Trane Technologies is betting that one coolant distribution unit can serve multiple generations of AI silicon. At Yotta 2026 in Swords, Ireland on Sept. 29, the company announced the LiquidStack CDU 2.X platform, an architecture-agnostic CDU rated for flow of up to 3,750 liters per minute at 3.5 bar — a figure the company says covers today's leading GPU platforms while leaving headroom as rack densities climb.

The core commercial argument is risk reduction. As GPU roadmaps shift on short cycles, data center operators face redesign costs each time a new platform lands. CDU 2.X addresses this with a single CDU design that supports multiple current and emerging AI silicon platforms, including NVIDIA's Vera Rubin generation and beyond, from a common architecture.

"Operators need cooling infrastructure that can adapt as GPU platforms and rack densities evolve," said Scott Smith, General Manager of LiquidStack, Trane Technologies. "CDU 2.X combines the performance and flexibility customers need today with the headroom to prepare for what comes next, allowing them to configure cooling around their facility and deployment strategy rather than designing the facility around the CDU."

That last point is deliberate positioning: LiquidStack wants the CDU to adapt to the facility, not the reverse. To that end, the platform offers flexible control valve, power-feed and redundancy options, and supports both end-of-row and rack-adjacent deployment — giving operators a choice in how they distribute liquid cooling as rack densities rise.

Power quality engineered in

One design detail stands out for operators watching electrical balance-of-plant costs. CDU 2.X uses an ultra-low-harmonics variable frequency drive (VFD) architecture that addresses harmonic distortion at the source. That reduces the need for additional power-quality components elsewhere in the system, simplifying both system architecture and deployment.

On the power side, dual-feed A/B and automatic transfer switch options let operators match the CDU configuration to their facility's resiliency requirements.

Efficiency also factors into the design. Lower approach temperatures and support for facility inlet temperatures up to 45°C enable more efficient liquid cooling system architectures — a meaningful lever for operators running high-density AI halls where cooling energy is a significant share of operating cost.

Building out the portfolio

The announcement follows the commercial availability of the LiquidStack GigaModular CDU platform, which Trane has expanded to 14 MW of capacity. Together with CDU 2.X, the LiquidStack portfolio gives data center operators flexibility in how they deploy and scale next-generation liquid cooling infrastructure — from modular high-capacity units to flexible distribution architectures.

Trane Technologies, which acquired LiquidStack and now runs it as a brand, has been steadily deepening its investment in advanced thermal management for high-density data center environments. The move positions the HVAC incumbent against specialist liquid-cooling vendors chasing the same hyperscale and AI infrastructure buildout.

Availability

The LiquidStack CDU 2.X is available for pre-order now, with shipments beginning in the second quarter of 2027. That timing puts delivery roughly in line with the ramp of next-generation GPU platforms the unit is designed to cool, meaning operators planning 2027 deployments can now specify a CDU intended to span both current silicon and what follows it.