Particle.news
Download on the App Store

IBM Commits $240 Million to Together AI for Large-Scale Inference Cluster on IBM Cloud

The agreement aims to give enterprises a fast, governed place to run open-source models under IBM’s cloud and sovereign-cloud controls.

Overview

  • IBM and Together AI signed a $240 million multi-year deal announced Tuesday to build a GPU-powered inference cluster that will run on IBM Cloud.
  • The cluster will use Nvidia HGX B300 systems and Spectrum-X networking and is targeted to be available in early 2027 with full deployment planned by 2027.
  • Together AI says the platform can deliver up to twice-faster inference and lower token costs and expects to serve large open-model workloads, a claim that has not been independently verified.
  • For enterprises, the partnership promises integrated, enterprise-grade inference with Red Hat-style governance and data-residency options that appeal to regulated sectors and sovereign-cloud customers.
  • The deal gives Together AI a clear enterprise sales channel and revenue visibility through 2027 and tests IBM’s strategy to compete on inference infrastructure rather than building proprietary models, with success hinging on real-world production performance.