The partnership and the launch target
Equinix announced on September 2, 2026 that it is building the Equinix Inference Exchange, a distributed AI inference program for global enterprises that combines NVIDIA Enterprise Reference Architectures with Together AI's inference platform. The service will run on Equinix's global data center footprint and connect to clouds, networks and AI providers through Equinix Fabric, with a targeted launch in Q1 2027. Together AI's platform supports more than 200 open-source models, a scale Equinix says is central to offering enterprises flexibility rather than a single-vendor stack.
What the program is designed to do
The exchange is framed as a way to bring inference closer to where data, users and applications already reside, reducing latency and the cost of shuttling workloads across regions. Equinix CEO and President Adaire Fox-Martin positioned the move as part of nearly three decades of operating what she called "the trusted exchange where the world's enterprises run, connect and orchestrate their most critical workloads." The announcement was made at Equinix Horizon, the company's inaugural customer and partner event, alongside the introduction of Equinix Fabric One, which is intended to simplify connections across globally distributed AI environments.
How NVIDIA and Together AI fit in
NVIDIA contributes validated Enterprise Reference Architectures — pre-tested hardware and software configurations intended to shorten the path from AI experimentation to production deployment. NVIDIA vice president of global AI clouds and infrastructure ecosystem Raj Mirpuri said the combination turns Equinix's digital interconnection platform into "a global fabric for AI inference" and treats accelerated compute as a strategic asset class. Together AI supplies the inference layer, giving enterprises a broad menu of more than 200 open-source models and an open-ecosystem posture that Equinix highlights as a counterweight to proprietary model providers.
Why the footprint matters
Equinix operates 281 data centers across 77 metros, a scale the company is leaning on to argue that inference can be deployed close to enterprise users rather than concentrated in a small number of hyperscale regions. The exchange is designed to connect through Equinix Fabric to multiple clouds and networks simultaneously, supporting hybrid and multi-cloud architectures. Equinix framed the program as part of a broader pattern of layering AI-specific capabilities on top of its interconnection assets, following earlier initiatives such as the Equinix AI Factory program and collaborations with companies including Cisco.
Other developments
Separately, NVIDIA and Equinix drew additional coverage as analysts weighed what the deeper inference partnership could mean for both companies' stocks, though no specific market reaction figures were reported in the available sources.
Share this article
