News
Nvidia estimates co-packaged optics could reduce power consumption by 70 percent
Sunday, March 1, 2026 at 09:47 PM
Nvidia has calculated the potential impact of implementing co-packaged optics (CPO) in its AI infrastructure. The company estimates that the shift from traditional pluggable optics to CPO can reduce power consumption by up to 70 percent and decrease connectivity costs to one-tenth of current levels.
Context
At the GTC 2026 conference, Nvidia reinforced its transition toward optical interconnects by detailing the efficiency gains of its co-packaged optics (CPO) platform. The company estimates that integrating silicon photonics directly into the switch package can reduce power consumption by up to 70% per 800G of bandwidth compared to traditional pluggable transceivers. This innovation is central to Nvidia's roadmap for scaling 'AI factories' to millions of GPUs, where traditional copper cabling faces physical distance and power density limitations.
Following the initial reveal of the Spectrum-X and Quantum-X photonics switches in 2025, Nvidia has expanded its ecosystem through collaborations with TSMC, Coherent, and SK Hynix. The roadmap now targets the 2028 release of the Feynman GPU architecture and NVLink8 CPO, which will bring these efficiency gains to scale-up interconnects. Industry analysts at TrendForce project that CPO penetration in AI data centers could reach 35% by 2030 as operators like CoreWeave and Lambda prioritize lower operational costs and 10x higher network resiliency.
Sources (10)
NVIDIA Corporation - NVIDIA Announces Spectrum-X Photonics, Co-Packaged Optics Networking Switches to Scale AI Factories to Millions of GPUs
Silicon Photonics Networking for Agentic AI | NVIDIAHow Industry Collaboration Fosters NVIDIA Co-Packaged Optics | NVIDIA Technical BlogCopackaged optics have officially found their killer app • The RegisterCo Packaged Optics (CPO) – Scaling with Light for the Next Wave of InterconnectMicro-LED co-packaged optics cut power consumption to just 5% that of copper cablesNVIDIA Compute Architecture Paves the Way for Scale-Up Optical Interconnects; CPO Penetration in AI Data Centers Expected to Rise Steadily, Says TrendForceNvidia updates data center roadmap with Rosa CPU and stacked Feynman GPUs — optical NVLink, Groq LPUs with NVFP4, and NVLink also on deck | Tom's Hardware
Related Companies
Nvidia
NVDA