Mavenir and Red Hat launch sovereign AI platform for telecom operators

Telecoms have quietly been some of the most strategically important infrastructure owners in AI—they sit on the pipes, but until now most of the value from generative models has accrued to cloud hyperscalers. Mavenir’s platform, built on Red Hat OpenShift, is an explicit attempt to flip that script by turning carriers into AI service providers in their own right. By defaulting to on-premise, operator‑controlled models and only routing the toughest workloads to external frontier systems, it addresses both sovereignty and unit economics in one architecture.

For the broader AI race, this is one of the clearest examples of ‘agentic infrastructure’ moving out of the lab into a regulated, high-scale industry. If operators can reliably meter token usage, enforce hard spend caps, and expose standard APIs, they become natural distribution channels for AI assistants, agents and vertical models—especially in markets where trust in US-based cloud vendors is fragile. It also creates a sizable new route-to-market for open-weight and regional models that can be hosted inside carrier data centers.

The competitive implication is that hyperscalers may increasingly be forced to share economics with carriers or risk being disintermediated at the edge. Over time, we could see national or regional telcos emerge as important counterweights to Big Tech in AI infrastructure, particularly outside North America.