Particle.news

Equinix Unveils Inference Exchange With Nvidia and Together AI

The service combines Nvidia-validated reference architectures with Together AI’s open-model inference platform, leveraging Equinix’s metro-edge footprint to enable low-latency sovereign inference.

Overview

  • Equinix announced the Equinix Inference Exchange at its Equinix Horizon event on Wednesday, Sept. 2, 2026, and said the service will be available in the first quarter of 2027.
  • The three-layer program pairs Equinix’s global colocation and Equinix Fabric connectivity with Nvidia’s Enterprise Reference Architectures and Together AI’s inference platform that offers access to more than 200 open-source models.
  • Together AI will act as the seller of record and bill end customers who use the Equinix Inference Exchange, while Equinix supplies data-center operations and interconnection across roughly 280–281 sites in 77 metros.
  • The announcement did not disclose financial terms or specific hardware purchase commitments, and the market reaction was modest with Equinix shares rising about 2% on the day.
  • Industry observers say the move highlights demand for metro-edge and sovereign inference—running AI close to data and users—and gives Equinix a way to monetize its neutral exchange footprint as inference grows as a share of AI compute.