News Categories

At the OCP Taiwan Summit, AVC Thermal R&D Vice President Fang-chou Lin shared the latest technical standards for high-density AI server thermal management. The speech focused on cold plate optimization and CDU upgrade directions, summarizing five core challenges for future liquid cooling systems. These include continuously rising chip TDP, 45℃ high-temperature water cooling trends, stricter coolant cleanliness requirements, lower CDU heat exchange temperature difference and the shift from component competition to full industrial chain integration. As a professional data center cooling CDU supplier, AVC delivers systematic solutions targeting high-heat-flux AI computing scenarios.

As AI accelerator rack power rises from today's roughly 100 kW per rack toward 600 kW and beyond, cooling architecture is fast becoming the defining constraint in data center design. Air cooling remains effective for high-density AI data centers in the low-to-mid density range, but as rack power continues to climb, direct-to-chip liquid cooling is emerging as the critical pathway for absorbing significantly higher power densities. Two-phase direct-to-chip liquid cooling, which exploits phase change within the cold plate, extends the density curve even further, delivering the scalability needed for rack power levels between 200 kW and 1 MW+. In this landscape, a forward-looking data centre liquid cooling system is no longer a future upgrade, but a strategic prerequisite for deploying next-generation AI infrastructure.

As AI chips push past 2,000W in 2026 and single-rack power approaches 500kW, air cooling has run out of breath. TrendForce data shows liquid cooling penetration in global AI data centers climbing from 14% in 2024 to 33% in 2025, with 2026 forecasts raised from 40% to as high as 53%; the global liquid cooling market is projected at US$12.7 billion in 2026 and US$42.2 billion by 2030. Yet few liquid-cooling concept stocks have real competitiveness. This article zooms in on the system's most valuable device, the AI GPU liquid cooling CDU, and uses it as a yardstick to measure Envicool, Shenling Environment and Goaland.

NVIDIA achieved USD 89 billion in 2026 Q3 data center revenue with a 117% year-on-year increase, reflecting robust AI computing growth. GB300 NVL72 racks support up to 142kW power load, making high-density stable cooling essential for full-load GPU operation. TrendForce upgrades the 2026 AI chip liquid cooling penetration rate to 53%. The rate stood at 33% in 2025 and is expected to near 60% in 2027. Liquid cooling has evolved from exclusive supercomputer customization into standard infrastructure for high-end AI systems. Industry competition focuses on full-chain capability rather than single liquid cooling distribution unit hardware. Leading vendors deliver low thermal resistance, zero leakage and stable mass delivery to realize sustainable revenue. The industrial chain covers chip heat exchange, fluid connection, cooling distribution, facility heat rejection and AI rack deployment. The global liquid cooling market will grow from USD 8.2 billion in 2026 to USD 29.5 billion in 2033. China’s market will expand from RMB 23.25 billion in 2026 to RMB 47.04 billion in 2028 with steady growth momentum.

AI server computing competition has shifted to high-precision terminal and connector manufacturing. Veteran industry practitioners reveal core supply chain features. Leading domestic connector manufacturers earn over half their revenue from one international giant, with end products supporting global AI computing facilities. Liquid cooling connectors achieve a 40% gross margin, outperforming 30% for air cooling connectors and 26% for high-speed backplane connectors. Only two mature mass-production liquid cooling lines operate worldwide, constrained by technical bottlenecks in liquid cooling plate microchannel processing and brazing. Reliable AI GPU liquid cooling CDU supplier delivers matched thermal control solutions to support full-load operation of high-power AI GPU racks.

MarketsandMarkets releases the latest industry forecast for global liquid cooling components. The global data center liquid cooling manifold market size will grow from 940 million USD in 2026 to 6.33 billion USD in 2033, representing a compound annual growth rate of 31.2 percent. The liquid cooling track has remained highly popular in recent years. Market attention is mostly focused on cold plates and cooling distribution units while manifolds are long underrated as a critical subsystem. Manifolds act as core vascular structures in liquid cooling systems. They precisely distribute and regulate coolant among servers, cold plates, racks and overall cooling loops. Stable flow and pressure control serve as fundamental guarantees for reliable heat dissipation in high density computing environments. The booming AI computing infrastructure strongly drives demand for professional AI data center liquid cooling manifold solutions across modern data center deployments.

The industry has defined three mainstream technical architectures for row-based rack-level liquid cooling manifolds, covering tube-drilled type, plate-brazed flow channel type and segmented modular splicing type. Each solution serves distinct data center power density and deployment scenarios with unique structural features, cost models and operational reliability. As a professional GPU cooling manifold manufacturer, we optimize manifold design targeting high-load GPU server heat dissipation, matching different AI cluster construction standards and long-term operational requirements. The tube-drilled type acts as the traditional and most widely adopted solution, while brazed plate and modular structures represent high-end and next-generation evolutionary routes respectively.

The Vera Rubin architecture adopted by NVIDIA has pushed single-chip TDP up to 1800W, lifting single rack power density from the traditional 20kW to 70–132kW with a future target of 200kW. Traditional air cooling has a maximum heat dissipation ceiling of only 40kW, failing to support high-power AI computing iteration. Stricter energy policies further accelerate industrial upgrades. National Development and Reform Commission mandates a maximum PUE of 1.3 for newly built large-scale data centers and 1.25 for eastern regional facilities. Beijing will implement tiered electricity pricing from 2026 for data centers with PUE exceeding 1.35, imposing additional charges for excessive energy consumption. Legacy data centers generally record a PUE above 1.8 and even over 2.5 for outdated facilities, with power costs accounting for more than 60% of total OPEX. Modern HVAC systems have become core competitiveness rather than auxiliary equipment for data center operation. Advanced energy saving liquid cooling CDU products effectively reduce overall power consumption, help facilities meet PUE compliance requirements, and deliver long-term operational cost reduction for high-density AI server clusters.

Power consumption of NVIDIA H200 and B200 single GPUs exceeds 700W, pushing traditional air cooling to its operational limit. Maximized fan speeds trigger severe bottlenecks in noise, energy consumption and temperature control, firmly capping AI computing performance. Liquid cooling has evolved from an optional upgrade to a mandatory infrastructure solution for mainstream NVIDIA AI servers, unlocking new value distribution in high-density AI computing. Advanced energy saving air liquid cooling CDU solutions effectively resolve the dual pain points of high power consumption and heat dissipation failure, delivering efficient energy optimization for mass-deployed AI server clusters.

Global AI infrastructure construction is currently dominated by hyperscale data center projects, with 100MW, 300MW and even GW-level campuses constantly setting new industry records. However, Kirk Killian, President of Partners National Mission Critical Facilities, points out that hyperscale deployment only represents part of AI infrastructure. Unlike hyperscale facilities with single-unit capacity ranging from 40MW to 300MW, most enterprise data centers operate at a scale of 0.25MW to 10MW. Although small in standalone capacity, enterprise users are 50 to 100 times more numerous than global hyperscale operators, widely covering finance, manufacturing, healthcare, scientific research, energy and government sectors. These enterprises continuously invest in core business IT infrastructure, making enterprise data centers an indispensable pillar of AI industrial deployment. The booming upgrade demand drives the rapid iteration of AI data center liquid cooling technologies tailored for medium and small-scale computing scenarios.