11M Transactions per second: Breaking the money movement bottleneck

What we did

Today, we’re announcing a groundbreaking performance achievement for the Radius Network: sustained settlement of 11 million transactions per second with finality in less than 55ms and a cost basis that makes any price point viable.

The internet is undergoing generational change driven by the rise of AI agents. With that, we have an opportunity to put something in its place that better serves humans. The ability for agents to transfer value, at any speed or price point, is critical to building an internet that supersedes today’s attention economy. 

Achieving this performance took considerable time and effort. Each additional transaction per second, each millisecond of saved latency, was hard won by our engineering team. Importantly, we took no shortcuts out of the belief that payment shouldn’t mean friction. We wanted your, or more likely, your agent’s, money experience to be on par with any high-performance service on the open internet. 

All transactions were settled in real time, leveraged the full USDC smart contracts, and were executed on commodity Google Cloud hardware. 

We set out to prove that the money layer doesn’t have to be a bottleneck to innovation. So we built the highest capacity, fastest, and cheapest way to move a stablecoin on the open internet. 

Why this matters

Agents will and should use the internet differently from humans. Today’s model of agents using tools to navigate experiences built for humans is short lived. The opportunity lies in what is built in its place. 

The attention economy that dominates today’s internet stems from the dominance of closed ecosystems. It’s inefficient for every business to transact directly with every potential user, so instead, we pay with our time. We are not the first to make the case that agents armed with stablecoins offer an alternative to this model. But we do hold a specific view of what has to be true of the money layer to make this a reality:

1. Scale to meet any demand
2. Latency consistent with agentic execution
3. Costs viable at any transaction price point
4. Programmability to support embedded policies 
5. Real-time settlement between any set of actors

While points 1-3 are views widely held by others working on this problem, we believe that points 4 and 5 are the true unlock to a model that is structurally different from what we have now. Without programmability, as with the EVM, agents must inherit policy enforcement from a trusted ecosystem. Without real-time settlement, scale is only achieved when significant volume flows through long-standing bilateral relationships, as with payment channels. 

We think of these tradeoffs as the big box retailer versus the open bazaar. With programmability, your agent gets a personal security guard that travels with it. With real-time settlement, it gains the ability to negotiate and transact with any vendor. Without these traits, our agentic internet will look a lot like the one we have today.  

Why this matters

The transition to an agent-native internet will impact every website and every economic actor, whether they sell or buy, whether its bits or atoms. 

If you offer data or APIs, your traffic volumes will increase dramatically. If you sell goods or services, you now have to reach a brand new customer, one that might come with little data and high expectations. If you’re a content producer, your reader will no longer look at the ads placed next to your product.

Agents already dominate internet traffic. Change will happen quickly, and the winners will be those who go beyond stop-gap solutions to deploy something purpose built for this new user. 

At Radius, we built the money layer that makes this level and scale of innovation possible. Come test us out at network.radiustech.xyz. We’re looking forward to helping you design a better internet.