Infrastructure built around your compute.

Explore how the NX brings power, cooling, space and network pathways together around your rack architecture.

Explore compute compatibility.

The four platforms shown are the base architectures our reference layouts are built around. Other rack platforms are supported, with configurability to 800 VDC and 600 kW per rack. Select a platform, NX configuration and rack architecture to see how a reference layout brings compute and facility requirements together.

Per block
16 × 230 kWCompute racks
16 × 50 kWSupport racks
16Open
NX-2 total
2,304GPUs
9.0MW IT load
11.2MW peak
64Racks
Reference layouts up to 230 kW per rack · NX configurable to 800 VDC and 600 kW per rack

Reference layouts from Flexnode IT reference architectures, September 16, 2026. Illustrative only. Each project is reviewed against its own requirements; a reference layout is not a project approval or a compatibility certification.

Rack platforms

Rack platforms with NX reference layouts. Power figures are per rack.

Rack platformRack powerPeak electrical design power (EDPp)Accelerators per rackCoolingRelease year
NVIDIA GB200 NVL72 DGX132 kW192 kW72Direct liquid cooling2025
NVIDIA GB300 NVL72 DGX142 kW192 kW72Direct liquid cooling2026
NVIDIA GB300 NVL72 MGX142 kW192 kW72Direct liquid cooling2026
NVIDIA Vera Rubin VR200 NVL72 DGX230 kW300 kW72Direct liquid cooling2027
NVIDIA Vera Rubin VR200 NVL72 MGX230 kW300 kW72Direct liquid cooling2027
AMD Helios MI455XRequest documentation
NVIDIA DGX B200Request documentation
NVIDIA DGX Rubin NVL8Request documentation
NVIDIA Vera Rubin NVL4Request documentation

Reference configurations

Each reference configuration provides two scalable units per NX block. Support rack positions are totals across both units.

Reference configurationCompute racksSupport rack positionsScalable unitsCooling
NVIDIA GB200 NVL72 DGX16162Direct liquid cooling
NVIDIA GB300 NVL72 DGX16162Direct liquid cooling
NVIDIA GB300 NVL72 DGX SuperPOD16242Direct liquid cooling
NVIDIA GB300 NVL72 MGX16162Direct liquid cooling
NVIDIA Vera Rubin VR200 NVL72 DGX16162Direct liquid cooling
NVIDIA Vera Rubin VR200 NVL72 DGX SuperPOD14162Direct liquid cooling
NVIDIA Vera Rubin VR200 NVL72 MGX16162Direct liquid cooling
AMD Helios MI455XRequest documentation
NVIDIA DGX GB200 SuperPODRequest documentation
NVIDIA DGX B200 SuperPODRequest documentation
NVIDIA DGX Rubin NVL8 SuperPODRequest documentation

Built around the racks you deploy.

The NX brings the essential facility interfaces together around the requirements of your compute architecture.

01

Power

Facility distribution to the rack interface.

02

Cooling

Cooling interfaces and their connection to facility heat rejection.

Your rack + computeInside the NX
03

Space

Rack positions, clearances and service access.

04

Network pathways

Routes and entry points for the customer's network.

Customer, OEM, Flexnode and partner responsibilities are defined for each project.

Bring us your architecture.

Can Flexnode support a rack platform not shown here?

Bring the platform's power, cooling and space requirements. Each architecture is reviewed against the NX design and the site before any commitment is made.

What information is needed for a first review?

Rack platform, power per rack, cooling method, rack count and target timeline. Network and space requirements refine the layout.

Does a reference layout confirm compatibility?

A reference layout shows how a platform can fit within an NX module. Compatibility is confirmed through a project-specific review.