Nvidia has invested over $900 million to bring Enfabrica’s CEO, Rochan Sankar, along with key team members, on board while licensing the startup’s specialized networking technology.
Enfabrica, a Silicon Valley AI hardware startup, tackles a core challenge in scaling AI: connecting tens of thousands of GPUs efficiently so they operate as a single system without bottlenecks. Their technology reportedly supports clusters of up to 100,000 GPUs before performance drops.
This move reflects a growing trend in AI: acquiring talent and IP through “acqui-hires” rather than full acquisitions, likely to sidestep lengthy regulatory hurdles.
Sankar, an Indian-origin technologist with prior experience at Broadcom and other semiconductor firms, will lead Nvidia’s integration of Enfabrica’s accelerated compute fabric (ACF) and SuperNIC architecture into its AI infrastructure.
The broader goal: Nvidia is doubling down on controlling both GPU hardware and the networking/memory fabrics needed to efficiently scale generative AI workloads, solidifying its edge in full-stack AI infrastructure.
