anner Electronics announced it will unveil its most powerful Edge AI server to date, the ECA-6050, at Computex 2025. Built on NVIDIA Hopper GPUs, this 2U front-access server supports up to four 600W GPUs with NVLink, enabling high-throughput, low-latency AI inference for applications like LLMs, real-time video analytics, and multi-tenant AI workloads. Designed for enterprise and telecom edge deployments, the ECA-6050 offers 1,024GB DDR5 memory, BlueField-3 DPU support, and NVIDIA AI Enterprise compatibility, delivering compute density, connectivity, and scalability for Edge AI use cases.
NEW TAIPEI CITY, Taiwan, May 20, 2025 /PRNewswire-PRWeb/ -- Lanner Electronics, a global leader in network appliances and edge AI servers, today announced the upcoming launch of its most powerful Edge AI platform to date — the ECA-6050, powered by NVIDIA Hopper GPUs. Purpose-built for enterprise and telecom infrastructure, the ECA-6050 delivers the compute performance, GPU scalability, and low-latency network processing essential for next-generation AI inference at the edge.
The ECA-6050 is a 2U front-access edge AI server purpose-built to support up to four 600W PCIe GPUs, including either four NVIDIA Hopper or four NVIDIA PRO RTX 6000 Blackwell Server Edition GPUs. This exceptional GPU density and memory bandwidth—rare in edge deployments—enables the ECA-6050 to deliver high-throughput inference for large language models (LLMs), real-time video analytics, and multi-tenant AI workloads, making it an ideal platform for private generative AI applications in enterprise and telecom edge networks.
The ECA-6050 supports up to four NVIDIA Hopper GPUs with NVIDIA NVLink, enabling high-speed GPU-to-GPU communication for parallel processing of large AI models with minimal latency. With SR-IOV and Multi-Instance GPU (MIG) support, the platform allows secure, efficient GPU resource partitioning across multiple containers or virtual machines. These capabilities make the ECA-6050 ideal for multi-tenant inference, AI-as-a-Service, delivering scalable performance and workload isolation at the enterprise and telecom edge.
To ensure high-speed communication and efficient data handling, the ECA-6050 is equipped with up to 1,024GB of DDR5 memory and a dedicated PCIe expansion slot for NVIDIA BlueField-3 DPUs or NVIDIA ConnectX-7 network adapters. With BlueField-3, the platform supports hardware-accelerated networking and offloading of data movement, unlocking powerful capabilities for LLM inferencing at the network edge. The systems support the NVIDIA AI Enterprise software platform for enterprise-grade AI deployments.
"The ECA-6050 represents a strategic evolution in Lanner's AI server portfolio, supporting up to four data center-class GPUs for high-throughput, low-latency inference," said Jeans Tseng, CTO of Lanner Electronics. "With the integration of NVIDIA Hopper GPUs and BlueField-3 DPUs, this platform delivers the compute density, connectivity, and scalability required to power real-time, advanced AI inference and training across telecom and enterprise environments."
The ECA-6050 will make its global debut at Computex 2025, held at the Nangang Exhibition Center, Hall 1, Booth #I0201a. Visitors will have the opportunity to experience how Lanner's latest edge AI innovations are driving the future of enterprise AI networking, AI video analytics, private LLM deployments, and intelligent telecom infrastructure.
For more information, visit www.lannerinc.com or contact [email protected] to schedule a meeting at Computex.
About Lanner Electronics
Lanner Electronics is a leading global provider of high-performance network appliances and carrier-grade edge servers. With a focus on delivering cutting-edge solutions for telecom, enterprise, and industrial applications, Lanner is committed to providing innovative and scalable platforms that enable the next generation of networking and AI-driven applications. www.lannerinc.com
Media Contact
Brian Chen, Lanner Electronics, 886 +886-86926060, [email protected], www.lannerinc.com
SOURCE Lanner Electronics
Share this article