Super Micro Computer Inc. said Monday that Lambda has rolled out a broad portfolio of Supermicro GPU-optimised servers, including systems built around Nvidia’s new Blackwell architecture, to scale the start-up’s cloud-based artificial-intelligence infrastructure. The deployment features Supermicro’s AI Supercluster racks equipped with Nvidia GB200 and GB300 NVL72 accelerators as well as servers using HGX B200 and H200 GPUs paired with Intel Xeon processors. Lambda says the hardware is already powering a “large-scale AI factory” for training and inference workloads, while liquid-cooling technology is expected to cut energy and cooling costs. The collaboration, which began in June at Cologix’s COL4 Scalelogix data centre in Columbus, Ohio, is aimed at giving Midwestern enterprises faster access to high-performance compute. Supermicro senior vice-president Vik Malyala described the servers as an energy-efficient response to surging demand, and Lambda vice-president Ken Patchett said the company is working toward gigawatt-scale AI factories for labs, enterprises and hyperscalers. Supermicro shares rose 0.8 % to $44.22 in afternoon trading. Analysts forecast that Microsoft, Meta, Alphabet and Apple could collectively spend more than $400 billion on AI infrastructure by 2026, a market the Supermicro-Lambda-Nvidia alliance is positioning itself to serve.