Equinix Horizon 2026: The Recap

Equinix has been in business for 28 years and just held our first ever customer event this week. Equinix Horizon changed that, bringing together industry leaders, customers and partner alongside analysts and influencers, with NVIDIA CEO Jensen Huang joining the keynote by video. A day of keynotes, product news and hallway conversation kept circling back to one question: what it actually takes to run AI at production scale across a real enterprise network.

The Keynote: Connection as the Point

Equinix CEO and President, Adaire Fox-Martin opened Horizon with a direct claim: AI may be distributed, but its value depends on how well it connects. What the highlights of the keynote:

Keep the conversation going:

Catch all of the highlights and keep talk with Equinix experts and other attendees in the Equinix Community: Join the conversation

Two announcements, One Problem

Equinix used Horizon to introduce two products aimed at the same challenge: making distributed AI infrastructure easier to connect and run.

Equinix Fabric One is a managed connectivity service built for one job. Enterprises specify what they need connected, and Fabric One handles the routing, cloud connectivity, MACSec encryption, resiliency and failover automatically. It runs on an open interconnect specification co-developed with AWS and Google Cloud, so applications and AI agents can request connectivity through a portal, an API, an automation pipeline or a natural-language prompt via and MCP-compatible AI tool. Equinix built it in five months. Fabric One enters beta later this year with general availability planned for 2027 in North America.

Read the announcement: Equinix Unveils Equinix Fabric One | Product page: Equinix Fabric One

Watch this interview with Arun Dev, Vice President, Digital Interconnection Services:

Equinix Inference Exchange addresses a related question: where inference actually runs. Build with NVIDIA and Together AI, the program combines NVIDIA’s Enterprise Reference Architectures with Together AI’s inference platform, which supports more than 200 open-source models. Delivered through Equinix Fabric to clouds, networks and AI providers worldwide, it will support both multitenant deployments for shared efficiency and dedicated single-tenant environments for workloads that need capacity. Equinix Inference Exchange launches in Q1 2027.

Read the announcement: Equinix Accelerates AI Inference with NVIDIA and Together AI | Product page: Inference at the Metro Edge

Watch Bruce Owen, EVP Global Markets discuss Equinix Inference Exchange:

The Edge

Equinix’s neutral, colocation-first approach dates back to the company’s founding in 1998, long before hyperscalers or neoclouds existed, and it’s now applied across more than 10,500 interconnected business on its exchange. That history shapes where Fabric One and Equinix Inference Exchange run: Equinix’s metro footprint puts inference within 10 milliseconds of the majority of internet users worldwide, across 77 metros in 36 countries, proximity that directly affects inference latency and cost.

The Choice Message, on Camera

Adaire Fox-Martin took the model-choice pitch on camera after Equinix Horizon: the platform is build so companies can put one AI model up against another and switch based on performance and cost.

Customers on Stage

Hyundai AutoEver America CEO Paul Hager pointed to why the moment matters: as AI workloads and multicloud architectures spread across more environments, connectivity has to scale without adding operational complexity. Alembic and Texas A&M Engineering Experiment Station, two customers featured at Horizon, shared how private AI infrastructure is already changing what’s possible in their work: compute time measured in minutes instead of days, and AI-driven research on a more secure foundation.