High Power Densities · Made in Germany
High rack densities require the right cooling method
AI and HPC workloads deliver significantly more performance per rack. Traditional air cooling reaches its technical limits in this scenario, as GPU servers generate a great deal of waste heat in a small space.
Anyone operating GPU racks needs a data center designed for high power densities and appropriate cooling: with a reliable power supply, efficient cooling, high availability, and room for future growth.
GRASS-MERKUR offers colocation for AI, GPU, and HPC workloads at the HAN1 and HAN2 locations.
You choose the cooling method that best suits your kW requirements per rack: from air cooling to rear-door cooling to liquid cooling.
All methods can be combined and scale with your needs. This way, you can start with your current load and expand as soon as your workloads require more power: energy-efficient, scalable, and Made in Germany.
From 35 to 280 kW per rack
Air, rear door, or liquid: The right cooling method for every rack density
The cooling capacity per rack determines which method is appropriate. Three levels cover the entire range:
Chip Cooling
This is how the cold reaches the chip
Coldplate
Cools the CPU and GPU directly.
Rack-Manifold
Distribute the coolant among the servers.
CDU
Separates the circuits and regulates temperature and pressure.
Recooling
Dissipates heat.
Your Benefits
Cooling that scales with your GPU workloads
- Up to 280 kW per rack for GPU and HPC clusters
- Fewer fans, higher energy efficiency
- Higher density with Rear Door, no hardware modifications required
- Air, Rear Door, and liquid cooling - all from a single source
- Consulting on optimal rack density
- Data center in Hanover, Made in Germany
Capacities
Two Locations: Cooling Capacity for Your IT Workload
| Procedure | Locationt HAN1 | Location HAN2 | Total |
|---|---|---|---|
| AirStandard Server, Storage, Network | 1,35 MW | 4,0 MW | 5,35 MW |
| Rear DoorHigh-Density Racks | 1,35 MW | 4,0 MW | 5,35 MW |
| LiquidGPU-, AI- and HPC-Workloads | 1,7 MW | 4,5 MW | 6,2 MW |
+ 4.0 MW available in 2 years. As of 11/2026.
Frequently Asked Questions About Liquid Cooling
Cooling methods, server hardware, and available capacity
We offer three cooling methods: air cooling up to 35 kW per rack, rear-door cooling up to 65 kW per rack, and direct liquid cooling up to 280 kW per rack.
Yes. All three processes run side by side in the same data center. This allows you to operate standard systems and high-density GPU clusters in one location and expand step by step.
Not with Rear Door: You get higher density with the same server hardware—even as a retrofit. For Direct Liquid Cooling, we'll work with you to determine whether your hardware is designed for it.
This is the electrical power we can provide for your IT systems. The electricity used for cooling is not included in your allocation.
As of November 2026, a total of 5.35 MW will be available at HAN1 and HAN2 for air, 5.35 MW for rear door, and 6.2 MW for liquid. An additional 4.0 MW will be available in two years.
We keep a close eye on leaks, temperature, and flow at all times.
Everything from a single source
More than just cooling
Contact
Request a Liquid Cooling Consultation
Please describe your project to us—for example, power requirements in kW, the number of racks, and your timeline.
We’ll get back to you with a detailed assessment of the most suitable cooling method.
Direct Line
GRASS-MERKUR GmbH & Co. KG
Rothwiese 5
30559 Hannover