At the AI Infrastructure Summit 2025 in Santa Clara, Vladimir Stojanovic, CTO & Co-Founder of Ayar Labs, explains how optical I/O is breaking through copper’s limits to unlock massive GPU scaling, higher performance per watt, and new system architectures. This was one of the most anticipated conversations at the summit — and for good reason.
Timestamps
00:00 – Why scaling AI inference requires thousands of GPUs working together
00:22 – How Ayar Labs’ optical I/O connects GPUs across racks and clusters
00:48 – Performance-per-watt advantages: achieving 10–20x improvements
00:58 – Building a high-volume manufacturing pipeline for optical I/O in AI systems
👉 Starting this November, NextGenInfra.io is launching the Data Center Networking Showcase series. If your company is pushing the boundaries of AI infrastructure, join us to share your story with the industry. Want to be involved our video series? Contact info@nextgeninfra.io
https://ngi.fyi/ayarlabs-vs
