Two-phase cooling for AI and HPC
Thermal Intelligence for High-Density Data Centers
We’re making two-phase cooling deployable in the data centers you already have. STAR, our Stabilized Two-Phase Architecture, is how we do it.
The problem
The Cooling Ceiling
Compute is outgrowing the cooling around it. Rack power is climbing from tens to hundreds of kilowatts. Two-phase cooling, which absorbs heat by boiling, can carry that load; the hard part is deploying it reliably at rack scale.
Cooling power
Removing hundreds of kilowatts per rack strains facility power and water budgets.
Deployable capacity
Installed power stays stranded when racks can’t be cooled reliably under real AI workloads.
Deployment speed
Custom cooling loops and facility rework add months to every new site.
Our technology
Meet STAR
Our Stabilized Two-Phase Architecture (STAR) keeps two-phase cooling stable under fast-changing GPU loads, and fits the racks and CDUs data centers already run.
Design targets.
How we build
One Integrated Design Loop
Analysis, testing and expertise work as one. The PhaseQ team links system-scale analysis, component-scale analysis and experiments, so each informs the others and every design decision draws on all three. Agentic AI speeds up the routine work; our engineers make the calls.
Path to deployment
- Now
Modeling and design
System- and component-scale analysis, in-house
- Next
Lab-scale validation
Test data to confirm and refine the models
- Then
Rack-scale proof of concept
STAR running on existing infrastructure
- Later
Deployment
With data center partners
The founders
From Nuclear Reactors to AI Racks
In light-water reactors, boiling heat transfer keeps the core cooled and defines its safety margins. Two-phase cooling for AI racks depends on the same physics.
PhaseQ was founded by two nuclear engineering Ph.D.s with over 30 years of combined experience, from fundamental boiling physics to advanced reactor system design.
Partner with us
Deploy with Us
Let’s deploy it together. We’re looking for partners to help validate and deploy STAR.
- Data center operatorsPlanning high-density AI capacity
- GPU cloud and HPC teamsRunning racks at the edge of what cooling allows
- Hardware and CDU partnersBuilding the systems STAR plugs into