Equinix, the company that quietly operates 281 data centers across 77 metros worldwide, just announced its most ambitious AI play yet. The Equinix Inference Exchange, unveiled on September 2, combines Nvidia’s enterprise reference architectures with Together AI’s inference platform to create what amounts to a global marketplace for running AI models at the edge of enterprise networks.
The target launch is Q1 2027.
What the inference exchange actually does
Equinix’s pitch is straightforward: use its existing global data center footprint and its interconnection service, Equinix Fabric, to let enterprises run inference workloads close to their data and end users. Together AI brings support for more than 200 open-source models to the table, giving enterprises a broad menu of options without being locked into a single vendor’s proprietary stack. Nvidia contributes its validated Enterprise Reference Architectures, essentially pre-tested hardware and software configurations that reduce the guesswork in setting up inference environments.
The combination means an enterprise in Frankfurt could theoretically spin up an inference workload on open-source models using Nvidia-validated infrastructure, connected via Equinix Fabric to its existing cloud providers, without shipping data halfway around the world first.
