Share with your CTO
Ericsson is positioning India as the world’s primary proving ground for agentic AI deployment, with global CTO Erik Ekkuden arguing that India’s 5G infrastructure, startup density, and digital public stack (Aadhaar, UPI) create conditions no other market can replicate at scale. The company already exports 15,000 5G antennas and 9 million components annually from India to Europe. Its Aduna network API platform now includes three Indian telcos, with fraud prevention as the first monetizable use case.
What this means for your business
The argument Ekkuden is making isn’t really about India’s AI ambitions. It’s about the network layer becoming a billable surface for AI inference. Fraud detection running at the network edge, quality-on-demand for agentic workflows, real-time language translation baked into the pipe itself: these are services where Ericsson sits between the telco and the enterprise application, collecting toll. CTOs at companies with significant India operations, or evaluating API-based fraud prevention vendors, are deciding whether that infrastructure layer becomes a strategic dependency before they’ve fully mapped it.
The “India as proving ground” framing deserves scrutiny. Ericsson, speaking as a vendor whose revenue depends on telcos deploying its network APIs commercially, has an obvious interest in projecting an optimistic adoption timeline, which nudges the stated “it is too early to say how much is scalable” caveat to the margins. The concrete numbers (9 million components, 15,000 antennas exported) describe manufacturing traction, not AI deployment at scale. The agentic AI leadership claim rests on university research and startup momentum, neither of which has yet produced a named enterprise deployment. Impressive potential, thin proof.
What’s worth taking seriously is the architectural direction, not the timeline. If 6G networks natively run AI inference at the edge (meaning computation happens close to the data source rather than in a distant data center), the current assumption that enterprise AI workloads flow through centralized cloud providers gets materially complicated. A CTO renewing a hyperscaler contract in the next 18 months should be asking whether edge inference changes the make-vs-buy calculus on latency-sensitive agentic applications. The answer probably isn’t yes yet. But the renewal cycle that matters is the one after this one.
Concept deep-dive: Network APIs
A network API lets an external application query live telco data in real time, the way a weather app queries a weather service, except the data is signal quality, device location, or transaction anomaly flags. This matters for enterprise AI because it turns the carrier’s infrastructure into a programmable input layer. Instead of AI models working only on historical records, they can act on what the network knows about a call or transaction right now, which is the difference between flagging fraud after the fact and stopping it mid-handshake.
Based on reporting from India to lead world in adopting agentic AI: Ericsson CTO Erik Ekkuden | People, originally published 2026-10-11 14:08:00.
