appConfigEntityId

bodyType

customContents

entityId

entityUuid

isVideo

articlePageQuery

article

articleSchemaContent

content

hooksArticlePageAdvertisementConfigArticle

hooksArticlePageFloatingButtonConfigArticle

appConfig

filter

json

liveArticle

Wency Chen

China’s supply chain is stepping up investments in liquid cooling technology as the global AI build-out pushes data centre power densities to levels that air cooling can no longer efficiently handle.
Dozens of Chinese listed companies have in recent months announced plans to expand into, or ramp up, liquid-cooling systems as demand surges for technology that uses circulating fluids to keep racks of AI chips from overheating.
Among them are HVAC specialist Sanhua Intelligent Controls, electronics manufacturer Lansi Technology, power-systems maker Kehua Data, as well as thermal-solutions providers Shenzhen FRD and Goaland Energy Conservation.
“We’re actively expanding into the liquid-cooled server business and continuing to advance related partnerships,” Sanhua, a key Tesla supplier, said on Thursday on an investor platform, adding that its thermal management components can be deployed in liquid-cooling applications.
One of the most closely watched players in the sector is Shenzhen-based Envicool, a precision cooling specialist whose customers include Nvidia, Intel and domestic hyperscale cloud operators such as Alibaba and Tencent.
Its Shenzhen-listed shares have more than tripled over the past year, closing at 113.14 yuan on Friday.
Goldman Sachs and UBS have raised Envicool’s 12-month price targets in recent research notes to 121.10 yuan and 160 yuan, respectively.
“The digital economy and carbon reduction are long-term areas of focus globally, and technologies – such as precision cooling – that enable and improve energy utilisation efficiency will be increasingly adopted to achieve these development goals,” said Goldman Sachs analyst Jacqueline Du.
The momentum is being fuelled by a global wave of AI data centre construction aimed at meeting exploding demand for computing power.
As model sizes grow and workloads intensify, attention is shifting from semiconductors themselves to the infrastructure required to power and cool them, straining traditional supply chains and accelerating innovation across the hardware ecosystem.
Liquid cooling is emerging as a critical component as chip power consumption climbs.
Nvidia’s Blackwell B200, for example, operates at more than 1,000 watts per processor. The rack-scale supercomputing platform GB200 NVL72, with 72 Blackwell GPUs and 36 Grace CPUs, features a liquid-cooled design.

Data-centre liquid cooling generally takes two forms: direct-to-chip cold plates that circulate coolant through sealed loops attached to processors, and more advanced direct-contact systems, including immersion and spray cooling.
Upstream, the supply chain comprises components such as cold plates, cooling distribution units (CDUs), manifolds and quick connectors.
Key suppliers include US-based Vertiv and Boyd and Taiwan’s Cooler Master. Their customers range from liquid-cooling solution integrators to server manufacturers, with cloud service providers and telecoms operators forming the downstream end market.
UBS estimates the global liquid-cooling market could reach US$31 billion by 2030, at a compound annual growth rate of about 51 per cent.
Goldman Sachs forecasts liquid cooling will account for 74 per cent of AI training servers by 2026, while adoption in AI inference servers will remain comparatively lower. The bank expects the global server-cooling market to reach about US$14 billion this year.
US liquid-cooling supplier Vertiv saw its New York-listed shares jump to an all-time high this week after reporting strong full-year results, with orders, revenue and profit all rising and beating guidance for 2026. The company, which partners with Nvidia, forecast net sales of about US$13.5 billion.
The traction is also gathering pace in China, where authorities and tech companies have sped up the construction of large-scale computing clusters to underpin the country’s AI ambitions.
In a rapid expansion drive, Beijing has built 42 “10,000-card” intelligent computing clusters, with total intelligent computing capacity exceeding 1,590 exaflops – one quintillion calculations per second – according to the latest data from the Ministry of Industry and Information Technology.
Since 2022, the government has rolled out the “Eastern Data, Western Computing” initiative, aimed at shifting data-processing workloads from energy-intensive eastern regions to western provinces with abundant renewable resources.
Alongside broader green and low-carbon development targets, national policies impose explicit power usage effectiveness requirements on data centres, further encouraging the adoption of higher-efficiency liquid-cooling solutions.


Trade truce extension aside, was AI the real Xi-Trump summit breakthrough?

How Chinese firms can avoid a Europe-style backlash in Southeast Asia

How Alibaba is using AI to revamp its mapping app – and take on Meituan

China’s AI-led transformation is taking shape. Why are some economists urging caution?

Artificial intelligence (AI)

China moves deeper into liquid cooling as AI data centres push their thermal limits

An employee conducts maintenance in a server room of the China Mobile Hohhot Data Centre in Hohhot in Inner Mongolia. Photo: Xinhua

Goldman Sachs forecasts liquid cooling will account for 74 per cent of AI training servers by 2026. Photo: Shutterstock

As Chinese AI and GPU demand heats up, a Hong Kong data centre turns to liquid cooling

Server racks featuring the new direct-to-chip liquid cooling solution at Global Switch’s Hong Kong facility in Tseung Kwan O on September 12, 2024. Photo: Global Switch

Global Switch’s Hong Kong data centre in Tseung Kwan O on November 15, 2017. Photo: Global Switch

The technical area of Global Switch’s Hong Kong data centre, with power distribution cabinets along the walls. Photo: Global Switch

Global data centres to drive demand for liquid cooling systems as AI boom heats up servers, energy consumption

Server rooms in a big data storage and cloud computing centre. Photo: Shutterstock

OpenAI CEO Sam Altman gestures during a session of the World Economic Forum in Davos in January 2024. Photo: AFP

An AirTrunk hyperscale data centre outside Sydney. Photo: Handout

China’s thirsty data centres, AI industy could use more water than size of South Korea’s population by 2030: report warns

Data centres consume water directly through cooling processes and indirectly from electricity generation. Photo: Xinhua

Tech

Chinese manufacturers are moving aggressively into liquid-cooling technologies as hyperscalers expand data-centre infrastructure


Chinese manufacturers are moving aggressively into liquid-cooling technologies as hyperscalers expand data-centre infrastructure.


China moves deeper into liquid cooling as AI data centres push their thermal limits

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}Chinese manufacturers are moving aggressively into liquid-cooling technologies as hyperscalers expand data-centre infrastructure

Chinese manufacturers are moving aggressively into liquid-cooling technologies as hyperscalers expand data-centre infrastructure