DriveNets secures $410M Series D to scale AI networking

DriveNets’ $410 million Series D is another sign that the AI gold rush is spilling heavily into the less glamorous parts of the stack: networking and interconnect. As GPU clusters balloon in size, the limiting factor increasingly becomes how fast you can move data between accelerators, not just how many chips you have. DriveNets’ Ethernet fabric targets precisely this chokepoint, promising higher utilization and lower cost per token for hyperscalers, NeoClouds, and large labs building their own clusters.

For the race to AGI, investments like this underpin the physical substrate on which future frontier models will be trained and served. A cash‑flow positive infra vendor in Israel attracting this scale of capital, including participation from AMD, suggests hyperscaler‑style disaggregation is moving into mainstream AI cluster design: commodity switches plus specialized software instead of monolithic proprietary boxes. That could ultimately lower barriers for well‑funded non‑US players—state labs, sovereign clouds, or late‑stage startups—to stand up cutting‑edge compute.

It also deepens the symbiosis between chipmakers and networking software. AMD’s presence in the round indicates that GPU vendors understand that without a high‑performance fabric, their silicon can sit idle. If DriveNets can deliver on promised performance and multi‑vendor interoperability, it will quietly help both expand global AI capacity and reduce the per‑unit cost of experimentation, pulling more AGI‑relevant work into the feasible zone.