Cooling AI & HPC Infrastructure - Wherever You Deploy

WATERLESS
CHIP-LEVEL COOLING

ZutaCore HyperCool® delivers advanced, waterless cooling designed for modern AI data centers and high-performance computing (HPC) environments. Built around a two-phase, direct-to-chip architecture, it enables efficient thermal management for high-power CPUs and GPUs from AMD, Intel, and NVIDIA.

How Two-Phase Cooling Works

At the chip level, HyperCool uses a dielectric heat transfer fluid that absorbs heat through a controlled boiling process. As the fluid vaporizes, it carries heat away from the processor and flows to the Cooling Distribution Unit (CDU), where it is condensed and recirculated.

This closed-loop process enables precise, on-demand cooling that responds instantly to changing workload conditions—making it ideal for High Frequency Trading (HFT), AI training, inference, and HPC applications with sustained compute demand.

Engineered for Next-Generation Compute

As processor power continues to increase, traditional cooling approaches struggle to maintain performance without throttling. HyperCool is designed to support these next-generation requirements, enabling:

• Sustained operation of high-power CPUs and GPUs

• Stable thermal performance under continuous load

• Higher rack densities in AI and HPC deployments

By removing thermal constraints, HyperCool allows compute infrastructure to operate at its full potential.

GPU Cooling for AI PCIe Servers and HPC Workloads

ZutaCore OmniTherm is a two-phase, direct-to-chip cold plate designed for PCIe-based AI and HPC servers, enabling efficient cooling of high-power GPUs without requiring changes to standard server architecture. By cooling both the GPU die and high-bandwidth memory, it supports sustained performance at up to 600W per GPU, while reducing dependence on airflow and fan power.

• Single-slot compatibility for standard PCIe servers
• Enables higher GPU density per rack without redesign
• Maintains stable performance under sustained AI and HPC workloads
• Reduces airflow requirements in high-density environments

In-Rack Cooling Distribution Units (CDUs)

ZutaCore’s in-rack Cooling Distribution Units (CDUs) manage cooling within each server rack by circulating dielectric fluid between the cold plates and the heat rejection system.

They are designed for AI data centers and HPC environments where rack power densities exceed the limits of air cooling.

Rack Cooling Capacity

• 20kW per rack - Air-based CDU, no facility water required

• 60kW per rack - 3U facility water CDU

• 70kW per rack - 6U facility water CDU

• Up to 120kW per rack - dual 3U facility water configuration

Installs directly in standard 19" racks and scales linearly from single rack to clusters.  Seamlessly integrates with the 1.2MW end-of-row cooling for larger deployments.

End-of-Row Cooling Distribution Units (CDUs)

The End-of-Row CDU manages heat removal at the row level by receiving vapor from multiple racks, condensing it back into liquid, and interfacing with facility water.

• Aggregates heat from multiple high-density racks

• Condenses vapor back to liquid for reuse

• Transfers heat to facility water outside the white space
  
  

Cooling Capacity and System Performance

• 1.2MW - 2.0MW cooling capacity per unit
• < 0.35 L/kW flow rate for efficient operation
• < 3 bar system pressure
• < 4.5kW system power consumption

Keeps facility water outside the data center white space and includes N+1 redundancy across pumps, power, and controls.

HyperCool Cloud
Monitoring and Optimization

HyperCool Cloud is a cloud-based plaform that monitors and manages ZutaCore cooling sysyems across racks, rows, and sites.  It collects thermal, power, and system data from CDUs and controllers, providing centralized visibility across deployments, and enables remote monitoring and control of cooling performance.

• Real-time monitoring across multiple sites
• Correlates cooling performance with compute output
• API-based integration with server telemetry and DCIM systems
• Secure, encrypted communication with role-based access control
• Enables remote diagnostics and performance optimization
• Integrates with existing infrastructure through standard protocols