Airsys LiquidRack is a new, next-generation rack-level liquid cooling platform aimed at mid-density AI, data center, telecom, and edge deployments, with the company now taking candidates for pilot deployments. The pitch is a simpler liquid cooling architecture intended for retrofits and distributed sites where operators want higher rack density without committing to large, centralized liquid cooling infrastructure.
LiquidRack is built around a rack-level approach that integrates fluid distribution, pumping, and control at the server level, rather than using shared-tank immersion or direct-to-chip systems that depend on more complex facility piping. Airsys also describes the design as “hazardous free” and “maintenance friendly,” with a goal of making upgrades from air to liquid cooling more straightforward and lower cost.
On the thermal side, Airsys says LiquidRack is designed for high-temperature operation with dry coolers, enabling compressor-less architectures intended to reduce mechanical complexity, lower energy consumption, and eliminate water usage. For facilities teams, that’s a consequential combination if it holds up in the field, because it directly affects the cooling plant footprint, maintenance profile, and how much site capacity can be devoted to IT rather than mechanical systems.
For IT load sizing, Airsys puts LiquidRack in the 0.5–8 kW per server range and up to 80 kW per rack, positioning it for CPU-focused and mid-density AI deployments and distributed workloads. Heat rejection options span dry coolers, adiabatic systems, and traditional chiller-based architectures, which is meant to let operators integrate the platform into existing facilities or use it in rack-level architectures for AI factory environments.
At the server level, Airsys describes a “spray-based” cooling approach intended to deliver targeted heat removal from CPUs and GPUs while keeping thermal management localized at the rack. The company also claims reduced fluid handling compared with immersion, stating LiquidRack uses up to 80% less dielectric fluid than immersion systems. The hardware format is described as a 2U cassette-based design with integrated pumping and a serviceable architecture, targeting incremental rollout in retrofit and new-build environments.
One blunt takeaway: the stated ceiling of 80 kW per rack squarely targets the messy middle of deployments that aren’t hyperscale training clusters, where space, staffing, and mechanical retrofit constraints tend to be the limiting factors—not the availability of exotic cooling expertise.
“Most data center operators are looking to a practical, flexible and seamless path from air to liquid cooling to meet the challenge of high density, stranded power and ESG goals,” said Yunshui Chen, Founder and CEO of Airsys. “By integrating key cooling functions at the rack, we simplify deployment and enable customers to scale performance within the constraints of their existing infrastructure.”
On density expectations, Alex Cordovil, Research Director at Dell’Oro Group, said, “A meaningful share of enterprise and edge workloads is expected to sit within the 40–80 kW-per-rack range, inside facilities that were never designed for liquid cooling. Rack-level liquid cooling architectures can offer a pragmatic path to higher densities without the infrastructure overhaul that full retrofits require.”
Airsys also tied LiquidRack to its Power Compute Effectiveness (PCE) concept, with Dr. Rand Talib, Research Analyst at Uptime Institute, saying, “PCE is a metric developed by cooling system provider Airsys to provide greater transparency into power allocation within a data center’s provisioned electrical envelope. Cooling architecture decisions directly influence how much of the provisioned power envelope is allocated to facility systems and how much remains available for IT compute.”
