Brae Systems builds data centers underwater. They don’t need to build software on the side.
Brae Systems is a Chicago startup building underwater data centers. It uses Plexus to see what its pods are doing underwater and what its servers are doing in production.
The challenge
Brae is building data centers meant to sit on the seabed. Its prototype pod has been lowered into a river for test runs, and it now runs a rack of servers for a paying customer. That means two very different streams of telemetry. On the pod, Brae tracks accelerometer, gyroscope, and stress and strain readings, plus cooling-loop flow rates and temperatures. In the data center, it tracks power usage, server health, and server temperatures.
Vish, Brae’s CEO, is a software founder himself. His first instinct was the tool he knew from his last company, Datadog. He also remembered paying those bills. Datadog is built for software teams, around views like P95 and P99 latency and endpoint counts. Making it work for a hardware system on a riverbed is a different problem.
Brae’s other option was to build its own. Vish knew the hidden and upfront cost of that.
“You could clone Plexus. I don’t want to. Building it is maybe the first three months. Then you maintain it for the life of the company.”
He describes software as a wheel. Every feature you add is another layer of rubber that has to go all the way around. Skip that work and the wheel goes floppy. For a hardware company, every hour spent on that wheel is an hour not spent on the hardware.
Why Plexus
Brae looked at more established, better-funded tools. Plexus won because it wasn’t trying to sell a one-size-fits-all product.
“What drew us to Plexus was Ann saying: let us build this in the shape of your hardware, and let us onboard you by hand.”
Vish’s logic is to buy something built for your use case instead of reverse-engineering something you have to maintain. It also mattered that Plexus is a startup. “All else being equal,” he said, he was going to choose the early-stage company building alongside him.
What changed
See it: the pod, live, as it goes into the water
Plexus shows Brae’s pod on the dashboard in 3D, moving in real time. Vish’s favorite moment is watching it bob around as the team lowers it. The same view gives the team the forces and the jerks the pod takes. That’s data Brae needs to design a pod that can stay on the seabed.
“My favorite thing is watching that thing bob around on the dashboard as we lower it. Collecting that data has been super helpful for us.”
Prove it: from test runs to a paying customer
When Brae signed with Plexus, it was running prototype tests. About three months later, the same platform monitors a production rack for a pilot customer: power usage, server health, and server temperatures, all in one place. There was no new tool to buy and no migration.
Decide what’s next: building iteratively, like a software team
Most hardware work still runs waterfall: build everything, then review it. Brae works the other way. It even set up regular check-ins with its third-party engineering reviewers instead of waiting until the end, and the firms were grateful. Plexus fits that approach, because every test run produces data the team can act on before the next one. Next, Brae plans to overlay a stress and strain heat map on the pod view to see exactly where the forces land.
The results
- From prototype test runs to a paying customer's production workload in about three months, on one platform
- No in-house telemetry software to build or maintain, so engineering time stays on the hardware
- A live view of the pod's motion and forces during deployments, which feeds directly into design decisions
“I wouldn’t discount the value of not owning software.”