Lambda completes world's first large-scale CPO deployment linking 10,000+ Blackwell Ultra GPUs
TheZachMueller · x · 2026-11-17
Lambda announced the world's first large-scale co-packaged optics (CPO) deployment at its Quebec campus, connecting 10,000+ NVIDIA Blackwell Ultra GPUs via NVIDIA Photonics switches (144x 800Gb/s ports, liquid-cooled, 18 removable external light sources).
Key results vs. pluggable optics:
- Up to 10% lower latency and up to 5x better power efficiency
- Optical engines sit beside the switch ASIC, cutting electrical paths from inches to millimeters — less signal loss, no retimer chips
- Error rates stayed stable as the fabric grew, meaning fewer interruptions for long training jobs and more power budget for compute
Announced at SC26, following Lambda's early CPO adoption first disclosed in November 2025.
More from Infra
- DeepSeek Flash on Mac Studio M5 Ultra (512GB): 54 t/s generation with 128K context — vhthc · 2026-11-17
- Peter Diamandis: The Country Building Gigawatts Fastest Will Set the Price of Intelligence — PeterDiamandis · 2026-11-17
- Gemini Managed Agents Adds Self-Hosted Sandboxes: Google Runs the Loop, You Run the Tools — _philschmid · 2026-11-17
- RWKV-8 RPU preview hits 12,500+ tokens/s real-time inference — kastnerkyle · 2026-11-17
- Your AI budget is too generous and too restrictive at once: routing saves, flat caps kill rewrites — ruslansv · 2026-11-16
- Edge inference with llm-d: prefill-decode disaggregation on Qualcomm Cloud AI 100 Ultra — HowDevelop · 2026-11-16