Ayar Labs and Wiwynn Partner for Development of Co-Packaged Optical AI Infrastructure

Why It Matters

The solution removes copper’s bandwidth bottleneck, enabling hyperscale AI workloads to scale faster and operate more efficiently, a critical need for cloud providers and AI‑intensive enterprises.

Key Takeaways

• Co‑packaged optics replace copper in rack‑scale AI servers
• System scales to 1,024 accelerators, >100 Tbps per accelerator
• Liquid‑cooling and ELSFP support high‑power, hyperscale workloads
• Ayar Labs secured $500 M Series E funding from major chipmakers
• Partnership leverages Wiwynn’s manufacturing for rapid deployment

Pulse Analysis

The shift from copper to silicon‑photonic interconnects is reshaping data‑center architecture, especially for AI workloads that demand terabit‑scale bandwidth. Traditional copper links suffer from signal attenuation and power inefficiency beyond a few meters, limiting rack density and scaling potential. Co‑packaged optics integrate photonic components directly with compute dies, dramatically shortening the optical path and reducing latency, while maintaining the high data rates required for modern transformer models.

Ayar Labs’ TeraPHY engines, paired with Wiwynn’s rack‑level system design, create a modular platform that can host over a thousand accelerators in a unified fabric. Each accelerator enjoys more than 100 Tbps of optical connectivity, facilitated by SuperNova remote light sources and ELSFP modules that simplify laser management. The liquid‑cooled chassis ensures thermal stability for high‑power AI chips, addressing one of the biggest operational challenges in hyperscale environments. This combination of optical bandwidth, power efficiency, and serviceability positions the offering as a compelling alternative to legacy copper‑based servers.

The partnership arrives at a pivotal moment as AI adoption accelerates across industries. Ayar Labs’ recent $500 million Series E round, featuring investors like Nvidia and AMD, underscores market confidence in photonic solutions. By leveraging Wiwynn’s manufacturing scale, the duo can rapidly deliver system‑ready products, potentially setting a new performance baseline for cloud providers. Competitors are also exploring silicon photonics, but the integrated, rack‑scale approach may give Ayar and Wiwynn a first‑mover advantage in the emerging optical AI infrastructure market.