UniSCool
Open server with UniSCool copper cold plates mounted directly on the processors and coolant hoses running between them

Liquid cooling for 2000 W chips

Smart Cool Fins sit directly on the processor and adjust to where the heat actually is. Patented in Lleida, tested for five months in a live server room.

See the fins adapt

EIC Accelerator 2026
OCP Global Summit 2025, FTS winner

Smart Cool Fins

Patented direct-to-chip technology

Inside the cold plate, self-adaptive fins change their position with the local heat load, in time and in space. More coolant contact where a core is working hard, less resistance where it is not. Scroll to watch it happen on a simulated die.

Infrared thermal map of a processor die showing heat spread across the cores

A chip under AI load

Cores that work hardest become hotspots. Without enough heat extraction they set the limit on how hard the whole chip can run.

Fixed fins, uneven die

A conventional cold plate treats every square millimetre the same. Coolant warms as it crosses the chip, so downstream cores run hotter.

Fins lift where it is hot

Smart Cool Fins rise over the hotspots and lie flat where the die is cool. Heat is pulled out locally and the temperature evens out.

The load moves, the fins follow

AI workloads shift from core to core. The self-adaptive fins respond as the load shifts. Move your cursor over the die to shift the load.

Hottest point
96°C
Spread across die
58°C
Die temperature
2565105 °C

Illustrative model of the principle, not measured data.

Five months in a live server room

70%

less energy spent on cooling with Smart Cool One in real operating conditions.

30%

lower total system energy, measured on the same installation.

2000W

dissipated per socket, for the most power-dense CPUs and GPUs.

Smart Cool One is modular and plug-and-play, which makes the move from conventional cooling to liquid straightforward.

Smart Cool One

Where the heat is going

Data centers already spend up to 40% of their electricity on cooling. Rising power density is pushing operators to single-phase direct-to-chip liquid cooling, from AI clusters to the chip itself.

Server board with UniSCool cold plates on each processor socket
AI & HPCUp to 2000 W per socket for GPU and CPU clusters.
UniSCool test installation with transparent coolant lines and a heat exchanger
Data centersUp to 50% energy savings over microchannel and jet impingement plates.
Infrared thermal image of a semiconductor die
SemiconductorsCooling integrated into the silicon for next-generation chipsets.

Battery and power electronics

Electric vehiclesLiquid cooling for performance and longer battery life.

Data center operators

Try & Buy

Run Smart Cool One on your own servers and measure the savings before you commit to a rollout.

Apply for Try & Buy

Server and system OEMs

Custom solutions

Cooling designed around your hardware, your socket power and your coolant loop.

Discuss a design

Chip designers

In-chip silicon cooling

Bring adaptive cooling into the silicon itself for thermal performance at the source.

Explore in-chip
Winners of the 2025 Future Technologies Symposium at the OCP Global Summit, and backed in 2026 by a €2.5 million EIC Accelerator grant to industrialise the technology.
  • University of Lleida
  • Université de Sherbrooke
  • European Innovation Council
  • CDTI · NextGenerationEU
  • ACCIÓ Catalonia
  • The Collider
The UniSCool team in white branded polo shirts

Test it on your own racks

The Try & Buy programme is open. Tell us about your servers and the team in Lleida will size a pilot with you.

Contact the team