ZutaCore Unveils Two-Phase Boiling Cold Plates as Thin as 11.6 mm

ZutaCore has launched a two-phase boiling technology built into cold plates as thin as 11.6 mm and as light as 300 g. The company says the design keeps the hottest AI chips — and the hot spots that shift within them — at a stable, uniform temperature inside a form factor that fits a single-wide PCIe slot.

The technology, called Two-Phase Matrix Flow Boiling, sits inside ZutaCore's OmniTherm family of cold plates. It went public alongside deployment plans for this quarter.

How the boiling matrix works

The design rests on a manifold that feeds a matrix of small boiling regions. Common headers push fluid straight down onto many of these regions at once. Vapor escapes straight up from each one.

ZutaCore frames this as a combination of two known cooling approaches. Pool boiling is stable but hard to scale. Flow boiling scales well but is less stable. The company says Matrix Flow Boiling captures the stability of the first and the scalability of the second.

The underlying system is closed-loop two-phase cooling. The coolant is allowed to boil and then condense, changing its phase. Phase change removes heat more effectively than simple heat conduction through a single-phase loop.

What does the cooling respond to?

The short answer: the location of heat creation itself. Because the manifold feeds many boiling regions in parallel, cooling action concentrates where the silicon actually generates heat, including hot spots that move across a die during operation.

My D. Truong, chief product and technology officer at ZutaCore, laid out the system-level framing: "AI cooling is now an end-to-end system challenge: you have to cool the silicon while keeping entire racks and rows stable."

He added: "Matrix Flow Boiling combines the stability of pool boiling with the scalability of flow boiling, so cooling responds exactly at the location of heat creation. OmniTherm brings that to today's GPUs, scales to the multi-kilowatt chips on the roadmap, and gives us a scalable technology path into the package and directly to the silicon."

What does this change for operators?

The immediate deployment target is concrete. ZutaCore said OmniTherm deployments are underway this quarter for the Nvidia RTX PRO 6000 Blackwell Server Edition.

The physical specifications matter for rack engineering. The OmniTherm hybrid copper-aluminum cold plates can be as thin as 11.6 mm and weigh as little as 300 g. That thinness is what allows the plate to fit within a single-wide PCIe slot rather than forcing bulkier server layouts. The stated scalability path points toward multi-kilowatt chips on vendor roadmaps and, eventually, cooling built into the package and directly onto the silicon.

Who backs the company

ZutaCore's investor list includes names from both IT infrastructure and HVAC. Taiwanese IT infrastructure provider Wiwynn and HVAC firm Carrier are investors in the company. Earlier this year, Mitsubishi Electric, Carrier Ventures, and Samsung Electronics also invested in the firm.

That investor mix signals where two-phase cooling sits in the market. It spans data center server design on one side and building-scale thermal systems on the other.

The takeaway

The launch is a specification story more than a research announcement. An 11.6 mm, 300 g cold plate that boils coolant in a parallel matrix, ships this quarter for a named Blackwell GPU platform, and claims a route to direct-silicon cooling. Operators running dense AI racks now have one more direct liquid cooling option whose value proposition is stability under moving hot spots — at a size that does not force a redesign of the PCIe slot layout.