Digital Realty: Building the Foundation for Enterprise AI

What happens when AI infrastructure becomes as programmable as the applications running on it?

According to Digital Realty, that concept could help enterprises simplify the deployment and operation of private AI environments at scale.

To support this vision, the data centre giant has unveiled ServiceFabric Model Context Protocol (MCP), a new programmable layer designed to help enterprises deploy and operate private AI infrastructure at scale.

The launch extends Digital Realty's ServiceFabric interconnection platform and introduces support for the emerging Model Context Protocol standard, which allows AI systems and agents to interact with infrastructure through standardised interfaces.

Digital Realty positions networking and infrastructure management as programmable services that AI systems can access directly.

Chris Sharp, Chief Technology Officer at Digital Realty, says: “Our strategy is simple: provide the foundational infrastructure enterprises need for sustained AI workloads, while enabling flexible scale as demand grows.

“ServiceFabric MCP extends the foundation of AIPx with programmable controls and agent-ready interfaces, and our patent position reflects the long-term investment we’ve made in this architecture.” 

Building a programmable AI foundation

At the heart of the announcement is AI Private Exchange (AIPx), Digital Realty's private interconnection architecture for AI workloads.

The platform combines orchestration and policy technologies designed to help enterprises manage AI infrastructure across multiple locations and environments.

ServiceFabric MCP sits on top of this architecture, creating what Digital Realty describes as an AI-native control surface.

The platform is designed to operate across more than 800 Digital Realty and third-party data centres, allowing organisations to connect workloads, infrastructure and datasets through a common framework.

The move reflects a growing challenge facing enterprises deploying AI.

As workloads become larger and more complex, organisations must coordinate resources spread across colocation facilities, cloud platforms and on-premises environments while maintaining governance and security controls.

According to industry analysts, that operational complexity is becoming a defining issue for production AI deployments.

“Enterprise adoption of Private AI infrastructure has reached an inflection point,” says Mary Johnston Turner, Research VP at IDC.

“Production AI workloads now demand control over data movement, policy enforcement and partner integration that public cloud APIs alone cannot deliver.

“Providers combining global footprint with programmable, agent-ready interconnection are well positioned to support this next wave of enterprise AI investment.” 

From networking to operations

ServiceFabric MCP introduces four core capabilities intended to simplify AI infrastructure management.

The platform enables intent-based network design and provisioning, allowing connectivity resources to be configured through APIs and MCP interfaces.

It also provides real-time visibility into infrastructure performance, including capacity, topology, throughput and latency metrics.

Security functions include identity management through OAuth 2 and programmable access controls over network connectivity.

Operational features extend into diagnostics and troubleshooting, with integrations available for platforms including Slack, Microsoft Teams, Splunk and Datadog.

Digital Realty says the technology has already been tested across its own operational environments as well as customer deployments.

Validation through real-world AI workloads

One example comes from healthcare AI specialist See All AI, which uses Digital Realty's infrastructure to support medical imaging applications.

T. Michael Thornton, the CEO, says: “At See All AI, we are developing advanced medical imaging AI systems that demand both massive compute performance and highly scalable data infrastructure.

“Digital Realty's Borton campus and ServiceFabric provide the high-bandwidth, low-latency connectivity required to support our NVIDIA DGX B200 environment, enabling secure movement of large imaging datasets, dynamic connectivity to cloud resources and the operational resiliency needed for production healthcare AI.”

Digital Realty has also developed AI infrastructure solutions with partners including Lenovo, Dell and ePlus, supported by technologies from NVIDIA and AMD. Additional ecosystem partners are currently being developed.

Beyond networking

While ServiceFabric MCP initially focuses on programmable networking and connectivity, Digital Realty views it as the first stage of a broader Foundation for AI strategy.

The company expects future development to expand into areas such as power management, facility capacity, inventory visibility, ecosystem integration and sovereign deployment requirements.