AMD and NVIDIA release CPU-only racks for AI tool execution
AMD and NVIDIA announced CPU-only racks on October 12, designed to handle AI agent tasks like tool calling and program execution. AMD’s Helion rack stacks up to 32,768 cores (6th-gen AMD CPUs) and connects to separate GPU-based inference infrastructure, with shipments starting in November. NVIDIA released design specs for its Vera CPU rack at the OCP conference, focusing on orchestration and…
Key points
- AMD’s Helion rack offers up to 32,768 CPU cores for AI tool execution, shipping November 2026
- NVIDIA released Vera CPU rack specs for orchestration but omitted core-count details or performance data
- Neither company disclosed cloud provider orders or benchmark comparisons with GPU alternatives
The move reflects a shift toward modular AI infrastructure, where tool execution and model inference are physically separated. AMD’s emphasis on core count contrasts with past GPU-centric designs, while NVIDIA’s specs align with CEO Jensen Huang’s March 2026 comments on CPU efficiency for orchestration. Industry analysts note this could reshape AI workload planning, though actual adoption and performance remain unconfirmed.
Model pages: Muse Wonder → · Muse Wonder Full → · Claude Opus 5.5 →
The story so far
5 episodes →- AMD and NVIDIA release CPU-only racks for AI tool executionthis story
SecondSource Evening Edition 2026/10/13 | AMD and NVIDIA announce CPU-only racks for AI tool execution—neither company's announcement mentions performance or orders
note.com · 13 October 2026
Loading the full article…
This text was published by note.com and written by SecondSource. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
2sources- Is AMD Even a Real Threat to Nvidia or Just a Pricing Lever?Press · 247wallst.com ·
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Hardware & Compute
All →- Nvidia reportedly invests in AI inference chip rival d-Matrix · 5 src
- NVIDIA adds cuPotential library to CUDA-X for AI chip simulations · 2 src
- Microsoft unveils Azure Maia 300 chip for 40% more tokens per megawatt in 2027 deployments · 10 src
- Epic Games unveils AI-assisted launcher with faster load times · 1 src
- Salience Labs raises $125M Series B to develop optical switches for AI datacenters · 2 src
Comments
via GitHub Discussions