AI servers generate far more heat than traditional servers, and many new data centers are turning to liquid cooling to manage it. Liquid transfers heat efficiently, but it also raises the stakes. Pumps, valves, sensors, and long piping loops must operate in sync, and small shifts in system behavior can affect cooling across an entire room of servers.
The team at Nautilus Data Technologies® encountered these challenges well before liquid cooling became a broader industry focus. Years of daily operation revealed what held up under load, what failed, and what a full liquid cooling loop required to remain stable over time.
How Nautilus Technologies Got Started
Nautilus Data Technologies launched its first liquid-cooled facility in December 2019 in Stockton, California. Built directly on the San Joaquin River, the data center was designed to operate sustainably by using the river as a heat sink, an approach few operators were willing to attempt at the time.
The facility’s primary cooling system is an open loop. River water is drawn in, passed through heat exchangers, and returned to the river with minimal temperature change and no environmental impact. The river water never mixes with internal facility systems, eliminating the need for chemical treatment, additives, or discharge management.
By avoiding cooling towers and chillers, Nautilus significantly reduced energy use, water consumption, and operational complexity. In 2019, very few data centers were attempting anything comparable.
As Nautilus’ Vice President of Product Augustin Roca puts it, “We were developing and operating facility-level liquid cooling years before liquid cooling became mainstream. We were liquid cooled before it was cool.”
Nautilus Data Technologies' first liquid-cooled data center sits on the San Joaquin River in Stockton, CA.
Turning Operating Experience Into OEM Products
Operating the Stockton site gave Nautilus something most cooling manufacturers lacked: years of hands-on experience running a fully liquid-cooled data center. That perspective provided early insight into where the industry was heading.
“We built what we needed to run our own facility, but we knew the market would eventually catch up to those same requirements,” says Roca.
As liquid cooling gained traction in new builds, Nautilus transitioned from operating data centers to developing commercial products, using the Stockton facility as the foundation for its OEM liquid cooling systems.
An example system with cooling skids and liquid-to-chip cooled servers.
What Operating Their Own Facility Taught Them About Control
Running a fully liquid-cooled data center at scale gave Nautilus insight most OEMs never gain firsthand.
Day-to-day operation revealed how liquid cooling systems respond to real world operating conditions and variations in IT load, how interactions between components propagate through the hydraulic loop, and which failure modes have meaningful operational impact.
“Being an operator gives us a unique edge. We see what holds up under real load, not just what works on paper,” says Roca.
Experience drawn from hundreds of thousands of unit hours of operation at the Stockton data center reinforced a simple truth: cooling hardware is only as reliable as the controls running it.
Over those operating hours, the team validated critical behaviors under live IT load. The system maintained stable operation as pressure shifted across varying loads, remained resilient to configuration errors that would challenge less robust systems, and continuously adapted as flow and thermal demand changed.
Maintaining stable pressure, coordinating pumps, handling fault conditions, and giving engineers a clear operational view of the cooling loop all depend on flexible and reliable control logic.
As Nautilus began engineering its next-generation facility scale cooling distribution unit, those lessons shaped every requirement. The team needed a platform that could move data cleanly, adapt as designs evolved, and provide deep visibility without relying on bolt-on systems.
A layered design of the floating data center, separating electrical systems, structural framing, and the data hall
Building Something New
Nautilus facility cooling distribution units were designed from the outset to serve entire data halls across a wide operating range. Earlier generations validated the core architecture, coordinating multiple subsystems while providing operators with a clear, real-time view of system behavior under live IT load. These systems established the foundation for hall-scale liquid cooling, but they were not designed as commercial products.
The EcoCore FCD™ represents the next step in that evolution. It was engineered as a purpose-built, facility scale platform for external deployment, with an emphasis on predictable behavior, operational visibility, and adaptability as system designs and workloads change.
Designed to operate stably across a wide range of flows and pressures, the FCD (Facility Cooling Distribution) remains resilient to configuration changes while providing operators with clear insight into system state without relying on external tooling.
The result is a facility cooling platform that translates years of operating experience into a product designed to be deployed, understood, and trusted by operators outside Nautilus.
Taking Control of...Controls
When it came time to select a control platform, the team focused on a short list of priorities: ease of development, clear data visibility, and reliable communication with building management systems and supervisory software.
“We needed something that was easy to work with, gave us clean access to the data, and integrated smoothly with the rest of the building,” says Roca.
Nautilus had already used Opto 22’s
groov EPIC platform for infrastructure monitoring at the Stockton data center, where it proved reliable. Familiarity helped, but it was not the deciding factor.
What mattered were
groov EPIC’s built-in tools.
Node-RED,
CODESYS®,
Ignition Edge®, and secure REST API access allowed engineers to build and adjust control logic directly, without additional middleware or toolchains. Development could happen quickly and in place, without unnecessary complexity.
Opto 22’s manufacturing presence in California also provided a supply chain Nautilus trusted.