Resolight.ai Bets Photonics Will Fix AI Networks

Resolight.ai Bets Photonics Will Fix AI Networks
Stealth startup Resolight.ai has emerged with a photonic processor it claims can eliminate the core bottleneck in AI infrastructure: the conversion of optical signals back into electrical form for routing and processing. Co-founder and CEO Ofer Shapiro says keeping data in the optical domain throughout the network delivers 10x higher bandwidth, 90% less power consumption, and 10x fewer switches required. The company positions itself against established players like Nvidia, Broadcom, Marvell, and Cisco, arguing their co-packaged optics approaches still rely on electrical switching. Resolight's so-called scale everywhere architecture aims to let software, not network topology, determine how GPUs are clustered across data centers.
Read the original article →