Lenovo Group on Tuesday unveiled an expanded suite of enterprise hardware and deployment services aimed at the rapidly converging frontiers of artificial intelligence (AI) and high-performance computing (HPC).
The rollout arrives as enterprise computing shifts from static model training toward continuous inference and autonomous, agentic AI workloads, shifts that are testing the physical and thermal limits of modern data centers.
At the center of the announcement are Lenovo’s next-generation ThinkSystem servers, powered by AMD’s 6th Generation EPYC SP7 processors. To tackle mounting power density challenges, Lenovo is leaning heavily into proprietary thermal engineering with the ThinkSystem SC755. The dual-socket system integrates Neptune liquid cooling, designed to extract up to 100% of system heat via water loops.
Lenovo claims the architecture yields up to 40% higher sustained compute performance compared to conventional air-cooled alternatives, while curbing facility-level water and evaporative cooling consumption. The server also introduces Dynamic Partitioning, allowing operators to dynamically reconfigure dual-socket systems into single sockets as workload requirements pivot.
For virtualization and memory-intensive enterprise tasks — including electronic design automation and financial modeling — Lenovo introduced the single-socket ThinkSystem SR755. The company positioned the machine as a high-density, lower-power option engineered to deliver favorable unit economics against traditional dual-socket setups.
Data throughput and pipeline bottlenecks are addressed through new ThinkSystem DE6800 and DE6500 high-speed block-storage arrays, alongside validation for Cornelis Networks’ CN5000 fabric interconnects.
“AI and high-performance computing are converging around a simple customer need: the ability to turn increasingly complex workloads into real outcomes,” said Scott Patti, vice president of Lenovo’s Infrastructure Solutions Group. Patti emphasized that enterprise adoption hinges on maximizing performance per watt and shortening the lag between hardware delivery and operational deployment.
Dan McNamara, senior vice president and general manager of Compute and Enterprise AI at AMD, noted that pairing 6th Gen EPYC silicon with high-density system designs provides the efficiency necessary to run modern AI inferencing alongside traditional scientific research at scale.
Beyond physical silicon, Lenovo introduced end-to-end consulting and lifecycle capabilities targeting large-scale “AI factories.” The expanded Hybrid AI Factory Services feature multi-node integration across Red Hat AI Enterprise and NVIDIA AI Enterprise with Run:ai orchestration. Lenovo also integrated NVIDIA Infrastructure Services (NVIS) expertise to streamline cluster validation for gigawatt-scale deployments.
To support enterprises constrained by localized power shortages and footprint limitations, Lenovo partnered with global data center provider Digital Realty. The agreement establishes a direct conduit to colocation facilities and the Digital Realty Innovation Lab, backed by procurement options through Lenovo’s TruScale Infrastructure-as-a-Service model.
