NVIDIA and Cisco unveiled USFP standard at OCP Global Summit 2026; NVIDIA contributed Vera Rubin rack designs and additional infrastructure specifications
NVIDIA and Cisco Propose USFP Optics Standard at OCP 2026
USFP, if adopted broadly, would substantially increase front-panel bandwidth density for AI data center networking. Opening the Vera Rubin rack design to OCP extends a pattern of NVIDIA sharing infrastructure specifications with the broader industry while retaining control of core technologies.
The full picture
At the OCP Global Summit 2026, NVIDIA and Cisco introduced the Ultra-Dense Small Form Factor Pluggable (USFP) standard, targeting 400Gb/s per lane, roughly four times the bandwidth density of current 200Gb/s-per-lane pluggable optics, in approximately the same footprint as the existing OSFP form factor. The module supports 16 full-duplex lanes with a roadmap to 32, and NVIDIA claims approximately 1.3x greater power efficiency compared to current 200Gb/s pluggables. NVIDIA and Cisco plan to submit USFP as an open industry specification through a community-led multi-source agreement, with TE Connectivity and Amphenol involved; no sampling schedule or availability date has been announced. Separately, NVIDIA contributed Vera Rubin rack-level designs to the Open Compute Project at the same summit, extending its MGX architecture that was first shared with OCP the prior year. NVIDIA also presented DSX AI factory software updates, MGX 1.2, 800 VDC power specifications, and Ethernet work developed with Meta and Google as part of its OCP contributions. NVIDIA describes its datacenter strategy in terms of 'giga-scale AI factories' that integrate compute, power, and cooling as a unified design.
How it developed
Sources
Related
- Split fromAI factory rack standards at OCP Summit
Want this in your inbox?
I send one email each morning with the stories that moved. If you would rather just read here, that works too.
Subscribe free