AI and Semiconductor Computing Scenario

Core Scenario Requirements

It needs to accommodate ultra-high heat flux density up to 1000 W/cm² and TB-level high-speed signal interconnection transmission demands for large AI model training and supercomputing center high-performance chips. It solves key industry pain points such as AI chip overheating and frequency throttling, high-speed signal transmission latency, and excessive data center energy consumption, serving as a core material supporting the upgrading of next-generation AI computing power.

Specific Applications of Six Major Products

1. Single/Polycrystalline Diamond Substrates & Heat SinksUsed as substrates and heat sinks for AI chips, GPU/CPU chips, as well as thermal structures for HBM stacked memory.With a thermal conductivity of 2300 W/m·K, single-crystal diamond heat sinks can be directly bonded to the backside of AI chips, reducing junction temperature by over 40 °C. They fundamentally suppress the self-heating effect of high-performance chips, avoid thermal throttling, improve chip computing output by more than 30%, and double service life. They are the core thermal management material for next-generation AI computing chips.
2. Diamond Composite Heat Sinks & EnclosuresApplied to heat sinks of AI server power modules, enclosures of 800G/1.6T high-speed optical modules, and heat dissipation housings for data center switches.Diamond-copper composite heat sinks reach a thermal conductivity of 1000 W/m·K, resolving heat dissipation challenges of high-power power modules and high-speed optical modules and improving long-term operational stability. Meanwhile, the electromagnetic shielding performance of the enclosures suppresses electromagnetic interference of high-speed signals and ensures stable high-speed communication in data centers.
3. M9/M10 Copper-Clad Laminates & TVG Glass SubstratesAs foundational core materials for high-speed AI interconnection, they are widely used in AI server computing mainboards, high-speed switch backplanes, GPU interconnection high-speed boards, and HBM memory substrates.M9/M10 laminates feature Df ＜ 0.002 and stable Dk 3.0~3.8, perfectly supporting 224Gbps/400Gbps PAM4 high-speed signal transmission. They greatly reduce signal loss and latency in high-speed interconnection and improve interconnection bandwidth and computing efficiency of AI computing clusters. Their ultra-high thermal conductivity up to 700 W/m·K also solves heat accumulation caused by densely arranged PCBs.TVG glass substrates are adopted for 2.5D/3D advanced packaging carrier boards of high-end AI chips. With excellent dimensional stability and low dielectric properties, they ensure high-speed interconnection between chips and HBM memory, increase the computing density of AI chips, and act as a key packaging material for next-generation AI chips.
4. Customized Diamond Heat Dissipation ModulesAs the core thermal solution for AI computing power upgrading, they are customized liquid-cooling heat dissipation modules for AI server GPU/CPU chips, supercomputing center processors, and large model training clusters.Optimized via dedicated AI thermal simulation, they form an integrated thermal solution of diamond heat sink — microchannel liquid cold plate — vapor chamber. Capable of handling ultra-high heat flux up to 1000 W/cm², they deliver over 5 times higher heat dissipation efficiency than traditional copper water-cooled plates, completely eliminating overheating and frequency reduction of large AI model training chips and stabilizing full-load operation. They also lower the PUE of data centers below 1.1, greatly cutting cooling energy consumption, and providing essential support for ten-thousand-card-level AI computing clusters.
5. Aluminum Silicon Carbide (AlSiC)Applied to structural frames of AI server chassis, thermal bases of liquid cooling systems, and packaging substrates of high-speed optical modules.Its lightweight and high-rigidity characteristics meet the high-density rack deployment requirements of AI servers; its excellent thermal conductivity improves overall heat dissipation efficiency and further reduces data center cooling power consumption.
6. Quantum DiamondUsed for quantum-AI hybrid computing chips, quantum neural network processors, and high-precision time synchronization systems in supercomputing centers.NV-color-center-based quantum diamond qubits can operate stably at room temperature without complex cryogenic refrigeration, making them a core material for constructing room-temperature quantum computing chips. Combined with traditional AI chips, they enable quantum-AI hybrid computing, delivering exponential computing acceleration in complex scenarios such as large model training, drug development, and cryptography.They also support picosecond-level time synchronization for ten-thousand-card clusters in supercomputing centers, significantly boosting parallel computing efficiency, and serving as a fundamental pillar of the next-generation AI computing revolution.