Blab AI
Telecommunications Providers Shift from AI-Driven Opex Savings to Monetizing AI Services
While most communications service providers (CSPs) have leaned on AI to trim operating costs, a new trend is emerging: turning their networks into revenue‑generating AI platforms.
The shift is already visible in the work of several major carriers. China Mobile, for instance, has reached Level 4 on the TM Forum Autonomous Networks maturity index in its network operations centres, deploying agentic and generative AI to automate maintenance tasks. A TM Forum case study shows the initiative cut operations‑and‑maintenance staff requirements by 30 % and reduced mean time to repair for faults and customer complaints by the same margin. Rakuten Mobile has achieved Level 4 autonomy for radio‑access‑network (RAN) energy efficiency—a first for a live Open RAN deployment—and expects a 20 % improvement in energy use. Swisscom has validated Level 4 processes in IP transport, delivering cost savings and faster time to market for new services.
These examples illustrate the current focus on operational efficiency. CSPs are using AI to prepare data, identify use cases, integrate systems, validate models and realign organisations. According to the TM Forum’s six‑level maturity framework, most carriers self‑report at Level 1 or 2, but a handful have demonstrated Level 4 “highly autonomous” processes.
According to Cisco Live presentations, the industry is now looking beyond internal savings. Cisco Chair and CEO Chuck Robbins said CSPs are beginning to see a path toward monetising AI. He noted that the bandwidth, distributed infrastructure, security and proximity to users that CSPs already provide are the same assets required for AI services. Robbins highlighted the potential to repurpose legacy central offices and mini‑data centres as AI compute nodes.
AT&T’s Andy Forester, general manager of strategic managed services, framed the current moment as a break from earlier edge‑compute cycles. He explained that enterprises are now deciding where AI workloads should reside—on‑premises, at the edge, in specialised AI data centres or in hybrid architectures. Forester cautioned that the market is still forming and that the optimal architecture depends on use case, data, latency, security and business outcome.
The rise of agentic AI adds complexity. Software agents that act on behalf of users or organisations will generate traffic patterns that are more dynamic, distributed and symmetrical. This makes the network a control point for performance, security, policy and cost, not just a transport layer.
While the present focus of AI‑enabled network automation is to reduce operating expenses and enforce discipline, the industry consensus is that the long‑term opportunity lies in monetisation. CSPs could offer sovereign AI services, GPU‑as‑a‑service, or edge‑based combinations of connectivity and compute capacity.
The industry is still early in this transition. Most carriers are still at the sub‑domain or domain level of automation, and the journey to end‑to‑end, cross‑domain automation remains a work in progress. Nevertheless, the combination of proven operational gains, a clear path to monetisation and the growing demand for AI workloads suggests that CSPs are positioned to become key players in the AI value chain.
In the coming months, observers will watch how CSPs expand their AI service portfolios, whether they partner with hyperscalers or develop proprietary offerings, and how they navigate the regulatory and security challenges that accompany the shift from internal optimisation to external service provision.
The current situation is that CSPs have demonstrated tangible operational benefits through Level 4 autonomous processes, are beginning to articulate a monetisation strategy, and are preparing to leverage their distributed infrastructure to support AI workloads. The next steps will involve scaling these capabilities, building market demand for AI services and addressing the technical and policy implications of turning telecom networks into AI infrastructure.
The shift is already visible in the work of several major carriers. China Mobile, for instance, has reached Level 4 on the TM Forum Autonomous Networks maturity index in its network operations centres, deploying agentic and generative AI to automate maintenance tasks. A TM Forum case study shows the initiative cut operations‑and‑maintenance staff requirements by 30 % and reduced mean time to repair for faults and customer complaints by the same margin. Rakuten Mobile has achieved Level 4 autonomy for radio‑access‑network (RAN) energy efficiency—a first for a live Open RAN deployment—and expects a 20 % improvement in energy use. Swisscom has validated Level 4 processes in IP transport, delivering cost savings and faster time to market for new services.
These examples illustrate the current focus on operational efficiency. CSPs are using AI to prepare data, identify use cases, integrate systems, validate models and realign organisations. According to the TM Forum’s six‑level maturity framework, most carriers self‑report at Level 1 or 2, but a handful have demonstrated Level 4 “highly autonomous” processes.
According to Cisco Live presentations, the industry is now looking beyond internal savings. Cisco Chair and CEO Chuck Robbins said CSPs are beginning to see a path toward monetising AI. He noted that the bandwidth, distributed infrastructure, security and proximity to users that CSPs already provide are the same assets required for AI services. Robbins highlighted the potential to repurpose legacy central offices and mini‑data centres as AI compute nodes.
AT&T’s Andy Forester, general manager of strategic managed services, framed the current moment as a break from earlier edge‑compute cycles. He explained that enterprises are now deciding where AI workloads should reside—on‑premises, at the edge, in specialised AI data centres or in hybrid architectures. Forester cautioned that the market is still forming and that the optimal architecture depends on use case, data, latency, security and business outcome.
The rise of agentic AI adds complexity. Software agents that act on behalf of users or organisations will generate traffic patterns that are more dynamic, distributed and symmetrical. This makes the network a control point for performance, security, policy and cost, not just a transport layer.
While the present focus of AI‑enabled network automation is to reduce operating expenses and enforce discipline, the industry consensus is that the long‑term opportunity lies in monetisation. CSPs could offer sovereign AI services, GPU‑as‑a‑service, or edge‑based combinations of connectivity and compute capacity.
The industry is still early in this transition. Most carriers are still at the sub‑domain or domain level of automation, and the journey to end‑to‑end, cross‑domain automation remains a work in progress. Nevertheless, the combination of proven operational gains, a clear path to monetisation and the growing demand for AI workloads suggests that CSPs are positioned to become key players in the AI value chain.
In the coming months, observers will watch how CSPs expand their AI service portfolios, whether they partner with hyperscalers or develop proprietary offerings, and how they navigate the regulatory and security challenges that accompany the shift from internal optimisation to external service provision.
The current situation is that CSPs have demonstrated tangible operational benefits through Level 4 autonomous processes, are beginning to articulate a monetisation strategy, and are preparing to leverage their distributed infrastructure to support AI workloads. The next steps will involve scaling these capabilities, building market demand for AI services and addressing the technical and policy implications of turning telecom networks into AI infrastructure.
