OCP Global Summit 2026: ZutaCore Unveils Two-Phase Matrix Flow Boiling to Cool the Hottest AI Silicon

ZutaCore, a player in waterless, direct-to-chip, two-phase liquid cooling, unveiled Two-Phase Matrix Flow Boiling, the technology inside its OmniTherm family of cold plates, at OCP Global Summit 2026.Built on a manifold that feeds a matrix of boiling regions, it holds the hottest AI chips – and the hot spots shifting within them – at a stable, uniform temperature, all from a cold plate thin enough to fit within a single-wide PCIe slot.

AI is rewriting the thermal rulebook. Chip power has roughly tripled in three generations, multi-kilowatt packages are arriving, and hot spots now move in milliseconds as workloads shift.

Two-phase cooling ushers in an era by reducing the complexity of increasing flow volume, mixed- material compatibility, and long-term OpEx associated with air-cooling and single-phase solutions – both of which are reaching their limits. Two-phase solutions can improve TCO by reducing flow volumes, weight, and OpEx while improving AI token economics.

Two-Phase Matrix Flow Boiling brings together the stability of pool boiling with the scalability of flow-boiling. With this approach, common headers feed fluid straight down onto many small boiling regions at once, while vapor escapes straight up from each one, delivering the benefits of both pool and flow boiling. Every region gets fresh liquid, the hottest regions automatically boil harder, and the cold plate works in any orientation.

Matrix Flow Boiling is already at work. OmniTherm deployments are underway this quarter for the Nvidia RTX PRO 6000 Blackwell Server Edition, delivering waterless two-phase cooling that increases GPU density and reduces operational burden at scale.

Why it matters for the data center

• Full performance, no throttling. Compute, memory and I/O are held at one consistent temperature as workloads shift, so GPUs deliver the performance they were acquired for
• Highest-density compute. The OmniTherm architecture allows cold plates as thin as 11.6 mm, opening up single-slot, sub-1RU and backside-cooled designs. It also enables hybrid copper-aluminum construction, bringing weight down to as little as 300 g (less than 1lb) with no risk of galvanic corrosion
• Waterless by design. The sealed dielectric fluid poses no risk to IT equipment, while low flow rates keep cooling power down
• Drops into proven infrastructure. As the next-generation cold plate for ZutaCore HyperCool, OmniTherm uses the same fluid, connections, racks and coolant distribution units already proven across more than 75 deployments, with no re-architecture required
• Built for the chips coming next. The same architecture scales beyond the multi-kilowatt packages on industry roadmaps and is designed to move into the chip package and, ultimately, onto the silicon itself. Development program is under way with partners

“AI cooling is now an end-to-end system challenge: you have to cool the silicon while keeping entire racks and rows stable,” said My D. Truong, CPTO, ZutaCore. “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.”