interact with financial infrastructure.
So, that makes for an odd opportunity:
Software could be the next big player in the crypto market.
Not a member of the Bitcoin community.Not a Bitcoin user.
Not a trader swapping.
An AI agent that needs to pay another service, purchase data, rent computing resources or settle a transaction automatically.
This could inadvertently form a new crypto economy in the shadows.
The wallet was designed for humans
Consider how a regular crypto wallet would operate.
You open it.
You look at your account.
It's your choice what to do.
You approve a transaction.
The network charges are collected.
You repeat.
This is because humans move slowly.
Machines aren't.
In many cases, an AI agent might have to make hundreds or thousands of small economic decisions without the need to wait for someone to click on a button each time.
It's there that the infrastructure gets fascinating.
This is already a trend that is taking shape. Coinbase has rolled out infrastructure for AI agents to have wallets and transact, while Circle launched its Agent Stack, featuring agent wallets and payment infrastructure.
This doesn't mean an “AI economy” has come to us.
However, there is proof that companies are constructing for one.
The real opportunity might not be the AI itself
What aspect of this do you find most interesting?
There is a general consensus that everyone is interested in the next great AI agent.
But perhaps what's below it is the bigger opportunity.
The following are the properties an autonomous agent must have:
A wallet.
A payment rail.
Identity.
Permission controls.
Settlement.
Security.
Data.
Reputation.
That is, a complete financial infrastructure stack.
In March, Coinbase's institutional research also made that claim, calling wallets, payment rails and settlement layers the infrastructure "that will support the emerging agent economy.
I think the most interesting crypto projects of this cycle don't always need to be the ones that are heavily branded, as AI.That's why I don't think the most interesting crypto projects of this cycle always need to be the loudest AI branded.
The unexciting infrastructure that no one talks about might be among the biggest winners.
Typically, that's the way technology operates.
The application is visible to people.
They leave out the plumbing.
Imagine an AI agent with its own budget
Here's a simple example.
Imagine you have an AI research agent.
You tell it:
“Find the cheapest reliable data source for this project.”
The agent searches several providers.
One charges $0.02.
Another charges $0.05.
A third charges $0.01.
Instead of asking you to enter card details, the agent could eventually pay the provider itself.
The transaction could be tiny.
The service could be used once.
And then it's finished.
This is where crypto becomes interesting because blockchains and stablecoins can support programmable, internet-native payments.
Research from Keyrock reported that AI agents had already settled millions of transactions using blockchain rails, although the overall market is still tiny compared with traditional payments.
So we're not looking at a finished revolution.
We're looking at the early infrastructure being assembled.
Stablecoins could become the machine's money
There is also an important part to that:
Stablecoins.
An AI agent doesn't have to bet on volatile possessions.
It must be much simpler.
A digital dollar which can be moved globally and by program.
This makes stablecoins especially noteworthy with regards to machine-to-machine payments.
Just recently, Circle announced that USDC's total circulation reached $73.3 billion, while the transaction volume on chain had increased significantly YoY. The company is also clearly shifting its focus towards AI-powered agents and digital payments.
The intriguing question is not:
Will AI be used with Bitcoin?
What is a more useful question?
What will the money be used for to pay other software?
That's a very distinct question.
But there's a huge problem
It's a thrilling idea to provide an AI with financial resources.
It's also very frightening.
Because an AI agent that makes a mistake is not like a human that makes a mistake.
An ordinary person might send $100 by mistake.
This could be repeated by an automated system thousands of times.
This is why the future generation of crypto infrastructure should have far improved controls surrounding:
Spending limits
Transaction permissions
Identity
Agent verification
Actions that can be undone or that are limited.
Monitoring
Fraud prevention
Reputation
In fact, this is precisely what a recent academic paper about the new “Web4” agent economy pointed out: Growth is on the rise but interoperability in identity, authorization and payment structures is still quite immature.
So the greatest problem is not just:
“Can we give AI a wallet?”
We already know the answer is increasingly yes.
The difficult question is:
“How much freedom should that wallet have?”
This could change what “adoption” means
Crypto has been working for years to convince the next billion people to use wallets.
But perhaps there's another way.
Rather than persuade a billion users to grasp blockchain transactions, the crypto infrastructure could be hidden under software which people are already using.
You wouldn't necessarily know that there's a blockchain going on.
An AI assistant can just buy something for you.
Application may be settled right away.
Your program can be programmed to buy data automatically.
You can get an agent to pay another agent for a service.
The blockchain acts as the back-end.
The user doesn't care.
But truthfully, that could be the type of adoption crypto has actually been seeking.
The next crypto cycle could be about infrastructure
In the light of the direction the industry is taking, I'm more and more sure that the next fascinating part won't just be:
Which token gets higher?
It could be:
So, who will fund the financial backbone of autonomous software?
Wallets.
Stablecoins.
Payment protocols.
Identity systems.
Agent marketplaces.
Settlement layers.
Security infrastructure.
These are far less interesting than meme coin.
However, they may be much more significant.
If AI agents are to be real economic actors, they'll need more than intelligence.
They'll need money.
But as new players enter the economy, the infrastructure must be created to enable them to do commerce.
Perhaps the greatest opportunity in crypto is to get more humans to use crypto.
Perhaps it's taking aim at the time when software will begin to use it for us.
