AI progress depends as much on electrons as on algorithms—and Rochan Sankar has spent his career proving it. As Founder and CEO of Enfabrica, he’s tackling one of the least discussed problems in AI: how data actually moves inside the data center.
After leading billion-dollar networking product lines at Broadcom, Sankar saw that existing architectures couldn’t keep up with the rise of accelerated computing. Enfabrica’s hardware moves data between GPUs, CPUs, and memory at multi-terabit speeds, while its systems disaggregate memory from compute to improve efficiency and lower the cost of AI processing.
Sankar believes true AI scale depends on this layer of infrastructure—the one most people never see. It’s slow, capital-intensive work, but it’s what allows the next generation of models to train faster, cost less, and reach more users.