Advanced electric drive. Intelligent control.
COOL•E integrates Orbis's patented electric drive technology with intelligent software controls to efficiently circulate liquid coolant through AI server cooling systems. The technology powers Orbis's own cooling units and can be integrated into cooling equipment from other manufacturers.
Engineered for both conventional AC power and emerging 800 V DC data center infrastructure, COOL•E enables operators to transition from AC to DC without replacing their cooling equipment.
AI demands more power. Cooling should use less.
Every data center gets a limited amount of electricity from the utility, and getting more can take years. Cooling and the AI servers share that supply, so every megawatt spent on cooling is a megawatt that can't earn money. COOL•E runs the cooling pumps with a more efficient kind of motor and smarter controls, so cooling takes less power and more of the site's existing power goes to the servers.
Recover power. Unlock revenue.
COOL•E E turns cooling power into computing power. Each megawatt it frees up can run AI servers worth an estimated $10M to $27M a year to the operator.
800V DC-Native
ASHRAE TC 9.9 A3/A4
NVIDIA 45°C Ready
Control Logic SW
Less electrical gear to buy, install and wait for
Every piece of equipment needs a circuit built big enough for the most current it can draw. COOL•E's cooling system needs about 80% less of that capacity than conventional cooling. That means thinner copper wiring (up to about 80% less copper), smaller breakers, panels and switchgear, and less power lost as heat in the wires.
Conventional systems also have to be wired for their worst day, when hot weather pushes the motors harder. COOL•E doesn't, so electricians size for normal running. The electrical design is simpler and the build goes faster.
Size
Single unit, beside or at the end of a row of server racks
Cooling capacity
2.0 MW
Electrical current needed
27 AMPS
Copper wiring saved
70 to 80%
Size
Pod: 12 units on one prefabricated platform
Cooling capacity
24 MW
Electrical current needed
588 AMPS
Copper wiring saved
80 to 85%
Built for NVIDIA's next generation, not just this one
NVIDIA's next AI system, Rubin, does roughly 10 times the work and needs much more coolant. That makes every freed-up megawatt worth even more. Conventional systems meet the extra demand by running the same motors harder, so their power needs climb fast. COOL•E adds one more motor instead, with each motor running at its most efficient speed, so its power needs rise slowly.
What Cool-E frees up in a 100 MW data center:
1 MW/YEAR= $10-27M
NVIDIA GB200
COOL•E DELIVERS:
3+ MW
Power freed for servers
3+ MW
21+ racks
Extra server racks it can run
$70-189 M
Yearly value of that extra computing
Yearly value of that extra computing
$30-81 M
NVIDIA RUBIN
COOL•E DELIVERS:
8+ MW
8+ MW
Power freed for servers
34+ racks
NVIDIA GB200
NVIDIA RUBIN
3+1 (N+1)
METRIC
Extra server racks it can run
Recover Power Without Adding Capacity
Think of a data center's electricity as a fixed budget that cooling and servers share. The electrical code requires every motor circuit to be built 25% bigger than what the motor is rated to draw. Conventional cooling motors are rated well above the job they actually do, so their circuits are big and the safety cushion on top is big too. The site pays for all of that capacity but never uses it.
COOL•E's motors are rated for the job they actually do, so their circuits are far smaller. The difference goes to servers, without asking the utility for more power.
The more powerful the chips, the more COOL•E saves
Rubin needs much more coolant than GB200. Conventional systems get it by pushing the same motors harder, which nearly doubles the electrical capacity they need (from 157 to 314 amps). COOL•E adds another motor running at its best speed, so its need rises only from 27 to 49 amps.
COOL•E stays about 83 to 84% below the conventional system either way. What changes is the size of the saving: 130 amps with GB200, 265 amps with Rubin. In a 100 MW data center, that is the difference between $30M to $81M a year of extra computing with GB200 and $70M to $189M with Rubin.
787
2+1 (N+1)
same
157 A
787 GPM
same
80% smaller and lighter than induction motors
476 GPM
about 84% less
same
83% less
Motors running + spare
Coolant pumped (gallons per minute)
Electrical capacity Cool-E needs
Size and weight
COOL•E needs less Amperage
Electrical interference
Cooling capacity
Conventional system (estimate)
476
2 MW, coolant brought within 2.4°C of the building's water
Runs on 800 V DC; AC supply optional
about 314 A
Power
same
Certified to CISPR 25, the auto-industry standard
DC-native cooling. Same footprint, less amperage, more compute.
Less electrical gear Smaller wires, breakers and switchgear, and a faster build.
Works on whatever power the site has The same hardware runs on standard AC or on DC. Start on AC, move to DC when the site is ready, and the cooling units stay as they are. No power is lost converting AC to DC, and no cooling equipment is left obsolete when the site's power system changes.
Each gallon carries more heat COOL•E lets the coolant warm up more on each trip through the servers (15 to 20°F, against 10 to 15°F for conventional systems). Each gallon carries more heat away, so there is less liquid to pump, the pumps use less power, and more of the site's power is left for computing
Same savings at any size One unit beside a row of racks or a full 24 MW pod: same design, same efficiency. Standard modules make design and installation faster, and power that would have been lost in drives and extra copper goes to more working GPUs.
White Paper: Unlocking Stranded Power Inside the Cooling Plant
