Cisco Sees Rising Role for Service Providers in AI's Inferencing Era

Why It Matters

The move positions service providers to capture a growing slice of AI spend, diversifying revenue beyond traditional bandwidth sales and accelerating edge compute adoption across the telecom ecosystem.

Key Takeaways

• Cisco predicts AI inference will be 25% of traffic by 2035
• Service providers can launch “inferencing clouds” to monetize edge AI workloads
• Comcast, Charter, Bell Canada already pursuing AI edge initiatives
• Megaport raises $594M to build inference cloud across 20 US cities
• AT&T, Cisco, Nvidia partnership brings on‑demand AI inference to the edge

Pulse Analysis

The AI landscape is entering an inferencing era where real‑time decision‑making and autonomous agents demand low‑latency, power‑efficient compute. Cisco’s recent study highlights that by 2035, AI inference could represent a quarter of all network traffic, a stark contrast to today’s data‑center‑centric training workloads. This shift forces network architects to rethink capacity planning, emphasizing edge proximity and traffic‑aware routing. Cisco’s deployment of sensors on provider routers aims to quantify these flows, promising semi‑annual reports that will guide infrastructure investment.

For telecoms and cable operators, the emerging demand translates into a tangible revenue opportunity. By repurposing existing fiber and edge sites into “inferencing clouds,” service providers can offer AI‑ready compute alongside connectivity, competing with GPU‑heavy neoclouds. Early adopters such as Comcast, Charter and Bell Canada are already embedding AI capabilities at the edge, while AT&T’s collaboration with Cisco and Nvidia showcases a model for on‑demand inference services. This evolution aligns with the broader 5G vision of mobile‑edge compute, enabling latency‑critical applications like autonomous vehicles and real‑time analytics to run closer to the data source.

Capital is flowing to support the transition. Australian‑based Megaport’s $594 million funding round and $329 million in AI infrastructure contracts underscore market confidence in distributed inference. Its strategy of deploying servers in 20 U.S. cities mirrors the telecoms’ push to localize AI workloads. As AI spend migrates from hyperscalers to edge locations, service providers that can bundle connectivity, compute, and AI‑specific SLAs will likely capture a significant share of the emerging AI inferencing market, reshaping revenue models and competitive dynamics across the industry.