San Jose, Calif. – – October 06, 2026 -- AEWIN Technologies, a unit of Qisda Group, will unveil a modular DC-MHS server lineup capable of supporting rack densities above 200kW using two-phase direct liquid cooling (2P DLC) technology from subsidiary Arivor at the OCP Global Summit 2026, running October 12–15 at Booth #D9.
New servers run AMD EPYC 9006 CPUs with PCIe Gen6 connectivity
The portfolio is built on AMD EPYC 9006 series server processors paired with PCIe Gen6 connectivity to address rising compute-density requirements in AI data centers. AEWIN positions the lineup to reduce power usage effectiveness (PUE) while extending hardware lifespan under sustained high-load conditions.
Flagship BG48-A20813 packs eight GPUs into a 4U chassis for agentic AI training
The newly added flagship, BG48-A20813, is a 4U AI server supporting up to eight dual-width full-height, full-length GPUs, engineered for large-scale agentic AI workloads and next-generation model training. The 2U BG28-A10710 model carries four dual-width FHFL PCIe 6.0 x16 slots rated for 600W GPUs, while the compact 1U BG18-A10710 integrates 2P DLC to cool a 600W TDP CPU alongside two single-width 600W GPUs.
Storage-optimized nodes target AI data pipeline bottlenecks
For data-intensive deployments, the BS28-A10710 offers 16 E3.S drive bays, and the BX28-A10710 scales to 12 U.2 NVMe SSDs or 24 SATA drives, covering a range of storage configurations for AI pipeline and data management use cases.
System management relies on OCP DC-SCM2 standard and Titanium-rated power
All platforms use OCP DC-SCM2-based system management powered by the ASPEED AST2700 BMC, combined with redundant 80 PLUS Titanium power supplies, targeting AI inference, HPC, cloud-native, and neocloud deployment environments.
AEWIN frames the launch as a demonstration of rack-scale system integration spanning compute, networking, storage, and thermal management under the Qisda Group umbrella, combining its server design expertise with Arivor's 2P DLC cooling technology.