Lately, when I look at Stripe’s moves, I feel like 2026 really is turning into a turning point for the machine economy. It’s not just automation anymore—it’s becoming realistic that a world where AI agents function as independent economic actors is actually within reach.

The “turbulent current” expression that Stripe co-founder John Collison talks about is exactly that. M2M transactions—automatic settlement between machines—are about to surge. But what matters here is that existing financial infrastructure can’t keep up with this pace. We need systems that operate 24/7 in real time, yet traditional banking systems are constrained by business hours. That’s precisely why USDC.

Choosing USDC as a stablecoin makes perfect sense. Its stable price means it can be used in commercial contracts, and it’s programmable. AI agents can automatically execute conditions like “buy if this data costs no more than $0.05 per megabyte.” With conventional credit card networks, small payments don’t work well due to minimum-fee problems, but on the blockchain, payments of under a cent can be processed efficiently.

You also shouldn’t overlook Tempo’s role. This was developed jointly with Paradigm, but when it comes to what the blockchain “trilemma” is—it's the balance problem among decentralization, security, and scalability. Existing chains are constantly forced into trade-offs among these three. But Tempo is optimized specifically for machine-to-machine transactions. It can process tens of thousands of transactions per second, offers predictable fees, and enables sub-second settlement. These are the requirements needed for AI agent commerce.

And then there’s the x402 protocol. I think this is quite innovative. It’s finally about making use of the HTTP 402 status code (“Payment Required”). A website returns a payment request → the AI agent reads it → it automatically pays with USDC. Once this loop is established, the web becomes programmable. Fully automated transactions—without any room for human intervention—become possible.

Changes on the user side are also interesting. Up until now, people have had to manually manage subscriptions or enter payment information each time. But going forward, it will be possible to delegate “spending authority” to AI agents. New agent-based wallets are emerging, equipped with fine-grained permission settings, whitelist features, and automatic replenishment.

It also seems likely that the monthly subscription model will shift. Paying $20 per month for a tool you only use occasionally is inefficient for machines. AI agents prefer usage-based billing models. They settle in USDC on the spot for whatever computing resources or data they consume. This kind of granular payment will generate large volumes of transactions.

The fact that Stripe is pursuing this kind of strategy is, I think, proof that the crypto industry is shifting from speculation to real-world usefulness. After understanding what the trilemma is, they’re building the infrastructure to solve it. In the era of the machine economy, the importance of stablecoin settlement and blockchain-based infrastructure isn’t just theory anymore—it’s moved into the implementation phase.