HOSPITAL

Hospitals — on-prem AI without a machine room

22–26 °C ROOMNO WARD WATER LINESDATA STAYS IN-HOUSE

Imaging AI, clinical NLP and research compute next to the PACS suite — in a corridor-side technical room, not a data center.

Worked example · public physics

What this looks like

TDP × 1.2 · room = 1.1 × IT
CLUSTER
2× 8-GPU nodes
IT LOAD (TDP × 1.2)
15.2 kW
ROOM MUST EXTRACT
16.7 kW ≈ 4.8 tons
DELIVERY
1 cabinet

A typical imaging-AI cluster fits one cabinet. The room AC must absorb ~16.7 kW — we check that with you before anything ships; if the room cannot, we say so.

In this room

What this looks like

Quiet by design

Acoustic targets set for clinical-adjacent technical rooms; no construction, no water lines through wards. Acoustic figures published from the demo unit — targets until then, not claims.

Data stays home

On-prem deployment keeps patient data inside the hospital network boundary — the reason these clusters exist at all.

IT-owned, not facilities-owned

Runs on standard building power and AC. No chiller project, no facilities budget fight, no machine-room conversion permit.

What we will NOT claim here

We do not claim medical-area certification (e.g. IEC 60601 environments) until it is tested and documented. Until then: technical rooms adjacent to clinical areas, and we help you place it correctly.

Configure for this scenarioThe cabinet