Research proposal, EGH404 Research in Engineering Practice, 2026
Where single-phase cooling runs out
Single-phase liquid cooling is validated to 52 kW per rack. Next-generation AI racks need 180 to 220 kW. I proposed a three-scale experiment to measure where single-phase stops being enough, and whether two-phase cooling extends the limit.
- Gap
- No study compares both modes on one boundary above package scale
- Method
- 270 controlled runs, an 8-branch rack-slice, then a live rack
- Preview
- Single-phase limit of 1.13 kW per GPU at the worst case, on synthetic data




