ZutaCore Z2O™

Zero H2O. near-Zero cooling energy.

The waterless cooling architecture for next-generation AI.

1.03
Annualized cooling pPUE, modelled1
Zero
Cooling water used
~97%
Of your power goes to compute, not cooling1

What Z2O changes for your data center.

Power is now the ceiling on AI, and water is the next constraint. Z2O gives more of your power to compute and takes water out of cooling altogether.

ENERGY

More power for compute.

More of your secured power goes to compute, not cooling.

Less power on cooling.
3% cooling energy
Cooling uses just 3% of IT power, in annual modelling.1
75% less cooling energy
Than best-in-class chilled-water cooling, in the same model.1
97% to IT
Of IT and cooling power goes directly to IT.1
No chiller plant
Nothing to purchase, power or maintain.
ZUTACORE Z2O™ IS

A unified loop, from chip to outdoor air.

Four steps, one sealed loop: ZutaCore direct-to-chip cold plates inside, Munters SyCool heat rejection on the roof.

01 · LIQUID IN

Liquid reaches every chip

Heat transfer fluid flows through rack manifolds to the cold plates in every server.

02 · BOILS

It boils on the chip

The fluid absorbs heat by boiling, at one constant temperature.

03 · VAPOR OUT

Vapor rises to the roof

Vapor carries the heat outdoors, largely by buoyancy and gravity. Pumps assist and add redundancy.

04 · CONDENSES

Outdoor air condenses it

Munters SyCool units condense the vapor with outdoor air for most of the year, with active assist on the hottest days. The liquid returns to the servers.

100%

All IT heat is captured and cooled.

No facility water loop
No chiller plant
No water-based cooling towers
No liquid-to-liquid heat exchanger
No CDU
WATCH
FUTURE-READY

Chips evolve. The loop stays.

Z2O is fungible thermal infrastructure, built for successive silicon generations. Upgrade supported cold plates on the same loop, with no re-plumbing.

TODAY
Field-proven HyperCool®
Running NVIDIA HGX B300 servers in the launch instance.
NEXT
OmniTherm™ Matrix Flow Boiling
Next-generation cold plates. Same fluid, same loop.
BEYOND
Future qualified generations
Thermal headroom for what comes next, on the same loop.

More tokens per second, per dollar, per megawatt.

More tokens per second from every megawatt secured, with less cooling infrastructure to build and operate. Where power is not the constraint, lower cooling energy reduces operating cost.

01 · MORE TOKENS

More compute, same grid.

More of a fixed grid connection goes to compute, so every megawatt you secure produces more tokens.

02 · CAPITAL COST

Less to buy and install.

Much of the mechanical plant behind a conventional liquid-cooled hall is simply not there to buy, install or commission.

03 · OPERATING COST

Less to run.

Near-zero cooling energy, and no water to buy, treat or maintain.

WHY TWO-PHASE

What two-phase does that single-phase can’t.

Z2O boils the fluid on the chip. Single-phase only warms it. That one difference shapes the whole system.

Z2O TWO-PHASE
SINGLE-PHASE
Temperature across the row
Constant. The last server runs as cool as the first.
Rises as the coolant flows, so hotspots are harder to manage.
Flow and pumping
Lower flow, because boiling carries the heat.
Higher flow rates and more pumping power.
Path to outdoor air
One loop, from chip to outdoor air.
Multiple loops, with heat exchangers between them.
AI operators have fought hard for every megawatt, and they should not be spending it on moving heat. Z2O takes the whole thermal stack, from the chip to the outdoor air, and collapses it into one two-phase loop. That is only possible with two-phase.
Dan Wood, Chief Product Officer, ZutaCore
Pairing Munters SyCool with ZutaCore's two-phase loop removes the water and much of the mechanical plant that AI data centers have relied on. We are proud to be ZutaCore's launch partner and to bring the first Z2O instance to market.
Keith Dunnavant, VP Offer Strategy & Portfolio Management, Munters
PLATFORM AND PARTNERS
MULTI-VENDOR

One architecture. Designed to scale.

Freedom to go waterless without being tied to a single vendor over time.

Standard racks, supported servers
Standard racks and supported servers from multiple manufacturers* run on the same Z2O architecture, with ZutaCore OmniTherm™ or HyperCool® cold plates, within qualified configurations.
Agnostic thermal infrastructure
One loop across supported server vendors and cold-plate generations, so you can change the IT without rebuilding the cooling.
Modular, with no fixed ceiling
No chilled-water plant or facility water to build. Repeat building blocks in prefab, modular or conventional halls to scale within validated design limits.

*First server partner: ASRock Rack's NVIDIA HGX B300 server.

LAUNCH PARTNER

Brought to market with Munters.

Munters SyCool is the heat rejection in the first Z2O instance, and Munters is bringing Z2O to market with us as launch partner.

FAQ

Questions we hear most.

TAKE THE NEXT STEP

See what Z2O does for your site.

Book a meeting

Run your own site, power and climate through the Z2O numbers with a ZutaCore engineer.

BOOK A MEETING

Reference architecture brief

The building block, the loop, the launch configuration and how it scales, in a short technical summary.

GET THE BRIEF

Become an ecosystem partner

Build heat rejection, servers, racks or modular data centers on the Z2O architecture.

BECOME A PARTNER

1 Annualized pPUE 1.03, modelled on the launch reference design with Munters (ASHRAE 2026 one-pager): hour-by-hour Dulles, VA bin data, 43 °C saturation temperature, 80/20 liquid/air split, 100% IT load. In the same model, best-in-class chilled water has an annualized pPUE of 1.115; Z2O uses about a quarter of its cooling energy. Power figures per 10 MW are arithmetic on the modelled pPUE. OmniTherm™, Two-Phase Matrix Flow Boiling™ and HyperCool® are trademarks of ZutaCore, Inc. Munters® and SyCool® are registered trademarks of Munters.