Nvidia's Ethernet platform for AI data centers, combining Spectrum switch silicon with SuperNIC adapters to build high-throughput GPU cluster fabrics.
NVIDIA and Cisco announced plans for USFP, a pluggable optic with 400Gb/s per lane and a stated ceiling of 12.8Tb/s per module. The first planned modules have 16 lanes, or 6.4Tb/s each way, so the ceiling needs 32 lanes or a two-way count.
Reality
Upscale's Token Fabric puts its SkyFabriX scale-up silicon and Nvidia Spectrum-X scale-out switches under one network OS, due in early 2027. For planners, the open question is whether one management layer holds across two suppliers' chips and two traffic regimes.
Reality
The accelerator is in full production and the headline number is a single-request generation rate at 100,000 tokens of context. That is a different purchase order than throughput.
Perspective Coverage
3 publishersReality
Rack-scale Secure AI Factory becomes orderable in September, with Supermicro compute from October. What is being sold is validated integration, and the loss line is idle capacity.
Reality
NVIDIA's open source Topograph discovers cluster topology from five named clouds or from ibnetdiscover on-premises, then republishes it as node labels, Slurm config or Slinky ConfigMaps whenever a watched part of the cluster changes.
Reality
A vendor panel at Supermicro's Open Storage Summit argues GPUs no longer set token cost alone. The useful corollary: chat, batch and agentic work each need their own capacity plan.
Reality