At OCP in San Jose this week, Cisco and NVIDIA announced plans for an Ultra-Dense Small Form Factor Pluggable (USFP), a new high-density 400Gb/s-per-lane optical module designed to equip the hardware ecosystem for the next generation of AI factories using pluggable transceivers.
𝐊𝐞𝐲 𝐃𝐞𝐭𝐚𝐢𝐥𝐬 𝐨𝐟 𝐭𝐡𝐞 𝐔𝐒𝐅𝐏 𝐀𝐧𝐧𝐨𝐮𝐧𝐜𝐞𝐦𝐞𝐧𝐭:
𝐍𝐞𝐰 𝐎𝐩𝐭𝐢𝐜𝐚𝐥 𝐒𝐭𝐚𝐧𝐝𝐚𝐫𝐝 - the USFP is designed to be the first module to support 400Gb/s-per-lane electrical signals.
𝐁𝐚𝐧𝐝𝐰𝐢𝐝𝐭𝐡 𝐚𝐧𝐝 𝐂𝐚𝐩𝐚𝐜𝐢𝐭𝐲 - it supports up to 12.8Tb/s of pluggable capacity for AI Ethernet fabrics, delivering 4x the bandwidth density and 1.3x the power efficiency of previous 200Gb/s-per-lane optics.
𝐄𝐜𝐨𝐬𝐲𝐬𝐭𝐞𝐦 𝐂𝐨𝐥𝐥𝐚𝐛𝐨𝐫𝐚𝐭𝐢𝐨𝐧 - Cisco and NVIDIA are working alongside hardware partners TE Connectivity and Amphenol to manufacture the required cages, connectors and modules.
𝐁𝐫𝐨𝐚𝐝𝐞𝐫 𝐎𝐂𝐏 𝐑𝐨𝐥𝐥𝐨𝐮𝐭 - the launch aligns with Cisco's broader release of new Cisco Silicon One systems (including G300-powered systems and 1.6T G302 silicon) to scale open AI infrastructure.
𝐔𝐒𝐅𝐏 𝐯𝐬 𝐗𝐏𝐎 - the difference between Cisco’s USFP and Arista’s XPO (backed by Coherent, Lumentum and the Open XPO MSA) comes down to competing visions on how AI data centers should solve the thermal wall and faceplate density limit.
Cisco’s USFP focuses on air-cooled and hybrid-cooled architectures. The company designed USFP around 400G-per-lane signalling to maximise electrical-to-optical efficiency and shrink module size, allowing operators to keep existing air-cooled rack designs without immediately overhauling facility plumbing. Meanwhile, Arista’s XPO is betting on direct liquid cooling. Each XPO module embeds a miniature liquid cold plate directly on the pluggable unit, allowing it to dissipate up to 400W per module.
𝐒𝐮𝐩𝐩𝐥𝐲 𝐂𝐡𝐚𝐢𝐧 𝐚𝐧𝐝 𝐒𝐢𝐥𝐢𝐜𝐨𝐧 𝐄𝐜𝐨𝐬𝐲𝐬𝐭𝐞𝐦𝐬 - Cisco collaborates closely with NVIDIA and is aligning USFP with NVIDIA’s Spectrum-X networking roadmap and native Silicon One switch chips. Arista, meanwhile, relies on Broadcom merchant silicon and an open optical component supply chain (Coherent, Lumentum) through the Open XPO MSA.
𝐀𝐧𝐚𝐥𝐲𝐬𝐭 Take
• The two standards serve two distinct infrastructure deployment applications.
• USFP is targeted at hyperscalers and enterprises that want maximum faceplate density without requiring immediate liquid-to-pluggable plumbing in every server row. It acts as an evolutionary step for air-cooled high-density networks.
• XPO is targeted at massive single-site GPU clusters where rack power density exceeds 120kW per rack, making liquid cooling mandatory down to the transceiver module. Ultimately, as speeds reach 400G/800G per lane, both pluggable form factors (USFP and XPO) will face competition from CPO.
