Nexthop AI announced a $500 million Series B funding round that values the Santa Clara-based company at $4.2 billion, alongside the launch of new AI data center switching platforms and a “Disaggregated Spine” architecture designed for hyperscale and NeoCloud operators. The round was led by Lightspeed Venture Partners with participation from Andreessen Horowitz, Altimeter, and existing investors. The capital will support development of next-generation AI networking systems optimized for hyperscale infrastructure.
The company also introduced a portfolio of switches targeting scale-out, scale-across, and front-end networking for cloud and AI data centers. These platforms support open network operating systems such as SONiC and FBOSS and include turnkey software options through Nexthop’s SONiC-based network operating system. Nexthop said the systems emphasize deployment speed, energy efficiency, and operational visibility for large AI clusters.
Central to the launch is Nexthop’s Disaggregated Spine architecture, developed with a hyperscale partner. The design separates traditional chassis-based systems into functional tiers, including a scale-across leaf layer facing the data center fabric and a spine tier facing the data center interconnect. Nexthop says this approach reduces cost and power consumption by approximately 30% compared with legacy chassis architectures while enabling open networking software and hyperscaler-driven customization.
• Nexthop AI raised $500M in Series B funding, valuing the company at $4.2B
• Funding led by Lightspeed Venture Partners with participation from Andreessen Horowitz and Altimeter
• New switching platforms target scale-out, scale-across, and front-end AI data center networking
• Disaggregated Spine architecture separates leaf and spine tiers to optimize performance and efficiency
• Architecture supports open networking platforms including SONiC and FBOSS
• New platforms include:
– NH-4010: 51.2 Tbps switch using Broadcom Tomahawk 5 silicon, designed for 15–20% power savings
– NH-4220: 102.4 Tbps air-cooled switch based on Broadcom Tomahawk 6 silicon
– NH-5010: deep-buffer scale-across spine switch based on Broadcom Qumran 3D silicon
• Systems include real-time telemetry, congestion management, load balancing, and optics monitoring
• Platforms are already shipping to leading hyperscalers
“Our relentless focus on innovation and deep customer partnerships has driven the development of highly customized JDM solutions for the largest operators and turnkey platforms for NeoClouds,” said Anshul Sadana, founder and CEO of Nexthop AI.
🌐 Analysis: Nexthop AI’s funding and product launch reflect the accelerating demand for new network architectures designed specifically for AI clusters. Hyperscale operators are increasingly exploring scale-across and disaggregated network designs to manage power consumption, fabric congestion, and east-west traffic growth as GPU clusters expand to hundreds of thousands of accelerators. Nexthop’s focus on open networking software and co-development with hyperscalers mirrors broader industry trends across AI infrastructure, where vendors such as Broadcom, NVIDIA, and Arista Networks are also introducing platforms designed to support 800G and emerging 1.6T Ethernet fabrics.
🌐 We’re tracking the latest developments in AI infrastructure. Follow our ongoing coverage at: https://convergedigest.com/category/ai-infrastructure/
🌐 We’re launching the “Data Center Networking for AI” series on NextGenInfra.io and inviting companies building real solutions—silicon, optics, fabrics, switches, software, orchestration—to share their views on video and in our expert report. To get involved, send a note to [email protected] or [email protected].
