Rothera, the CFTC-regulated event contracts exchange and clearinghouse founded as a joint venture between Robinhood and Susquehanna International Group, has launched its trading and clearing platform on Adaptive’s Aeron Exchange Accelerator technology, going live in under five months, in time for the 2026 World Cup. The platform reports $3 billion contracts traded during its first two months of operation and is now approaching a 10% share of US event contract market volume.
Control & Ownership
“What’s really important about the Exchange Accelerator for people in prediction markets is that they want control and ownership of the finished product,” says Matt Barrett, the company’s CEO & Co-Founder, in conversation with TradingTech Insight. “One of the characteristics of prediction markets is that no one knows how they’re going to evolve. So it’s very risky to tie yourself to a product offering that, even if it says it’s perfect for prediction markets today, may not be in six or twelve months’ time. If you were building on top of someone else’s IP that you couldn’t control and own, and therefore change directly, you’re making a bet that the product vendor is going to go in the same direction you need them to. That’s really risky when something is evolving as rapidly as prediction markets are.”Barrett’s argument inverts the usual case for vendor technology. Buying a platform is normally the risk-reducing option; the vendor carries the development burden, the upgrade path and the operational track record. However, in a market where nobody can specify the product twelve months out, the vendor’s roadmap becomes the exposure rather than the mitigation. Equally, a regional exchange might want ownership because the alternative is a modernisation programme priced and paced by a vendor it would rather not be locked into.
“Ownership and control is hugely valuable to prediction markets. But different exchanges have different tech priorities,” Barrett says. “A prediction market might use our accelerator to rapidly launch entirely unique asset classes from scratch. A well-established regional equities exchange, on the other hand, faces different challenges – heavy competition from MTFs and investment banks means they must modernise legacy stacks without the massive cost and lock-in of standard vendors. The accelerator model gives both segments high-performance foundations and control over their destiny, minus the risk of multi-year development cycles.”
What the two have in common is the reluctance to trade control for delivery speed, a trade the Exchange Accelerator is built to avoid.
A Layered Approach to Exchange Architecture
“In terms of our architecture, the bit that comes with open source Aeron is the low-latency messaging, the persistence of messages to disk, and the clustering, which gives you the resiliency,” Barrett says. “That’s what anyone can get if they pick up open source Aeron. Our development layer that sits in the middle provides developer productivity, operational productivity, and operational tooling – to be able to manage a system built on top of open source Aeron in production, and to develop it very quickly. And then in the Exchange Accelerator, they get a domain model, with a matching engine and things you can point at and say: that’s an account, that’s an order, here are different order types, here are risk limits. You get those sorts of concepts.”
In terms of how much code accounts for each part of the stack, Barrett states that 25% is Aeron open source, and 60% is proprietary platform code. Although Aeron Exchange Accelerator is only 15% of the code base, it leverages the underlying open source and the proprietary development and operational tooling, neither of which have any business domain specific code, although they do include things like FIX infrastructure.
“It’s a very clean separation of concerns, and it lets our clients focus on the things that matter to them – the differentiation – without having to worry about the necessary but not sufficient underlying layers,” Barrett says. “Obviously an exchange needs to be performant and resilient. We provide the foundations of that performance and that resilience through the open source layer, and then through the proprietary development and operational layer that sits in the middle. And then we provide a fast, performant matching engine implementation that leverages all of that but sits on top of it.”
Whereas Aeron itself is open source, Adaptive licenses and retains the IP to the middle layer, and the client typically owns the top layer, in this case the Aeron Exchange Accelerator. For Rothera, Adaptive extended and customised this layer, while Rothera built its own reference data hierarchy – prediction markets being novel enough that the data model itself was greenfield work.
“Because of the nature of the engagement, they own what we’ve built at that exchange, clearing and risk engine layer – even though we’ve built it,” Barrett says. “We are not just a product firm, we’re also a professional services firm. So we’ve done that implementation for them. They bought a copy of the Exchange Accelerator, and we have extended it, customised it, made it fit for purpose for them. And then they retain ownership of all of that.”
Build Priorities
The regulatory work led the build sequence. “The regulatory-driven requirements that allow a firm to launch as a regulated exchange were the highest priority,” Barrett says. “And a lot of what’s required for a regulated exchange comes out of the box. So it’s not so much about going back and implementing things that are required by the regulator – it’s about having those already, but surfacing them in the operational context in which the regulator needs to see them. Because the regulator isn’t interested in seeing a software demo; they’re interested in seeing a business demo. Those things are obviously very early in the project roadmap and the project timeline.”
The shape of a fixed-length build follows from what the accelerator already contains. Out of the box, the platform arrives with a high-performance matching engine supporting standard and enhanced central limit order book features (a range of order types with time-in-force attributes across limit, cost limit, market and stop limit orders), a risk engine for fully-funded spot contracts, order management and market data through FIX and web endpoints, a market operations API with a foundational user interface, and historical data through FIX drop copy and database integrations. The codebase also ships with unit and acceptance tests, micro and system benchmarks, deployment tooling for local, cloud and on-premise environments, and technical design documentation.
What gets built during the launch window is the client-specific work that sits on top: protocol integrations, enhanced risk models such as margin trading rather than cash, bespoke matching features such as RFQs, and deployment into the venue’s chosen environment. Further out sit the roadmap items – additional contract and asset types, expanded risk and collateral management, and more exotic order types such as multi-leg orders and implieds. For a venue launching against a hard date, the division is what makes the date achievable: the regulated core is already present, and the differentiation is the part that gets built within the window.
Engineering for Continuous Operations
Event contracts don’t observe market hours, they have to operate around the clock and prediction markets have grown up as continuous markets by default. The same pressure is now reaching further into the mainstream, with established asset classes moving towards extended and eventually continuous trading on the back of tokenisation and shifting participant expectations. Building for that from the outset is a different exercise from bolting it on later.
“The approach we’ve taken to 24/7 operation is through building the platform on top of a highly performant asynchronous implementation of Raft, which is a well-known consensus algorithm,” Barrett says. “We’ve made a lot of additions and extensions to the Raft protocol itself to fit 24/7 operation, whilst maintaining very tight SLAs around throughput and latency.”
Automated failover was the first consequence. “With a simple primary-secondary implementation, you typically need a human to intervene to decide when to fail over,” Barrett says. “And that immediately blows your SLA. If a human’s involved, it’s 10 to 15 minutes. If you have a system that’s down, your uptime targets are lost, your SLA is lost, and you’re not running 24/7 any more. So that was one of the first decisions we made.”
Hardware maintenance presents the same problem in a different form. A venue that never closes has no window in which to replace a node.
“One of the things we can do, almost out of the box, is handle infrastructure upgrades,” Barrett says. “You can have a cluster of nodes running your matching engine, and you can take one of those nodes down – ideally not the leader, but one of the secondary nodes – change the hardware, and bring the node back up. You’ve upgraded the infrastructure without impacting the SLA of the system. People using the system have no ability to perceive that you have interrupted service or changed infrastructure at any time.”
The same mechanism handles code deployment. Move the leader, replace the node it was on, and the venue absorbs a latency spike measured in milliseconds rather than an outage. The coordination overhead this imposes on development and operations is the cost of patching an exchange without downtime.
Looking Ahead
“We’re very clear about where we’re playing, and we see a huge amount of interest in the prediction market space at the moment,” Barrett says. “We’re seeing a huge amount of interest in other asset classes more broadly, due to 24/7 and tokenisation, and that’s also driving a lot of uptick in the Exchange Accelerator business line for us.”
As prediction markets, tokenised assets, and perpetuals increasingly demand continuous operations and unprecedented flexibility, issues of ownership and control are likely to become increasingly important in the future development of digital markets. By offering a path that avoids the rigid constraints of traditional vendor platforms while bypassing the extended timelines of in-house builds, the Aeron Exchange Accelerator aims to provide a compelling alternative for modern trading venues.
Subscribe to our newsletter
