Eminence Grey and Lumen Partner to Connect Sovereign AI Across the U.S.

An AI system’s security boundary does not end at the server rack. When models draw on sensitive information across offices, data centers and operational sites, the connections between those locations become part of the infrastructure that organizations must control.

Eminence Grey and Lumen Technologies are addressing that challenge through a partnership announced October 8, combining Eminence Grey’s sovereign AI platform with Lumen’s edge footprint and programmable network. The collaboration targets government agencies, enterprises and AI developers seeking to connect distributed workloads while retaining control over their data and operating environments.

The companies describe the offering as a foundation for dedicated, private AI infrastructure across the United States. Customers will be able to reserve pre-architected or custom-designed systems, with availability, configuration and implementation determined by their requirements.

Sovereign AI needs a network as well as compute

The partnership addresses three connected questions: where data is processed, who can access it and how it travels between systems. A deployment can use dedicated GPUs yet still depend on connections or management paths that introduce additional exposure.

Eminence Grey contributes AI nodes, GPU and high-performance computing clusters, its Whitehorse command-and-control plane, and Outrider access appliances. Lumen provides the programmable connectivity intended to link those environments, supporting distributed workloads, resilience and operational continuity.

The companies say the combined platform is designed to deliver dedicated infrastructure, verifiable isolation and zero internet exposure across the data lifecycle. Those are stated architectural objectives; the release does not provide deployment-specific security assessments or independent performance benchmarks.

“Sovereign AI isn’t just about where data lives. It’s about how data moves,” said Jim Fowler, Lumen’s chief technology and product officer, in the announcement. His point captures the partnership’s central engineering challenge: preserving control as information moves beyond any single location.

What Whitehorse adds to the infrastructure

Eminence Grey’s website describes Whitehorse as a proprietary, patent-pending control plane coordinating its private AI fabric across locations. It is designed to bring facilities, networking, compute and security into one operating environment, with orchestration, segmentation, policy enforcement and visibility at the workload level.

The company presents its broader business as Infrastructure-as-a-Service: designing, building and operating private AI environments for organizations that require control and resilience. Its engagement process runs from assessment and architecture definition through pilots, expansion and optimization.

That approach extends the offering beyond a purchase of GPU capacity. The operational value would come from aligning placement, access rules, monitoring and continuity across the infrastructure that supports an AI application.

Consider an enterprise agent that retrieves internal documents and triggers a workflow in a business system. Faster inference alone does not determine whether that deployment is suitable. The organization also needs to know which systems the agent can reach, which data boundaries apply and what evidence remains when it takes an action. A control plane is where infrastructure policies can be coordinated across those moving parts.

The announcement identifies Outrider as an access appliance but does not specify its protocols or detailed capabilities. Its precise role and configuration will therefore need to be established as part of a customer’s system design.

Lumen supplies the connectivity layer

Lumen’s Connectivity Fabric provides useful context for the programmable network side of the partnership. The company describes a self-service portal and lifecycle automation for purchasing, deploying and remotely managing consumption-based network services.

Its broader network portfolio combines fiber infrastructure, edge cloud and connectivity services for distributed applications. These existing capabilities help explain Lumen’s role in a collaboration that must connect AI capacity across multiple sites, although the announcement does not specify which individual services each customer configuration will include.

For distributed AI, networking affects how quickly data can reach compute and how reliably applications can continue operating when conditions change. The release says the systems are designed for resilient, low-latency workloads using distributed GPUs and high-performance computing resources. It provides no universal latency target, capacity specification or deployment timetable.

The practical benefit depends on matching network design to the workload. A system processing information near its source has different requirements from a large transfer between computing facilities. The partnership’s custom-design option leaves room for those distinctions.

Security claims require deployment-level evidence

Eminence Grey’s company description also includes private fiber connectivity and post-quantum encryption. The release does not identify specific cryptographic algorithms, certifications or implementation details for the combined offering.

Similarly, keeping infrastructure off the public internet describes an exposure boundary. It does not, by itself, establish how identity, software updates, administration, logging or recovery are handled. Those details are part of what customers will need to evaluate when translating the platform’s design into an operating environment for their own AI workloads.

For organizations protecting proprietary research, regulated records or mission-critical information, the partnership’s significance is the attempt to bring compute and data movement under a coordinated operating model. Eminence Grey and Lumen are joining complementary infrastructure layers at a point when AI deployment increasingly requires both.

The next measure of the collaboration will be the systems customers can actually configure and operate: their isolation boundaries, network behavior, audit evidence and ability to maintain service through disruption. Those specifics will determine how the promise of sovereign AI becomes a workable production environment.