We're excited to invest in Nexthop's $500M Series B. AI is forcing a complete rebuild of data center infrastructure, and networking is becoming one of the biggest bottlenecks. @nexthopai is building high-performance Ethernet switches designed natively for AI workloads. AI training and inference rely on thousands of GPUs communicating constantly. But the networking layer between those GPUs is struggling to keep up: - AI clusters require thousands of switches routing traffic between GPUs, servers, and storage nodes across multiple network layers - Each new GPU generation demands dramatically higher bandwidth, lower latency, and greater reliability from the underlying network - Despite this surge in demand, the data center switching market has been consolidating while engineering complexity continues to rise Nexthop is taking a different approach. Instead of adapting legacy networking products built for the cloud, Nexthop designs switches from the ground up for AI traffic patterns, with hardware and software co-designed for hyperscale AI networks. Their platform combines high-performance hardware with deep support for open-source operating systems and open networking standards, giving hyperscalers the flexibility to extend and integrate the network into their own environments. Anshul Sadana spent more than 15 years helping build Arista into the defining networking company of the cloud era. He and the Nexthop team are now building the networking company for the AI era. By @RaghuRaghuram, @shangdaxu, and @appenz(a16z)