Hello,

This week, Ericsson shipped agentic AI to the cloud, NVIDIA confirmed telecom AI spending hit escape velocity, and MTN went live with autonomous antennas in Ghana. Three continents. One direction.

In today's edition:

Let’s dive in.

Ericsson launched its Agentic rApp as a Service (rApp aaS) on AWS Marketplace on February 16. The system uses agentic AI for reasoning and coordination of RAN optimisation workflows. It introduces a natural language interface powered by generative AI, allowing operators to talk to their networks in plain language, translating speech into automated instructions.

Vivo Brazil, with 103 million connections, is the first operator to field-test the service. The rApp aaS connects to the non-RT RIC through the O-RAN Alliance's R1 interface - the open standard that links rApps to live network systems. Ericsson's existing AI optimisation portfolio already handles more than 100 million inferences daily across 11 million cells serving over 2 billion subscribers.

Three days later, Ericsson announced a partnership with Mistral AI on February 19. Initial use cases target automated legacy code translation, AI-assisted development for 6G research, and custom AI agents for internal workflows. The vendor also unveiled AI-ready radios with neural network accelerators - programmable matrix cores inside beamforming chips that run AI workloads at the cell site. Both launches headline Ericsson's MWC 2026 presence in Barcelona, March 2–5.

The timing is deliberate. Ericsson is assembling a full-stack AI proposition - from silicon at the radio edge to cloud-hosted agentic applications - ahead of Barcelona. Omdia analyst James Crawshaw noted that no other vendor offers a comparable rApp-as-a-Service solution, giving Ericsson a genuine differentiator. By building on the O-RAN R1 interface, the platform remains vendor-agnostic, which matters as operators resist lock-in. The Mistral partnership signals that telecom-specific foundation models are now a vendor priority - not generic AI applied to networks, but models shaped by radio and infrastructure data.

What this means for you: If you run a multi-vendor RAN, the rApp aaS model changes your procurement calculus. Test whether cloud-hosted agentic rApps reduce your optimisation cycles before committing to on-premise alternatives. MWC will be the proving ground.

NVIDIA's fourth annual State of AI in Telecommunications survey, released February 18, confirmed the spending shift. From roughly 1,000 respondents: 89% plan to increase AI budgets this year, up from 65% a year ago. Ninety percent reported that AI is already driving revenue growth while cutting costs. Network automation has overtaken customer service as the top investment area, with 54% now using AI for network planning and optimisation - a 17 percentage-point rise year on year. Seventy-seven percent expect AI-native networks to launch before 6G arrives.

The spending is translating into products. Amdocs launched aOS on February 3, an agentic operating system purpose-built for telecoms. Sitting on top of any BSS/OSS stack, aOS orchestrates AI agents across end-to-end workflows for over 350 operators globally. It targets the gap between isolated AI experiments and full-process automation - what Amdocs calls the shift from "digital to cognitive" operations.

Deutsche Telekom is pursuing a different path. Its multi-year partnership with OpenAI entered pilot phase in Q1 2026, deploying ChatGPT Enterprise across operations serving 261 million mobile customers. The German operator is simultaneously building an AI Gigafactory with NVIDIA, with construction starting this year and an Industrial AI Cloud already live as Europe's first sovereign enterprise AI platform.

The NVIDIA data confirms a structural change, not a spending cycle. When 9 in 10 operators report measurable revenue gains, scepticism becomes expensive. The shift from customer service AI to network automation AI reflects maturity - operators now trust AI with infrastructure decisions. Amdocs and Deutsche Telekom represent two approaches: a horizontal platform play versus a deep vertical partnership with a frontier AI lab. Both converge on the same target: autonomous, self-healing networks.

What this means for you: Benchmark your AI spending against the 89% figure. If your network automation budget has not grown year on year, you are falling behind the peer group. Evaluate horizontal platforms like aOS alongside vertical partnerships before budgets lock in for H2.

MTN Ghana and Huawei completed the world's first large-scale deployment of the Alpha Antenna on February 14. Post-deployment tests recorded a 6.8% rise in regional traffic and a 30x improvement in operations and maintenance efficiency. The integrated Antenna Information Sensor Unit enables fully automated, real-time retrieval of antenna parameters, eliminating manual site visits entirely. Remote multi-dimensional beam adjustment allows the network to adapt coverage dynamically. Field reporting indicates maintenance cost reductions of 97%.

Ghana does not rank among Africa's top ten mobile internet markets by speed. That makes this deployment a leading indicator, not an outlier. AI-RAN adoption across the continent is building: Vodacom is working with NVIDIA and Nokia on AI-enabled network management platforms, while MTN has run Open RAN proof-of-concept trials with Rakuten Symphony across South Africa, Nigeria, and Liberia. The Alpha Antenna validates that AI-embedded radio hardware works at commercial scale in markets where cost efficiency is existential.

For emerging market operators, the 30x maintenance efficiency gain matters more than the traffic uplift. Fewer truck rolls, fewer manual audits, and fewer configuration errors directly reduce opex in regions where skilled field engineers are scarce and travel distances are long. Huawei gains a live reference deployment for its autonomous network portfolio targeting Africa's 1.2 billion mobile subscribers. MTN Group, serving over 300 million customers across 16 countries, now has a template to replicate.

What this means for you: Operators in cost-sensitive markets should request Alpha Antenna performance data from Huawei. The 97% maintenance cost reduction, if replicated across MTN's West African footprint, reshapes the total cost of ownership model for RAN upgrades across emerging economies.

Telefónica and Mavenir signed a memorandum of understanding on February 19 to create a joint AI Innovation Hub. The facility serves as a production-grade testbed for AI-driven autonomous network orchestration, intent-based services, and monetisation frameworks. It emulates live traffic patterns so both companies can validate next-generation solutions before commercial rollout. Live demonstrations are scheduled for MWC 2026 in Barcelona, March 2–5.

This builds on Telefónica's Autonomous Network Journey programme, launched in 2021. The operator finished 2025 with 12 Level 4 use cases - a TM Forum classification reflecting highly advanced automation with minimal human intervention - operational across Telefónica Spain, O2 Germany, and Vivo Brazil. The programme exceeded initial targets. Separately, Vivo Brazil is now field-testing Ericsson's Agentic rApp aaS, connecting Mavenir core network work with RAN-layer automation within a single operator footprint.

Telefónica is building autonomous capability across every network layer simultaneously - core with Mavenir, RAN with Ericsson, enterprise AI via OpenAI tools. The multi-vendor approach de-risks the transition while maintaining competitive pressure on each supplier. For the broader industry, Telefónica's cross-layer approach may become the model for how tier-1 operators adopt autonomous networking without single-vendor dependency.​

What this means for you: Track Telefónica's Level 4 use case count as a benchmark. If your operator has fewer than five, the gap is widening. MWC demonstrations from the Mavenir hub will reveal which core network functions are closest to autonomous operation.