Chinese semiconductor startup Yisheng has raised $14.9 million to develop diamond composite materials that cool AI chips, according to Dealroom. The round targets one of the hardest physical constraints in accelerated computing: heat.

Why does chip cooling need new materials?

AI accelerators now push power densities that conventional thermal interface materials struggle to manage. As processors draw more current per square millimetre, the layer that transfers heat from silicon to heatsink becomes a bottleneck. Diamond, with thermal conductivity several times higher than copper, offers a route to move heat out faster.

Yisheng's approach uses diamond composites — engineered combinations of synthetic diamond with other materials — as thermal components for AI chips. The $14.9 million raised will fund development of this technology.

What does the funding signal?

The round points to growing investor interest in thermal management as a distinct category within AI infrastructure. Compute demand has driven record spending on GPUs and accelerators; cooling represents the adjacent problem that determines how much of that compute can run reliably.

• Yisheng secured $14.9 million, Dealroom reports.
• The company's technology applies diamond composites to chip cooling.
• The target market is AI processors, where thermal output limits performance.

China's semiconductor sector has expanded into materials science niches that support AI hardware, and thermal components sit squarely in that push. Diamond-based cooling remains a young field, and funding at this scale typically marks an early-stage company moving from prototype toward qualified production parts.

Who benefits if it works?

Data centre operators face a direct trade-off: higher chip power delivers faster model training and inference, but only if thermal systems keep silicon within operating limits. Materials that improve heat transfer can raise effective compute density without redesigning facilities.

For chipmakers, better thermal interfaces can unlock higher clock speeds or denser packaging. That puts suppliers of advanced cooling materials in the supply chain conversation alongside substrate and packaging vendors.

No details on investors, valuation or timeline were disclosed in the report.

$14.9 million now backs the claim that diamond, not just silicon, decides how fast AI chips can run.