Ayar Labs Builds 1,024-GPU Rack Platform

Ayar Labs and ODM Wiwynn on Wednesday unveiled a rack-scale reference platform designed to stitch more than 1,024 GPUs using co-packaged photonics, the companies said. The design targets 100–200 kW per rack versus 600+ kW for copper-based systems, increasing reach and bandwidth while enabling disaggregated multi-rack accelerator clusters; Ayar will show the platform at next week's OFC conference.
Key Points
- 1Announces rack-scale design stitching over 1,024 GPUs using co-packaged photonics and optical interconnects.
- 2Reduces per-rack power to 100–200 kW versus 600+ kW, increasing reach and bandwidth.
- 3Enables disaggregated, multi-rack architectures, letting datacenters scale accelerators without copper limitations.
Scoring Rationale
High novelty and industry-wide scope, but still a reference design rather than immediate, widely available deployment.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
