NVIDIA Opens AI Factory Rack Designs and Grid-Resilience Specifications
OCP contributions cover rack interfaces, facility design and how large computing loads respond to grid disturbances.
NVIDIA’s October 12 Open Compute Project announcement extends its AI infrastructure contributions beyond processors to rack design, facilities and grid behavior. The company contributed Vera Rubin rack and compute-tray designs through MGX and highlighted open specifications developed with hyperscale operators.
The MGX 1.2 design includes standardized power and cooling interfaces, blind-mate liquid cooling and support for racks up to 250 kilowatts. These are architectural capabilities, not a statement that every data center can accommodate that load without electrical or cooling upgrades.
NVIDIA highlighted the Open Data Center Specification v0.7, whose document is dated August 27, and the OCP Grid Resilience Specification. These are design references discussed at the October summit, not proof of a new October revision. Grid resilience addresses how computing facilities remain connected and recover load after disturbances. Separately released DSX components cover network configuration, cluster validation, rack management and bare-metal provisioning. The materials do not establish a construction contract, customer order or completed campus.
NVIDIA OCP account, October 12 · Reviewed October 20, 2026. AI Infrastructure.