The AI revolution runs hot. We keep it in control.
AI has a thermal bottleneck. Conventional cooling is not solving it.
+250%
GPU power consumption in the next two years.
<30°C
Coolant required by conventional cooling — costly and unsustainable at scale.
$1M /MW
To upgrade cooling infrastructure as the technology evolves.
60–70%
Temperature restricts GPU throughput by 60 to 70% of peak performance.
0
Current liquid cooling solutions treat all chips identically, unable to adapt to chiplet complexity or real-time thermal spikes.
Smart thermal control from chip to data center.
workloads for peak performance and efficiency, modular and built for the
future of AI.

DaTEG®
DaTEG is an active cold plate that sits directly on the processor.
Supervisor™
Supervisor is the software intelligence of the ecosystem.
%201.png)
Not a better cold plate.
Warm‑Fluid Operation (>43°C)
Operates without chiller pre-cooling. This removes a major infrastructure cost and a significant point of energy waste from the data center.
40% higher efficiency
Eliminates the use of chillers. The result is up to 40% lower energy consumption across the entire installation.
Future‑Proof Architecture
Designed for three to five chip generations with input from chip manufacturer. Reduces infrastructure investments as technology evolves.
Differential chiplet-level control
Manages HBM and compute zones independently and simultaneously. Conventional cold plates treat the entire chip surface as a single thermal zone. This distinction matters more with every new chip generation.
One integrated ecosystem.
Hardware, software, and a development tool are designed together. Not a point solution added to someone else's infrastructure stack.
Patented. Tested. Deployed.
US Patent
Granted protection for GemaTEG's differential thermal control architecture, covering independent temperature management of HBM and compute zones. Worldwide patent pending.
Tested for over a year
DaTEG 2.0 was validated through extended real-world testing before commercial announcement. We chose readiness over rushing to market.
Deployed with organizations that cannot afford failure
Mission-critical HPC under strict energy and sustainability requirements
Validated on an NVIDIA GPU platform under the SEED program
Maintaining temperature within ±0.5°C for particle physics workloads; no other solution met this specification
