Lenovo and Cisco Unveil Pre-Validated AI Data Center Infrastructure to Accelerate Deployment
December 15, 2025
As enterprises race to deploy artificial intelligence, data center teams are grappling with the complexity of building high-performance, scalable AI infrastructure from disparate components. The challenge lies not only in achieving the necessary compute density but also in ensuring seamless, low-latency networking to prevent AI model training from becoming bottlenecked by data movement.
To address this, Lenovo and Cisco have jointly launched a technically pre-validated infrastructure solution designed to simplify and accelerate AI deployments. The solution combines Lenovo's ThinkSystem SR675 V3 servers, configured in a "2-8-5" design (2 AMD EPYC CPUs, 8 NVIDIA GPUs, 5 PCIe Gen5 network adapters), with Cisco's Nexus 9364D-GX2A switches, which provide a 400G high-bandwidth, lossless fabric. This architecture, validated for use cases ranging from new AI clusters to augmenting existing data centers, is engineered to scale linearly up to 96 GPUs per deployment unit without major re-engineering.
The rigorous joint validation process tuned key performance parameters, including Non-Uniform Memory Access (NUMA) domains and BIOS/UEFI settings, while automating the configuration of Quality of Service (QoS) and Priority Flow Control on the Cisco switches. Performance benchmarks demonstrated an effective throughput of 391 Gbps on a 400 Gbps link, showcasing near line-rate capability critical for distributed AI workloads. Meghan Kachhi, Technical Marketing Engineering Technical Leader who led the validation, stated the partnership delivers "a technical foundation trusted by IT and business stakeholders."
For the industry, this pre-validated approach signifies a shift towards turnkey solutions that reduce the time, risk, and engineering overhead associated with building AI "factories." By providing a blueprint that ensures predictable performance from day one, the partnership aims to help organizations move faster from AI proof-of-concept to production. The integrated stack promises operational simplicity, lower total cost of ownership through space and power efficiency, and centralized management via Cisco's Nexus Dashboard, allowing IT teams to focus on strategic initiatives rather than infrastructure troubleshooting.
Source: cisco
