
Ayar Labs develops co-packaged optical interconnects that replace copper I/O to scale AI clusters beyond the rack.
ayarlabs.comCrawled Oct 1, 2026
Explore Ayar Labs' blog for expert commentary on optical I/O advancements for AI scale-up, tech trends, and company updates.
Vladimir Stojanovic, Ayar Labs co-founder, discusses overcoming AI architecture challenges with advanced optical I/O solutions for improved AI training and inferencing efficiency.
What’s the difference between in-package optical I/O and co-packaged optics (CPO)? Learn more about advances in optical interconnects.
Discover how Optical I/O enhances AI infrastructure with scalable, energy-efficient designs that reduce costs and boost performance.
Ayar Labs is solving the I/O bandwidth and power bottlenecks by moving data using light. See how we are enabling the next phase of Moore’s Law. Learn more on our Media page.
At Ayar Labs, you'll work at the frontier of photonics and AI infrastructure to shape the future of AI scale-up. Explore open roles and apply today.
Ayar Labs is solving the I/O bandwidth and power bottlenecks by moving data using light. See how we are enabling the next phase of Moore’s Law. Learn more on our AI Infrastructure Performance Comparison Tool page.
Meet our Executive Team, Board of Directors, and Advisors here.
Discover Ayar Labs’ journey in revolutionizing data transfer with co-packaged optics. Learn about our vision, leadership, and more.
Contact us via email or phone for general information or inquires. We’d love to hear from you.
Ayar Labs’ co-packaged optics for AI scale-up. Built and backed by semiconductor ecosystem leaders.
Read our Privacy Policy here.
OpenAI hardware vice president Richard Ho discusses Jalapeño, the company’s inference accelerator, in an interview with Dr. Ian Cutress. The article describes a single balanced design with 216 GiB of HBM4 and reports HotChips 2026 claims of 1.5–1.9× higher peak-throughput performance per watt and 1.7–3.6× lower latency than NVIDIA GB200/GB300. Ho also explains how OpenAI models assist chip design and kernel optimization, and recounts his prior work on D. E. Shaw’s Anton supercomputers and Google’s TPU program.
Sep 2026 · More Than MooreNews article
Ayar Labs raises a $150 million extension to its $500 million Series E, announced in March 2026, bringing the financing to $650 million in total. The Silicon Valley startup is developing technology that links AI chips directly through optical connections.
Sep 2026 · ReutersNews article
Ayar Labs raises an additional $150 million from an unnamed investor, extending its Series E to $650 million. The optical networking startup says TeraPHY replaces copper connections with optical interconnects to link AI chips, while co-packaged optics place its optical engine in server ASIC sockets. CEO Mark Wade says the funding supports manufacturing-ready products, foundry and packaging work, and engineering capacity as the company targets volume-production qualification by the end of 2027.
Sep 2026 · SiliconAngleNews article
Ayar Labs joined the NVIDIA NVLink Fusion ecosystem, making its co-packaged optics optically and electrically compatible with NVIDIA technologies. The announcement positions the technology for rack-scale AI infrastructure.
Jun 2026 · ayarlabs.comNews article