Jiangsu Haifa Machinery Manufacturing Co., Ltd.

As GPU single-rack power evolves toward hundreds of kilowatts or even megawatt levels, AI data center cooling is undergoing a comprehensive shift from air cooling to Direct-to-Chip Liquid Cooling, Water Cooling, and Megawatt CDUs. NVIDIA latest AI infrastructure has advanced an 800 VDC high-voltage direct current architecture, while Google continues to upgrade its Project Deschutes CDU. The reliability of pumps, heat exchangers, and control systems is becoming a critical metric for AI Factory. Under this transformation, High-Voltage DC Electronic Canned Motor Pumps, high-power electronic canned motor pumps, CDU Circulation Pumps, Sealless Liquid Cooling Pumps are entering the industry spotlight. Compared with traditional "AC motor + coupling + mechanical seal" configurations, electronic motor pumps highly integrate the motor with the pump hydraulic components, eliminate dynamic shaft seals, further reduce leakage points and equipment footprint, and adapt to variable flow regulation, N+1 Pump Redundancy, low noise and round-the-clock continuous operation requirements. For cooling such as deionized water and glycol aqueous solutions, Low Leakage, Vibration, Wide-Range Speed Control, andant Compatibility have become new selection keywords. The megawatt-class CDU trend is already very clear. In 2026, Vert's 2.3 MW and LG's 2.6 CDUs successively obtained NVIDIA DSX Ready qualification, Nc launched a  kW In-R CDU, and Google Deschutes is oriented toward 2 MW-class cooling and high-availability redundant architecture. Jiangsu Haifa Machinery Manufacturing Co., Ltd..jslgpump.com can leverage its technical accumulation in API 610 process pumps, API  magnetic drive pumps, canned motor pumps, and high-efficiency circulation pumps to extend its expertise in hydraulic modeling, sealless structures, rotor dynamics, and material compatibility into the AI liquid cooling field, with a focus on 800VDCU Pumps high-power canned motor circulation pumps, N+1 redundant liquid cooling pumps, and megawatt-class CDU secondary-side circulation solutions. The AI computing supply chain is expanding from "chip competition" to coordinated competition in power and thermal management. In the future, the core of liquid cooling pumps will upgrade from mere flow rate and head to efficiency, redundancy, zero leakage, low harmonics, online monitoring, full lifecycle reliability. Chinese References: 1. NVIDIA: "800 VDC Architecture for AI Data Centers," 2026 — 800DC architecture, AI Factory high-power power distribution, and ecosystem development. 2. Google Cloud: "Enabling 1 MW IT racks and liquid cooling" — Project Deschutes CDU, redundant pumps/heat exchangers, and approximately 99.999% CDU availability experience. 3.su Haifa: "NVIDIA 800VDC Empowers Liquid Cooling CDU: High-Voltage DC Electronicanned Motor Pump Technology Upgrade and Deployment," 2026. Jiangsu Ligong Group. 4 Panasonic: "Liquid Cooling System Business for Generative AI Data Centers," 6 — 400 kW 800 kW, and capacity CDU development. English References: . NVIDIA, Why Scaling AI Compute Performance Requires a New Power Architecture, 6. 2., Liquid Cooling AI Factories, 202 — 45°C-water liquid cooling and fully liquid-cooled next-generation infrastructure. 3 Google Cloud, Enabling 1 MW IT racks and liquid cooling at OCP EMEA Summit. . Nidec, Project Deschutes CDU Based Google OCP Specification — 2 MW class / 80I. 5.iv, 2.3 MW CoolChip CDU Qualified as NVIDIA DSX Ready, 2026.6. LG Electronics, 2.6 MW CDU Qualifies as NVIDIA DSX, 2026.

www.jslgpump.com