Equinix, NVIDIA and Together AI to launch Inference Exchange inside metro data centers in Q1 2027

The partnership and platform

Equinix has joined with NVIDIA and Together AI to launch Inference Exchange, a platform that places AI model inference capabilities inside Equinix's metropolitan data centers. The service is scheduled to begin rolling out in the first quarter of 2027 and was announced at Equinix's inaugural Horizon conference. The offering is positioned to separate inference workloads from centralized AI training, allowing enterprises to run model serving close to their data sources and end users.

Fabric One underpins the exchange

Inference Exchange is built on Equinix's Fabric One intent-based networking layer, which is intended to simplify deployment and provide low-latency, compliant connectivity between data, users and inference compute. Equinix has described Fabric One as the networking foundation that lets the exchange operate across multiple metro sites while meeting enterprise data-handling requirements.

Metro footprint and enterprise focus

By locating inference inside Equinix's existing metro data center footprint rather than at a limited number of hyperscale campuses, the partners are targeting latency-sensitive enterprise use cases in major metropolitan areas. The approach is framed as an alternative to routing every inference call through distant centralized clouds, with the goal of reducing round-trip times and keeping regulated data within established Equinix facilities.

What's next

The Q1 2027 rollout timeline sets the next verifiable milestone as the first operational deployments of Inference Exchange inside Equinix metro sites using Fabric One. Further detail on participating metros, pricing and early customer commitments is expected as the launch window approaches.

Keywords: Equinix, NVIDIA, Together AI, Inference Exchange, Fabric One, metro data centers