Ningchang launches a 1MW single-phase cold-plate liquid-cooled cabinet achieving 5000W per-chip cooling and 1.2 LPM/kW flow efficiency, analyzing how MW-density infrastructure shifts AI data center economics from token-level optimization to power-thermal co-design.
Baidu Intelligent Cloud has opened its Tianchi supernode reference architecture, a production-validated design for high-density AI compute clusters, to lower barriers for industry-wide adoption by providing a reusable blueprint covering interconnect, power, cooling, and maintenance principles proven across hundreds of cabinets running trillion-parameter models.
This article details the architecture of a 2U four‑node liquid‑cooled server, explaining its component layout, serial flow design, heat‑capture calculations, and the engineering of CPU, memory, SSD, PCIe/OCP, IO, and PSU cold‑plate solutions that together remove about 95% of heat directly via liquid and the remaining 5% through a rear‑mounted air‑liquid heat exchanger.
Amid China's accelerated "new infrastructure" drive, high‑density computing is pushing data centers toward liquid cooling, a technology that not only boosts compute density and cuts floor space but also reshapes server and facility design, lowers failure rates, and promises broader industrial applications.
