Aewin unveils rack-scale AI infrastructure & cooling

AEWIN Technologies unveiled rack-scale AI infrastructure and two-phase direct liquid cooling systems aimed at large-scale AI and data centre workloads.

The line-up includes AI server platforms, dense storage systems and network appliances, alongside a rack-scale cooling design developed with subsidiary Arivor Technologies. The cooling system is designed for AI and high-density computing infrastructure above 100kW per rack.

The server range centres on systems built around next-generation AMD EPYC processors. These include a 1U dual-socket direct liquid cooling server with support for two GPUs, a 2U AI server supporting up to four GPU cards, and a 5U liquid-cooling system supporting up to 16 GPU cards.

Storage is another part of the offering. AEWIN's high-speed storage servers can be configured with 24 U.2 or 32 E3.S SSDs for workloads including generative AI, model training, inference and large-scale data processing.

Cooling focus

AEWIN placed particular emphasis on thermal management as chip power demands rise in AI infrastructure. Its two-phase direct liquid cooling design uses low-boiling-point phase-change cooling technology intended to support next-generation chips above 3000W while improving thermal efficiency and lowering data centre power usage effectiveness.

The company also outlined wider uses for the cooling approach beyond mainstream server deployments. It said it would demonstrate Microchannel Lid applications for chips above 5000W, alongside high-end gaming systems intended to show the scope of the cooling technology.

The broader portfolio also includes platforms for edge AI, storage and cybersecurity workloads. Among them are the BIS-5231, a dual-socket general-purpose server with up to eight PCIe Gen5 slots, the BAS-6101 compact edge AI platform, the MIS-5131 2U2N high-availability storage server, and the SCB-1953 network appliance built on Intel Xeon 6 processors with Intel QAT acceleration.

That mix reflects a market in which infrastructure suppliers are trying to address not only compute demand but also networking, storage and power constraints. As AI deployments grow in scale, equipment makers are increasingly presenting integrated server, storage and cooling designs as data centre operators look for ways to manage rack density and electricity use.

Liquid cooling has become a more prominent part of that shift. Air cooling remains common in many enterprise environments, but higher-density AI systems and more power-hungry processors have pushed vendors to develop alternatives that can remove heat more effectively at rack level.

AEWIN's latest systems fit that trend, linking computing hardware, storage and cooling in a single infrastructure offering. The company also tied cybersecurity into the line-up through network appliances aimed at encrypted traffic and secure data processing environments.

The inclusion of Intel QAT acceleration in the SCB-1953 underlines how specialised hardware is becoming part of security infrastructure design. Vendors are increasingly trying to balance AI processing requirements with encryption and traffic inspection workloads, especially in edge and distributed computing settings.

For storage, the MIS-5131 and the SSD-based server options reflect demand for fast access to large datasets in AI training and inference environments. Dense flash configurations have become more important as model sizes increase and data pipelines require lower latency between storage and compute resources.

AEWIN said the combined architecture is designed to help enterprises build AI-ready environments for real-time analytics and large-scale data processing. Its approach brings together computing, cooling and cybersecurity technologies for next-generation data centre infrastructure.