Cisco is widening its bet on the AI data center. According to SiliconANGLE, the company's Secure AI Factory with Nvidia is expanding to rack scale for 2026 — moving the joint offering from individual servers toward complete, pre-integrated racks of AI compute.
A separate report from Portal ERP says Cisco is expanding its AI infrastructure partnership to include not just Nvidia but also Supermicro, a server maker that has become one of the more visible suppliers of dense, GPU-packed systems.
The phrase "AI factory" is industry shorthand for a data center built for one purpose: training and running AI models at scale. Rather than buying networking, servers, storage, and security separately and hoping they cooperate, customers get a validated bundle designed to be racked and switched on. Cisco's contribution is largely the networking and security layer; Nvidia supplies the accelerated computing that AI workloads depend on.
The available reporting is thin on specifics — the source items announce the expansion and the widened partnership without detailing pricing, availability dates, or performance figures beyond the 2026 timeframe attached to the rack-scale push.
Still, the direction is the point. As enterprises move from AI pilots to production deployments, the bottleneck shifts from model quality to plumbing: power, networking, and security around expensive GPUs. Vendors that once sold components are increasingly selling the whole assembled room, and Cisco pulling Supermicro into the arrangement suggests it wants more hardware options inside that room rather than fewer.
It matters because the companies that own the standard AI rack will shape what running AI costs — and who gets to sell it — for years to come.