Asia Vital Components Ships Double-Sided Cold Plates
The Taiwanese cooling supplier is expanding liquid cooling production for AI servers as demand for higher-power chips grows.
Updated on Nov. 11, 2026 in Data Centers
Asia Vital Components has started shipping double-sided cold plates designed for vertical power delivery in AI servers. The company says customers increasingly prefer the design as chips push past 2,300 watts of thermal design power.
Why it matters
Cold plates carry heat away from high-power AI chips using liquid flowing through internal channels, making them central to data center cooling. The shift toward liquid cooling reflects the rising heat output of accelerator chips.
The cold plates are designed for chips with a thermal design power above 2,300 watts, and the company estimates a 40-50% share of the cold plate market for Nvidia's GB200 and GB300 platforms. Whether capacity expansion will keep pace with the forecast demand has not been confirmed.
The players
Asia Vital Components
A Taiwan-based maker of thermal management products whose cold plates are used in AI servers, and which has worked with AWS, Google, Meta, and OpenAI.
Nvidia
The U.S. chip designer whose GB200, GB300, and Vera Rubin platforms drive demand for liquid-cooled AI server components.
The details
The company reported record third-quarter 2026 revenue and attributed margin gains to demand for higher-value liquid cooling products and a better sales mix. Most of its planned growth is expected from facilities in Vietnam, supported by planned capital expenditure of NT$15-17 billion.
Timeline
Third quarter 2026: record revenue and margin gains reported.
October 2026: record monthly revenue of NT$22.1 billion.
November 10, 2026: earnings call where the company laid out its outlook.
2027: planned capacity expansion and expected ASIC revenue surpassing GPU revenue.
End of 2027: expected full adoption of liquid cooling in GPU and ASIC AI servers.
The Tech Race
Asia Vital Components holds an estimated 40-50% share of the cold plate market for Nvidia's GB200 and GB300 platforms and is a preferred supplier for the Vera Rubin architecture. The company's outlook that ASIC revenue will surpass GPU revenue in 2027 signals a broader shift in the AI server market.
For data center operators and cloud customers, the shift to liquid cooling changes how AI compute capacity is built and maintained. Users of AI services may see the effects indirectly through the pace at which new high-power server capacity comes online.
The takeaway
Liquid cooling is moving from specialty equipment to a baseline requirement for high-power AI servers, so cooling capacity is becoming a key constraint alongside chip supply. Tracking supplier capacity plans offers an early read on how quickly AI infrastructure can scale.
Further reading
For more on the infrastructure behind this shift, see Data Centers.
Source note: This article includes information reported by Crypto Briefing.
