Nvidia has formed a three-way partnership with Equinix and Together AI to deliver open-model inference services to enterprises, extending its compute ecosystem from training into deployment. Equinix will provide colocation and network connectivity, Together AI the inference software platform, and Nvidia the GPUs and software stack, packaged as one integrated offering aimed at running open models closer to enterprise data and users. On the same day Equinix introduced Equinix Inference Exchange, an enterprise-grade distributed AI inference service scheduled for official launch in the first quarter of 2027. The design targets latency-sensitive and regulated workloads in sectors such as financial services and healthcare that require local data residency rather than distant hyperscale clouds. Equinix operates more than 280 data centers across 77 major cities, with roughly 230 cloud connection nodes and more than 10,500 enterprise customers, and says eight of the world’s top ten AI model companies and nine of the top ten AI cloud providers already use its infrastructure. The move sits alongside Nvidia’s broader open-model and rack-level push, including a planned $12.9 billion acquisition of Hugging Face, Groq LPU technology licensing for next-generation architecture, and a $3.5 billion MediaTek convertible-bond stake tied to custom XPUs and NVLink Fusion systems, as inference revenue has already overtaken training revenue at Nvidia.