Скачать или смотреть #AIInfraSummit25

How do you fix GPU under-utilization at scale? In this short interview, Moshe Tanach, Co-Founder & CEO of NeuReality, explains their 1.6T Ethernet NIC and in-network acceleration approach to boost GPU utilization and reduce bottlenecks. Recorded at the AI Infrastructure Summit 2025 (Santa Clara, CA).

Time-stamps
0:09 – NeuReality’s 1.6T NIC for AI scale-out
0:37 – Addressing GPU wait times & system bottlenecks
1:03 – Double the throughput, lower latency, in-network compute; utilization target
1:31 – ROI: recapturing CapEx/OpEx from idle GPUs
1:36 – Gen-1 vs. next-gen notes & ecosystem
1:49 – Ethernet approach vs. InfiniBand for large-scale deployment

Want to spotlight your innovations? Join our Data Center Networking Showcase Series on NextGenInfra.io starting in November. Limited slots for companies shaping AI & cloud networking—reach out to collaborate.