Lenovo Bets on Liquid Cooling and AMD EPYC for AI Factories

Lenovo's new Neptune liquid-cooled ThinkSystem SC755 server removes up to 100% of system heat through water and delivers up to 40% higher sustained performance than comparable air-cooled systems, according to Lenovo's internal testing. That headline number anchors a broad portfolio refresh announced November 10, 2026: ThinkSystem servers running 6th Generation AMD EPYC SP7 processors, new storage, validated interconnects, and expanded AI factory services spanning customer data centers, hybrid environments and colocation facilities.

The message behind the launch is systemic. Lenovo treats AI and HPC as an infrastructure problem where performance depends on memory bandwidth, storage, data movement and cooling — not just raw compute. Scott Patti, vice president of Lenovo's Infrastructure Solutions Group, said the goal is "getting more useful performance from every watt," while addressing power and cooling constraints and speeding the move from deployment to operational AI.

What does the new hardware actually deliver?

Three systems carry the compute story:

• ThinkSystem SC755 — the dual-socket, Neptune liquid-cooled flagship. It also offers Dynamic Partitioning, letting customers convert two dual-socket systems into single sockets and back, so compute allocation flexes with workload demand.
• ThinkSystem SR755 — aimed at compute density and memory bandwidth, targeting agentic AI infrastructure, electronic design automation, financial modeling, engineering and large-scale virtualization. Lenovo's internal analysis reports lower power consumption and better economics than dual-socket configurations.
• ThinkSystem DE6800 and DE6500 storage — both built for latency-sensitive block-storage workloads. The DE6800 targets high-throughput reliability; the DE6500 offers economical performance with simpler deployment and management.

Lenovo has also validated the Cornelis CN5000 fabric, giving customers another high-performance interconnect option intended to cut data-movement bottlenecks in demanding AI and HPC environments. One caveat worth keeping in mind: Lenovo's performance and efficiency claims are vendor-test results, and the company itself notes they vary with configuration, workload, facility water temperature and data-center conditions.

Services: the part after the hardware arrives

The services expansion covers two scales. Hybrid AI Factory Services now handle GPU deployment, configuration and optimization, with multi-node support for Red Hat AI Enterprise and NVIDIA AI Enterprise with NVIDIA Run:ai — two software paths for deploying, governing and scaling workloads, including agentic AI and physical AI.

At the larger end, expanded Gigafactory Services span concept and build phases through deployment and production optimization, with NVIDIA Infrastructure Services expertise included for specialized deployment and cluster validation. Lenovo can also connect organizations to Digital Realty's colocation network and the Digital Realty Innovation Lab, where they can validate AI infrastructure, applications and workload requirements before committing.

For organizations needing colocation space, Lenovo offers a single point of contact with scalable infrastructure, modular services and flexible financing through TruScale Infrastructure as a Service. The underlying bet: installation is only one part of reaching operational AI capacity.

Proof points from Florida and Hollywood

The University of Florida's fourth-generation HiPerGator cluster, deployed with Mark III Systems, combines Lenovo, NVIDIA and AMD technology. Lenovo's case study counts 60,000 AMD EPYC server CPU cores and 1,552 NVIDIA GPUs, handling 33 million research requests in one year. The TOP500 entry for HiPerGator AI lists 17.49 PFlop/s of measured Linpack Rmax performance against a theoretical 19.00 PFlop/s peak — installed in 2025 and ranked 125th on the June 2026 list.

DreamWorks Animation shows the facilities side. It replaced a traditional air-cooled render farm with Lenovo ThinkSystem servers using Neptune liquid cooling. Lenovo's case study reports up to 100% effective heat removal, a 6-degree increase in the chilled-water set point, and up to 10% lower annual data-center power costs. The higher set point cuts energy use and creates headroom for additional liquid-cooled infrastructure.

What does this change for operators? For anyone power-constrained, the practical implication is clear: cooling, networking and deployment services can matter as much to usable AI capacity as the processors themselves — and Lenovo now sells the whole stack.