NVIDIA DGX H100
8 × H100 SXM5 per node, with 8 compute rails at 400G. That is 3.2 Tb/s of scale-out network per node.
- GPUs per node8 × H100 SXM5
- Compute rails8 × 400G
- Scale-out bandwidth3.2 Tb/s
- Storage links2 × 400G
- In-band links2 × 100G
- BMC ports1
- Rack units8U
- Typical power10.2 kW
- Maximum power10.2 kW
What this means for the fabric
8 compute NICs means 8 rails. In a rail-optimized design each of those rails is sized on its own against the node count, then multiplied back up, so the leaf tier is driven by the rail count and not by the total endpoint count.
Storage and in-band management are separate fabrics with their own switches and their own optics, and the 1 BMC port per node feed the out-of-band network, which also has to carry every switch and every PDU in the cluster.
Power and racks
At 10.2 kW typical draw, this node fills a rack's power budget long before it fills the space. A 45 kW rack takes 4 of them on power alone, against 5 on rack units. Power is what actually sets the rack count.
Notes
The reference rail-optimized node. A SuperPOD scalable unit is 32 of these (256 GPUs), which fills the 32 downlinks of one leaf switch per rail exactly. Rack density is usually power-limited rather than RU-limited.
Source: NVIDIA DGX H100 documentation and DGX SuperPOD data center design guide, verified 2026-08. 8U and ~10.2 kW confirmed. Confirmed 2026-08-06