← Back

Wire · opportunities

Analysis: AT&T bets on routing, open model to tame AI costs

Published

28 July 2026

Topic

opportunities

Sectors

Telecom & Connectivity

Geography

United States

Source

Read at mobileworldlive.com

Verified

Fusion42 · 28 July 2026 · Fusion42 review

AT&T has deployed a proprietary AI gateway using cache-aware routing to cut inference costs by up to 90%, and released OTel 2.0, an open-source telecom-specific model trained on 400+ billion tokens, to address accuracy demands at scale across its 45 billion daily token processing.

This Wire brief sits within Fusion42's coverage of Telecom & Connectivity.

◆ The Wire takeaway

AT&T just proved that cheap, task-matched inference beats frontier models on real infrastructure costs. If you sell AI infrastructure, routing layers, or domain models to telcos or large enterprises running high-token-volume workloads, the 90% cost cut changes your customer's buying decision: they now need you to plug into their gateway, not replace it.

Coverage

1 source · 28 Jul 2026

Related on Wire

Topics

Telecom & Connectivityai-inference-costsmodel-routingdomain-adapted-modelstelco-aiopen-source-models