Nebius brings NVIDIA Groq 3 LPX online; Rubin NVL72 hits up to 30x throughput per MW on agentic coding
demian_ai · x · 2026-08-25
Nebius announced it is the first AI cloud to bring NVIDIA Groq 3 LPX online, putting ultra-fast inference on the same platform developers use to run leading open models in production.
The framing is the "AI factory": power is the scarce resource, and the metric is completed sequential agent steps inside a fixed megawatt — not peak FLOPS. Vera Rubin NVL72 reportedly delivers up to 30x higher throughput/MW on real agentic coding trajectories (context growth, tool calls, subagents) vs the previous generation.
The platform is codesigned for the full loop: Vera CPU feeds the GPUs with orchestration and tool use, Rubin does the heavy lifting, and LPX handles the low-latency generation step where sequential calls used to pile up.
More from Infra
- TSMC earnings note cross-referenced with WFE model and power checks in CS Atlas — BenBajarin · 2026-10-16
- Fluidstack hires Electrical Engineer for Modular R&D to build frontier AI infrastructure — MxMnr · 2026-10-16
- Jensen Huang joins BG2 podcast: does AI revenue justify capex, '27 bottlenecks — elonmusk · 2026-10-16
- Mistral's on-device infra rack revealed: DGX Spark, Mac Studio and more — gregmushen · 2026-10-16
- NVIDIA and Sweden's Wallenberg launch Sferical One, a shared AI supercomputer — nvidia · 2026-10-16
- Next infra frontier: self-evolving systems that adapt and self-verify — _shreya_s · 2026-10-16