The race to cool increasingly power-hungry AI accelerators just gained a new materials contender. Coherent Corp. has introduced Thermaditeโข 800 Liquid Cold Plates (LCPs), a thermal management platform designed specifically for next-generation high-power compute systems.
The photonics and advanced materials company says the new cold plates deliver thermal conductivity of 800W/(mโ K), roughly twice that of copper while maintaining a low coefficient of thermal expansion and high dimensional stability. As AI processors push power densities higher, those characteristics allow operators to reduce chip temperatures and maintain stable thermal interfaces across increasingly complex accelerator packages.
Thermal constraints have emerged as one of the biggest infrastructure challenges in AI infrastructure. GPU clusters now generate heat flux levels that conventional copper cold plates often struggle to dissipate efficiently. As a result, infrastructure architects increasingly look toward new materials and advanced liquid cooling architectures.
Diamond-Integrated Material Targets AI Thermal Bottlenecks
At the core of the product lies Thermadite, a composite material formed by integrating diamond into a silicon carbide (SiC) matrix. The approach creates a rigid structure with high thermal conductivity while maintaining mechanical stability under high thermal loads.
According to Coherent, Thermadite 800 liquid cold plates can reduce chip temperatures by more than 15ยฐC compared with traditional copper cold plates in high heat-flux AI accelerator environments.
Equally important for large-scale deployments, the material weighs significantly less. Thermadite delivers about 60% lower density than copper, which helps reduce system mass in densely packed compute platforms and rack-level liquid-cooling assemblies. Together, these attributes aim to support the next generation of accelerator platforms where rack densities continue climbing and thermal margins shrink.
The Thermadite 800 platform also enables complex internal cooling geometries. Coherent engineers design the liquid cold plates with microchannel architectures tailored to specific chip heat maps, directing coolant precisely toward localized hotspots on AI processors.
This targeted cooling strategy offers multiple benefits. It lowers thermal resistance across the chip-to-coolant path while minimizing pressure drops across the cooling loop. As a result, operators can reduce coolant flow requirements and improve overall system efficiency.
Lower pumping requirements also translate into lower operating costs for large AI clusters, where thousands of accelerators often share centralized liquid cooling infrastructure.
Engineering the Entire Chip-to-Coolant Thermal Path
Coherent executives frame the product as part of a broader shift in thermal design philosophy for high-performance computing.
“Effective cooling in high-performance computing depends on the entire thermal route from chip to coolant,” said Steve Rummel, Senior VP of the Engineered Materials Group. “By merging decades of material science with precision manufacturing, Thermadite liquid cold plates lower chip temperatures, optimize pressure flow, and minimize interface resistance – all specifically designed for the extreme requirements of modern AI chips.”
The statement reflects an industry-wide recognition that thermal performance no longer hinges on a single component. Instead, engineers increasingly optimize the full stack of materials, interfaces, and cooling channels connecting the processor package to facility-level cooling systems.
Thermadite 800 cold plates draw on Coherentโs vertically integrated manufacturing capabilities. The company controls multiple stages of the process from advanced material development to precision fabrication and high-volume production.
That level of integration may prove strategically important as AI infrastructure expands globally. Data center developers and accelerator vendors increasingly seek supply partners capable of delivering both specialized materials and scalable manufacturing.
Cooling solutions now sit at the center of AI infrastructure planning. High-density GPU clusters often exceed 80โ120 kW per rack, a level that pushes traditional air cooling beyond practical limits. Liquid cooling systems, including direct-to-chip cold plates, have therefore become a foundational technology for modern AI facilities. By introducing diamond-integrated thermal materials, Coherent aims to position itself deeper inside that emerging supply chain.
Industry Showcase at Major Thermal and Optical Conferences
The company plans to demonstrate examples of Thermadite 800 liquid cold plates at two upcoming industry events: the SEMI-THERM Symposium & Exposition and the Optical Fiber Communication Conference and Exhibition.
Both conferences attract engineers and infrastructure planners focused on high-performance computing, photonics, and advanced semiconductor packaging fields increasingly converging around AI infrastructure design.
As AI accelerator performance continues to climb, thermal engineering will likely become one of the defining battlegrounds for next-generation compute hardware. With Thermadite 800, Coherent signals that materials innovation, not just system design will play a decisive role in keeping AI processors within safe operating limits.
