Layer 1 · How the heat leaves
Two-phase immersion
Extreme density research deployments.
The shape of it
- Capital cost
- Highest
- Maturity
- emerging
- Density supported
- Highest available
- PUE
- Approaching 1.02
Verified 2026-09-06.
The catch
The fluids are expensive and several are under regulatory pressure as PFAS rules tighten. That is a supply and compliance risk sitting on top of an already niche choice.
What it actually decides
Thermally excellent, commercially awkward. Watch the fluid regulation before committing a fleet to it.
What this is powering
A GB200 NVL72 rack draws 120–132 kW and a GB300 pushes 135–200 kW, against a 2026 average rack of about 27 kW. The silicon decision and the power decision are the same decision, made eighteen months apart.
Layer 2 — Silicon · GB200 NVL72 · The 150 W inference option
Weighed against
- Two-phase immersion vs Single-phase immersion — Marginal thermal gain against a live PFAS regulatory question.
Others in how the heat leaves
- Air cooling — Up to ~30 kW per rack
- Rear-door heat exchangers — ~30-50 kW per rack
- Direct-to-chip liquid cooling — 50-200+ kW per rack
- Single-phase immersion — 100+ kW per rack equivalent
- Free and evaporative cooling — Depends on the paired system