Technical Guide · Cooling
Air, Liquid, and Hybrid Cooling for AI Infrastructure
The terms 'liquid-ready,' 'hybrid,' and 'direct-to-chip capable' mean different things. Here is how to read them.
· 8 min read
Loading
WORKLOAD — HPC & RESEARCH
INFRASTRUCTURE CONSIDERATIONS
Low-latency fabrics require careful row and cable planning.
Plan for successive hardware generations in the same footprint.
Parallel file systems sit close to compute.
Phasing aligned with grant and program timelines.
TYPICAL PROFILE
| Typical scale | Hundreds of kW to multi-megawatt |
|---|---|
| Cooling | Liquid-ready / hybrid |
| Deployment models | Dedicated suite, dedicated hall |
SOLUTIONS
RESOURCES
Technical Guide · Cooling
The terms 'liquid-ready,' 'hybrid,' and 'direct-to-chip capable' mean different things. Here is how to read them.
· 8 min read
NEXT STEP