Original film / Telecom Atlas

Data center cooling: the physical path of heat

A teaching film about air cooling, cold plates, coolant distribution, immersion and the path from silicon to heat rejection.

Original educational film with AI generated narration. It is a conceptual overview, not CFD analysis, equipment selection advice or installation guidance.

Nadim Najjar / Original educational work

The sequence, in words

A descriptive summary of the film, not a word-for-word narration transcript.

  1. The film begins with the reason cooling exists: electrical work inside computing equipment becomes heat that must leave the system.
  2. Air cooling carries heat from components into room air, then through cooling equipment toward an external rejection path.
  3. Direct liquid cooling places a cold plate close to high heat components. A coolant distribution unit separates and manages loops.
  4. Immersion places equipment in a suitable dielectric fluid. The film compares the physical paths without declaring one method universally best.
  5. The final section follows heat beyond the rack and explains that outdoor conditions, facility design, maintainability and workload all affect the choice.
Read the full narration transcript

English narration; Arabic subtitles are available in the player.

Data centers use two main cooling approaches: air and liquid. Many facilities combine both. The difference starts with what carries heat away from the electronics. First, air cooling. Fans move air across metal heat sinks, which spread heat from hot components. The warmed air carries that heat away. Liquid cooling includes three arrangements: rear-door cooling, direct-to-chip cooling, and immersion. Immersion has two branches: single-phase and two-phase. The rear door is a hybrid: air transfers heat to water inside a coil. Water stays liquid, never touching the electronics, and carries the heat onward. Direct-to-chip cooling uses a metal cold plate against the chip. Here, water-based coolant stays liquid inside its channels, not around the board; coolant choice varies by design. A coolant distribution unit, or CDU, transfers heat to a separate facility loop. In this depicted unit, heat crosses an exchanger wall; the liquids never mix. Immersion surrounds compatible electronics with purpose-designed dielectric fluid, not ordinary water. Dielectric means electrically insulating: the fluid can still carry heat. Single-phase immersion can use specially formulated mineral-oil or synthetic-hydrocarbon coolants. The fluid touches the electronics, absorbs heat, and stays liquid during operation. Here, circulation takes heat to an exchanger and a separate cooling loop. Two-phase immersion uses a suitable low-boiling dielectric working fluid around the electronics. At hot surfaces, liquid boils into vapor. A separately cooled surface, called a condenser, turns the vapor back into liquid. Droplets return to the bath, while the separate cooling circuit carries heat away. Boiling temperature depends on the fluid and pressure. Historically, Novec six forty-nine boils at forty-nine degrees Celsius at seven hundred sixty millimeters of mercury. That is a fluid property, not every tank or chip temperature. Every route needs onward heat removal, usually to the outdoors. Where conditions allow, some collected heat can instead be reused for heating. Compare each route with the same questions: what fluid carries heat first, what touches the electronics, does it change phase, and where does the heat go next?

What to take away

Follow the boundary between an observation and a completed change. A visual sequence makes the mechanism easier to discuss; a real design still needs site-specific engineering, operating procedures and verification.

Related reading: why lifecycle management is a data problem

Back to all films