What is an AI cluster?
An AI cluster collects computing resources to develop, test, and deploy artificial intelligence (AI) models. AI clusters can be customized to meet specific needs and workloads.
Learn More about AI
· State of the Market Report: AI-Driven Demand Poised to Propel Optical Communication Sector
· Corning’s AI push bolsters fourth quarter Optical Communications results
· Flexential report explores AI infrastructure challenges and investment priorities
As bandwidth needs for AI clusters continue to rise, so do sales of optical transceivers, linear pluggable optics (LPO), and co-packaged optics (CPO) for scale-out and scale-up networks used in AI clusters.
A new LightCounting report forecasts that the market for LPO and CPO for scale-out and scale-up networks used in AI clusters will double in size in two years: from $5 billion in 2024 to more than $10 billion in 2026.
The role of optics in AI is mainly seen in AI clusters.
An AI cluster collects computing resources to develop, test, and deploy artificial intelligence (AI) models. AI clusters can be customized to meet specific needs and workloads.
“We expect this rapid growth to moderate in 2026-2027 as the first wave of excitement about AI subsides,” LightCounting said. “Using LPO and CPO in scale-up networks will return this market to double-digit growth in 2028-2030.”
The Nvidia factor
A significant factor in optical transceiver sales in 2023-2024 was growth in Nvidia’s business.
New designs of Nvidia’s AI clusters require a lot of transceivers.
This points to Nvidia’s shift from copper to optical pluggable transceivers. Before 2023, Nvidia’s AI cluster systems used only copper and active optical cables (AOCs).
Last March, Nvidia unveiled NVIDIA Quantum-X800 InfiniBand and NVIDIA Spectrum™-X800 Ethernet, which offer 800 Gbps throughput.
Rise of pluggables
NDR (400G) InfiniBand connections use pluggable 400/800G SR4/SR8 and DR4/DR8 transceivers instead of AOCs. Also, the latest XDR systems will use 1.6T DR8 and 2xFR4 pluggables, including some LPOs.
Nvidia is also developing CPO, which it will use to scale NVLink connectivity over fiber.
LightCounting expects “LPO and CPO to be deployed in scale-up networks starting in 2026-2027 and reaching high volumes by 2028.”
For related articles, visit the Business Topic Center.
For more information on high-speed transmission systems and suppliers, visit the Lightwave Buyer’s Guide.
To stay abreast of fiber network deployments, subscribe to Lightwave’s Service Providers and Datacom/Data Center newsletters.
About the Author
Sean Buckley
Sean is responsible for establishing and executing the editorial strategies of Lightwave and Broadband Technology Report across their websites, email newsletters, events, and other information products.