The inference layer is where enterprise AI strategy meets operational cost. Running open models at production scale requires physical infrastructure, specialized compute, and serving software that most enterprises do not manage internally. Nvidia announced a data center deal on Wednesday with Equinix and Together AI to address that gap, giving enterprise customers a path to open-model inference.
Equinix has found its niche in the AI data center boom by occupying a layer distinct from what hyperscalers build for their own cloud platforms. The Wednesday deal draws Together AI's inference capabilities into that position. Nvidia anchors the hardware layer.
For enterprise buyers, the case for open-model inference is about operational control. Running models on infrastructure the customer manages is different from consuming a closed API at terms set by the vendor. That is the specific gap this arrangement targets, with Equinix's data center footprint as the physical layer underneath it.