CHICAGO -- November 13, 2026 -- AEWIN Technologies, a Qisda Group company, will show six server models and a rack-scale 2P DLC cooling system rated for up to 200kW per rack at SC26 in Chicago on November 17-19. Its three GPU servers, built around AMD EPYC 9006 SP7 CPUs, each hold two GPUs per rack unit, by FairsOnline's calculation.
AEWIN's three GPU servers hold two GPUs per rack unit in 1U, 2U and 4U chassis
The BG18-A10710 is a 1U server with an integrated 2P DLC solution. It supports CPUs up to 600W TDP and two single-width 600W GPUs. The 2U BG28-A10710 takes four double-width FHFL GPUs of up to 600W each in PCIe 6.0 x16 slots. The 4U BG48-A20813 takes eight double-width FHFL GPUs.
That works out to 2 GPUs per rack unit in each case (2 in 1U, 4 in 2U, 8 in 4U). Eight GPUs at 600W add up to 4.8kW of GPU power in the BG48 alone, before CPUs. The company describes the BG48 as built for next-generation model training.
Three storage servers cover E3.S, U.2 NVMe and 2.5-inch SATA drives
The BS28-A10710 has 16 E3.S drive bays. The BX28-A10710 supports up to 12 U.2 NVMe SSDs. The BX28-A10711 takes 24 2.5-inch SATA drives, twice the bay count of the U.2 model. AEWIN says the storage line targets HPC and enterprise workloads and AI data pipelines.
Subsidiary Arivor supplies 2P DLC cooling rated for 200kW per rack
AEWIN builds its rack-scale systems on 2P DLC solutions from its subsidiary Arivor. The system uses a phase-change coolant and a rack-scale CDU for dynamic system management. The company says it sustains up to 200kW per rack.
By FairsOnline's calculation, 200kW is about 42 times the 4.8kW maximum GPU load of a single BG48. The company did not disclose pricing, shipment dates or a PUE figure for the rack system.
AEWIN describes AI infrastructure as one of Qisda Group's four key AI business areas. It will demonstrate the servers and rack-scale cooling at SC26 in Chicago from November 17 to 19.