AI COMPUTE•NVIDIA•GPU CLOUD•DATA CENTERS•SEMICONDUCTORS•ENERGY•FUNDING•M&A•AI COMPUTE•NVIDIA•GPU CLOUD•DATA CENTERS•SEMICONDUCTORS•ENERGY•FUNDING•M&A•
Cloud

NVIDIA-backed Upscale AI introduces networking platform for mixed-chip data centers

Token Fabric aims to link AI processors from multiple suppliers while reducing networking complexity.

By GPU Data Hub Desk · GPU Data Hub·Editorial standards··United States

Upscale AI, a networking startup backed by NVIDIA and other investors, announced a platform called Token Fabric on 8 October 2026. According to Reuters, the system combines dedicated networking hardware with software intended to connect different suppliers' AI processors inside data centers.

The product is aimed at cloud operators and other builders of large AI computing facilities that want flexibility across accelerator brands. Upscale says its software can identify network bottlenecks and equipment needing attention; the goal is to reduce the time costly AI chips spend waiting for data.

Reuters reports that the first stage, connecting groups of AI chips within a data center, is planned for the fourth quarter of 2026, followed by a phased rollout through 2027. These are company plans, not independently verified performance results.

Why it matters: networking is a growing part of total AI infrastructure cost. Systems that support mixed accelerator fleets could give cloud providers more deployment options, but real-world interoperability and efficiency will need to be demonstrated.

Original source

This summary was written by the GPU Data Hub desk from reporting published by Reuters on 10 Oct 2026, 21:47. Read the full article at the original publisher.

Read Original Source →

By subscribing you consent to receive the daily briefing. Unsubscribe at any time. See our privacy policy.

LinkedInX

More from GPU Data Hub

Latest →

AI is a boon for 3 internet traffic cop stocks. Cramer says only 1 is a buy right now

Surging volumes of web activity generated by artificial intelligence agents are opening up expansion opportunities for content delivery and edge infrastructure providers. Digital delivery specialists including Cloudflare, Akamai, and Fastly are experiencing increased demand as networks handle the rising load from automated workloads.

Cloud··CNBC Technology

OpenStack Hibiscus Strengthens Trusted Infrastructure for the AI Era

The OpenStack community has released its 34th cloud platform iteration, dubbed Hibiscus or version 2026.2. The update is tailored to meet modern artificial intelligence demands by bolstering confidential computing, strengthening platform security, and improving operational efficiency across large deployments running accelerated workloads.

Cloud··HPCwire