Engineered for the GPU roadmap ahead
Designed for the 2,000 W+ TDP (thermal design power) class of chips arriving with NVIDIA Rubin Ultra, Feynman, and next-generation AMD silicon. The thermal envelope your next data center will be designed around.
Two-phase direct-
to-chip cooling.
Orbital's two-phase direct-to-chip cooling system, designed for Q4 2027, with AI compute at 2,000 W+ TDP (thermal design power). Built around a novel PFAS-free dielectric refrigerant, in a pumped, closed-loop architecture that uses phase change to cool the silicon densities single-phase simply cannot reach.
Key features
Chemistry and hardware, designed together
Orbital designs the refrigerant, the cold plates, the manifolds, and the rack-level architecture as a single integrated system, tuned to the thermal envelope of the next generation of GPUs.
Discovered with Orbital's agentic operating system
The novel refrigerant is the output of Orbital's agentic operating system, an integrated AI platform that unites domain-specific AI across materials science, hardware design, and manufacturing. Thermodynamic property models screened candidate molecules for boiling point, thermal conductivity, dielectric stability, and safety. Molecular dynamics simulations from Orb, our open-source AI simulation model family for advanced materials, validated the shortlist.
System specifications
Orbital's Two-Phase Direct-To-Chip system is designed for the rack densities, GPU thermal envelopes, and cooling overheads the next generation of AI compute will require.
Thermal envelope
0
W+
per GPU
Designed to cool the silicon classes arriving beyond Rubin Ultra and Feynman, a thermal envelope no current cooling architecture can reach.
Climate range
−0°C
to
+0°C
ambient
Designed to operate across the climate range data centers actually deploy in, from cold-climate facilities to high-ambient regions. No specialised site conditioning required.
Cooling overhead
~0.00
PUE
The cooling overhead the system is designed to deliver. Every watt spent cooling is a watt that can't be spent on compute.
Release and deployment
Our Two-Phase Direct-To-Chip release is aligned to the arrival of NVIDIA Rubin Ultra, Feynman, and next-generation AMD silicon.
Customers planning facilities for the 2,000 W+ TDP silicon generation can engage now to align roadmaps.
Talk to our team
Tell us your deployment timeline, site constraints, and target rack density. We'll come back with a technical scoping session and a proposal designed for your deployment.