Enterprise networking has become steadily more programmable while remaining dependent on specialist design and operations, and Equinix is now trying to move another layer of that complexity behind a managed service that accepts the outcome a customer wants rather than requiring every connection to be assembled separately.

Equinix has unveiled Fabric One, an any-to-any connectivity service intended to connect corporate infrastructure, cloud providers, networks, and AI environments while automatically handling routing, encryption, resilience, and failover. The product was announced at the company’s Horizon event alongside its new distributed AI inference offering.

Instead of configuring each network component individually, customers will be able to describe what needs to connect and specify the required outcome, after which Fabric One is intended to determine and maintain the underlying connectivity. Requests can come through a portal, APIs, automation workflows, software agents, or natural-language prompts.

The service is being built around the open-source OpenAPI 3.0 Interconnect specification, developed with collaboration from AWS and Google Cloud, which are serving as lead integration partners. Equinix says the approach is intended to preserve interoperability as organisations spread applications and AI workloads across more providers and infrastructure environments.

There is an important limitation for European customers, however. Fabric One is not yet commercially available, with beta planned later in 2026 and general availability in 2027 beginning in North America. Equinix has not announced a European production date, so the service is better treated as an indication of where enterprise networking is heading than as an immediately deployable European product.

Network automation moves up a layer

The problem Fabric One is trying to address is familiar to organisations operating multicloud estates. A new application may need private access to several cloud platforms, corporate infrastructure, software suppliers, security services, and AI providers, with each dependency carrying its own routing, encryption, resilience, and operating requirements.

Software-defined networking has already made many of those connections programmable, but programmability does not remove architectural work. Engineers still need to decide which services to use, how traffic should move, where redundancy is required, and what should happen when a provider or route becomes unavailable.

Fabric One attempts to shift the interface upwards by asking customers to express an intended outcome and allowing the service to translate that intent into lower-level network configuration. The model resembles changes elsewhere in infrastructure where cloud services progressively replaced manual component configuration with policy-driven requests.

That abstraction can remove repetitive operational work, although it also moves technical control towards the provider. A network team may spend less time building individual connections but become more dependent on whether the managed service accurately interprets requirements, exposes enough information for troubleshooting, and behaves predictably when the underlying infrastructure changes.

The open interconnect specification therefore carries more weight than a conventional product feature. If intent-based networking becomes another proprietary control plane tied closely to one provider, customers can simplify operations while making later migration harder. A shared specification offers a route towards greater portability, although practical interoperability will only become apparent once several implementations are operating in production.

AI agents become network consumers

Equinix is also designing Fabric One so applications and AI agents can discover, request, and provision connectivity programmatically. That creates a more dynamic operating model than the familiar process in which a network change starts with a human ticket and moves through design, approval, configuration, and testing.

An agent able to identify a service and request the network path needed to reach it could shorten deployment cycles, particularly where AI workloads move between infrastructure providers or capacity locations. The same capability creates a governance problem because network connectivity is itself a security boundary, and automatically creating new routes can expose data or services that were previously separated.

Enterprises will consequently need policies governing which automated systems are allowed to request connectivity and which changes still require approval. Intent-based provisioning only removes manual configuration safely when identity, authorisation, logging, segmentation, and policy controls are strong enough to prevent a software error from becoming a network exposure.

Equinix enters that market with an interconnection estate spanning more than 10,500 businesses and approximately 3,000 cloud and IT providers. The company also says eight of the ten largest AI model providers and nine of the ten largest neoclouds already have deployments within its ecosystem, giving Fabric One a substantial pool of destinations customers may want to reach.

The strategy fits a wider change in infrastructure management as businesses consume more external services. The work involved in connecting each provider becomes harder to justify as a collection of one-off engineering projects, particularly when application teams increasingly expect infrastructure to be available through software interfaces rather than long provisioning cycles.

Network abstraction tends to reveal its limits during failure, however. When latency rises, a connection disappears, or traffic follows an unexpected path, engineers still need enough visibility into the underlying system to understand what the managed layer has done. Fabric One will therefore have to reduce configuration work without making the resulting network opaque to the people responsible for operating it.

With European availability still unspecified, the immediate announcement is more architectural than commercial for Techopia’s core market. Even so, it captures a direction already visible across cloud and AI infrastructure: organisations increasingly want to state what their infrastructure should provide and leave software to assemble the components beneath it, while retaining enough control to prevent operational convenience from becoming another source of dependency.