Lenovo is introducing new ThinkSystem servers featuring sixth-generation AMD EPYC processors and new storage arrays. The company is also expanding its AI factory services, offering colocation through Digital Realty. This is intended to help organizations get AI and HPC workloads up and running faster.
Lenovo is building its updated portfolio around the convergence of AI and high-performance computing. According to the company, demand for underlying infrastructure is growing rapidly as continuous inferencing, agentic AI, and more complex scientific workloads become more common. The new servers run on AMD EPYC 9006 processors with an SP7 socket.
These chips, codenamed Venice, officially became available this month. They feature up to 256 Zen 6c cores, 16 memory channels, and 128 PCIe Gen 6 lanes per socket, with CXL 3.1 support. The top-of-the-line model, however, has a TDP of up to 600 watts. This means cooling is no longer an afterthought.
“That means getting more useful performance from every watt, overcoming growing power and cooling constraints, and moving faster from infrastructure deployment to operational AI,” says Scott Patti, Vice President of the Infrastructure Solutions Group at Lenovo.
Liquid cooling and flexible partitioning
The ThinkSystem SC755 is a dual-socket server that uses Neptune liquid cooling. The system dissipates up to 100 percent of its heat via water. Lenovo claims up to 40 percent higher sustained performance than comparable air-cooled systems. With Dynamic Partitioning, companies can convert two dual-socket systems into single-socket configurations, and back again.
The ThinkSystem SR755 is designed for large-scale virtualization, engineering, financial modeling, electronic design automation, and agentic AI. Lenovo promises high compute density and ample memory bandwidth, with lower power consumption than dual-socket configurations.
On the storage side, Lenovo is introducing the ThinkSystem DE6800 and DE6500, optimized for latency-sensitive block workloads with high throughput. The DE6800 is designed for applications that require reliable, high-throughput storage, while the DE6500 is the more cost-effective option with simpler management.
From hardware to running AI
But deploying hardware is only part of the challenge, according to Lenovo. That’s why it’s expanding its hybrid AI factory services to include GPU deployment, configuration, and optimization. New features include multi-node support for Red Hat AI Enterprise and NVIDIA AI Enterprise with NVIDIA Run:ai.
For large-scale rollouts, gigafactory services will be available, including expertise from Nvidia Infrastructure Services (NVIS) for deployment and cluster validation. And for those who need additional data center capacity, Lenovo offers access to Digital Realty’s colocation facilities. In the Digital Realty Innovation Lab (DRIL), organizations can validate their AI infrastructure and workloads. Financing is provided through Lenovo TruScale Infrastructure as a Service.
Tip: Lenovo AI Express delivers AI infrastructure within 15 business days