Phononic Thermal Kits Deliver Superior Al Cooling, Validated by 360Quadrant

Phononic, the preeminent global leader in solid state cooling technology for AI data centers, has expanded its data center cooling portfolio with new solutions for 1.6T pluggable transceivers, along with integrated Thermal Kits for GPU high-bandwidth memory (HBM) and co-packaged optics (CPO). Through these innovations, Phononic is redefining cooling as a real-time, chip-level control layer for AI systems. Purpose-built for modern AI data centers, these high-performance cooling solutions reduce thermal throttling, enhance compute performance, extend device useful life by up to 5×, improve PUE by up to 0.15, and deliver up to 3× return on investment. In addition, Phononic introduced its Thermal Fabric™, that forms a coordinated thermal control network that enables fleet-wide workload optimization across data center infrastructures. Phononic’s designation as a “Progressive Company” in the Data Center Cooling market on MarketsandMarkets’ 360Quadrants platform builds on a foundation of more than 40 million cooled pluggable optics modules deployed at leading hyperscalers worldwide, as the company extends its precision cooling platform through Thermal Kits for CPO and GPU HBMs.

Phononic’s Thermal Kits for GPU high-bandwidth memory (HBM) and co-packaged optics (CPO) combine high-performance thermoelectric coolers (TECs) with an integrated mechanical and thermal architecture, supported by API-accessible control firmware and software. Embedded firmware and Redfish-compatible interfaces deliver real-time analytics and system visibility, enabling predictive cooling, precise workload placement, and a deterministic thermal envelope. The solution provides targeted, performance-enhancing cooling at the node and chip level, directly addressing thermal hot spots instead of relying solely on bulk liquid cooling methods. Designed to integrate seamlessly with existing liquid or air cooling infrastructure, Phononic’s system responds to workload changes in milliseconds, delivering a 600× control advantage over legacy bulk cooling while maintaining consistent device temperatures and reducing the need for overprovisioned cooling capacity.

“Thermal is becoming the limiting factor in AI infrastructure,” said Ryan Parker, President and COO of Phononic. “We’ve already proven at hyperscale that we can control it in production. As CPO, GPU, and HBM power densities scale, that control becomes required to sustain performance. Extending it into these environments removes that constraint and provides real headroom on sustained performance, energy efficiency at the facility level, and device life.”

When deployed at thermal hot spots throughout the data center, Phononic’s Thermal Kits collectively form the Phononic Thermal Fabric™, enabling true workload orchestration through proactive, pre-emptive cooling and two-way telemetry. With millisecond-level response, the system continuously optimizes thermal performance to improve efficiency, boost compute performance, and reduce time-to-completion for the most demanding AI workloads. This architecture also enables data-center-wide increases in facility water temperatures while maintaining precise chip-level thermal control, allowing operators to improve energy efficiency, unlock additional performance, and extend system lifespan.