Equinix Partners With Nvidia and Together AI for Inference Exchange
The colocation giant is pivoting toward distributed AI inference to carve a niche separate from massive training clusters.
Equinix has partnered with Nvidia and Together AI to launch the Equinix Inference Exchange, a platform-as-a-service designed for open-model inference. The move positions the data center provider to capture the deployment phase of artificial intelligence by bringing compute closer to the end-user.
The new platform integrates Nvidia's Enterprise Reference Architectures with Together AI's open-source cloud platform, allowing enterprise customers to run AI models within Equinix's distributed infrastructure. Alongside this launch, the company unveiled Equinix Fabric One, a tool intended to simplify connectivity across various clouds, neoclouds, and AI providers. The Equinix Inference Exchange is scheduled to become available in the first quarter of 2027.
The Shift to Distributed Inference
While hyperscalers like Amazon and Google, as well as neoclouds such as CoreWeave, are investing in gigawatt-scale data centers for AI training, Equinix is pursuing a distributed colocation model. Training requires massive, centralized clusters, but inference—the stage where a model is deployed to make real-time decisions—benefits from lower latency. By utilizing smaller facilities in urban centers, Equinix can place compute resources closer to the sensors and data sources that feed AI applications.
Nvidia CEO Jensen Huang highlighted this strategic advantage, noting that the location of Equinix’s facilities allows users to be "close to where the action is, where all the sensors are."
Market Positioning and Financials
This strategy allows Equinix to avoid the high-burn growth and potential "AI bubble" risks associated with single-purpose training clusters. By focusing on interconnection and inference, the company leverages its existing urban real estate to support the high-speed communication required for agentic AI applications. Data center analyst Vlad Galabov noted that while this approach may mean missing some of the immediate boom, it reduces exposure to bubble risks, describing the alternative as a "high risk, high return" scenario.
The company's financial health remains strong as it pivots. Equinix reported second-quarter revenue of $2.63 billion, representing a 16% increase year-over-year, with net income falling between $477 million and $479 million.
Future Outlook
The success of the Inference Exchange will depend on the enterprise adoption of open-model AI and the demand for low-latency edge computing. As AI shifts from the training phase to widespread operational deployment, the industry will be watching whether Equinix's distributed model can outcompete the centralized scale of the hyperscalers. For now, the company is betting that the future of AI is not just in the massive cluster, but in the urban edge.