With the rapid development of generative AI and large models, the cooling demands for AI servers and smart terminals are growing. Liquid cooling technology is becoming the standard choice for data centers and high-performance computing equipment.
The main reasons include:
-
The power consumption of high-performance chips continues to increase, and traditional air cooling can no longer meet cooling needs. NVIDIA B200 GPU power consumption reaches 1000W, breaking the limits of air cooling.
-
Liquid cooling is highly efficient and can significantly reduce energy consumption. It can save over 90% energy compared to air cooling, helping to lower data center PUE values.
-
Policies require data centers to improve energy efficiency. New large data centers need to reduce PUE to below 1.3, and liquid cooling is a key technology to achieve this goal.
-
Liquid cooling can optimize the lifecycle cost of data centers. Although initial investment is higher, long-term operating costs are lower.
-
Downstream operators are accelerating the application of liquid cooling. The three major operators plan to apply liquid cooling to over 50% of data center projects after 2025.
Liquid cooling market participants mainly fall into two categories:
-
Server manufacturers and IDC construction companies, such as Inspur Information and Sugon Data Creation.
-
Professional temperature control system providers, including component suppliers and full-chain solution providers.
It is estimated that when AI server shipments exceed 1.4 million units in 2025, the liquid-cooled server market size will exceed 160 billion yuan, growing more than 8 times.
In the mobile terminal field, cooling technology is also constantly iterating:
-
Graphite cooling film: First widely applied in iPhone 4 in 2010, becoming mainstream.
-
Heat pipe cooling: Widely adopted since 2016, such as in Samsung Galaxy series.
-
VC cooling: Currently the most advanced mobile phone cooling technology, with better cooling effects.
With the popularization of new features such as 5G and wireless charging, the cooling demands of mobile devices are further increasing. Future edge AI devices will require more precise cooling, needing to iterate synchronously with chips.