NextFin News - Google is in active negotiations with Shenzhen-based Envicool and other Chinese manufacturers to secure liquid cooling systems for its artificial intelligence data centers, according to sources familiar with the matter. Representatives from Google’s Taiwan-based operations traveled to mainland China this month to meet with Envicool executives, signaling a critical supply chain pivot as the tech giant scrambles to support its next generation of AI infrastructure. The move highlights a growing desperation among U.S. hyperscalers to solve the thermal challenges of high-density computing, even if it means deepening ties with Chinese suppliers during a period of heightened geopolitical friction.
The shift toward liquid cooling is no longer a luxury but a physical necessity. Traditional air-cooling methods are proving insufficient for the heat loads generated by advanced AI processors, such as Nvidia’s latest Blackwell architecture and Google’s own Tensor Processing Units (TPUs). Liquid cooling systems, which circulate specialized fluids directly to the server racks, are significantly more efficient at heat dissipation. JPMorgan analysts estimate the global market for AI server liquid cooling will surge to over $17 billion by the end of 2026, up from $8.9 billion just a year ago. By engaging Envicool, Google is tapping into a supplier that has already scaled production to meet China’s massive domestic data center build-out.
Envicool, currently valued at approximately 98 billion yuan ($14 billion), has positioned itself as a frontrunner in this niche. During a recent industry conference, the company showcased a coolant distribution unit (CDU) reportedly designed to meet Google’s specific technical requirements. Goldman Sachs analysts noted that Envicool is preparing for a significant ramp-up in revenue, with a project pipeline that likely includes Google’s fifth-generation CDU orders. To mitigate potential trade risks, Envicool is expanding its manufacturing footprint beyond China, with new facilities in Guangdong province complemented by operations in Thailand and the United States.
The timing of these negotiations is particularly striking given the political climate. U.S. President Trump has maintained a complex stance on technology trade, alternating between sweeping export restrictions and case-by-case approvals for AI-related hardware. While the U.S. Department of Commerce has focused its "small yard, high fence" strategy on high-end semiconductors, cooling infrastructure remains a gray area. Unlike the chips themselves, cooling pumps and manifolds do not currently face the same level of export scrutiny, allowing Google a narrow window to leverage Chinese industrial capacity to maintain its lead in the AI arms race.
Google’s reliance on Chinese cooling tech also exposes the limitations of the domestic U.S. supply chain. While companies like Vertiv and Schneider Electric are major players, the sheer speed and scale of Chinese manufacturers like Envicool, Lingyi iTech, and Feilong Auto Components offer a competitive edge in a market where lead times are everything. For Google, the risk of a supply bottleneck in cooling equipment is just as existential as a shortage of GPUs. Without the ability to cool its clusters, the most advanced AI models in the world remain effectively paperweights.
The broader ecosystem of Chinese component makers is also benefiting from this infrastructure gold rush. Beyond cooling, firms like Innolight and Eoptolink are dominating the market for optical transceivers, while Victory Giant Technology continues to supply printed circuit boards to both Nvidia and Google. This interconnectedness suggests that despite political rhetoric of "decoupling," the physical reality of AI hardware remains deeply globalized. Google’s foray into Shenzhen is a pragmatic admission that the road to AI supremacy still runs through the heart of China’s manufacturing hubs.
Explore more exclusive insights at nextfin.ai.
