In one of the industry’s early efforts to offer structure, shared language, and practical context to how artificial intelligence and sustainability intersect in next-generation networks, ATIS’ Next G Alliance published Sustainable AI in Telecom: Promises and Challenges in 6G. Upon publication in 2025, AI adoption in telecom was accelerating; but many sustainability discussions remained high-level, often focused on individual technologies rather than system-wide impacts.

One year later, the pace of change has only increased. Generative AI, continuous inference workloads, and deeper automation across the Radio Access Network (RAN), core network, and data center domains have moved sustainability considerations from long-term planning discussions into immediate operational realities. Against this backdrop, the report has proven to be more than timely. It has served as a foundational reference point for how the industry approaches sustainable AI in telecom.

Balancing AI for sustainability and Sustainable AI

One of the report’s most significant contributions was its ability to establish a shared foundation for discussing AI and sustainability in telecom. By clearly distinguishing between AI for sustainability, or using AI to improve network efficiency and reduce environmental impact, and sustainable AI, ensuring that AI itself is designed and operated responsibly, the report helped ground the conversation in practical realities rather than abstract goals.

This distinction has become increasingly important as AI workloads scale. Efficiency gains enabled by AI can no longer be evaluated in isolation; they must be considered alongside the energy, water, and material costs associated with training, inference, and supporting infrastructure. By articulating this dual perspective early, the report provided a common reference that continues to underpin how organizations assess AI-enabled sustainability initiatives.

Equally important was the report’s emphasis on taking an end-to-end view across next generation mobile networks, from development and deployment through operation and retirement. This holistic perspective anticipated a reality now facing many network operators and technology providers: sustainability commitments increasingly require evidence across the full system lifecycle, not just improvements in runtime efficiency or isolated performance metrics.

From early insights to operational reality

Several of the concepts outlined in Sustainable AI in Telecom have moved rapidly from early insights to day-to-day operational considerations.

A clear example is the growing dominance of AI inference. The report highlighted that inference accounts for the majority of AI’s operational environmental footprint, particularly as AI models are deployed at scale across live networks. Over the past year, this observation has been reinforced by the widespread deployment of AI-enabled services that rely on continuous, real-time inference, driving sustained demand across communication and data center infrastructure.

This shift has sharpened the industry’s focus on model efficiency, right-sized architectures, and intelligent placement of AI functions across edge, core, and cloud environments. Rather than treating AI as a centralized capability, operators are increasingly evaluating where and how AI workloads should run to balance performance, energy consumption, and operational constraints.

Another area of acceleration is energy-aware and location-aware operation. This report discussed the potential benefits of aligning AI workloads with time-of-day usage patterns, regional energy characteristics, and infrastructure capabilities. Today, these considerations are becoming more tangible as networks and data centers encounter real limits on power availability, cooling capacity, and grid constraints. What were once forward-looking recommendations are increasingly relevant to near-term operational planning.

Broadening the sustainability lens

While energy efficiency remains central, the report’s attention to water usage, electronic waste, and embodied emissions has become even more relevant over the past year. Cooling-related water consumption and the environmental impact of specialized AI hardware are now more visible considerations in infrastructure planning and public discourse.

In this context, the report’s inclusion of circular economy principles, such as extending equipment lifecycles, enabling reuse, and managing retirement responsibly, provides an important foundation for addressing sustainability beyond operational energy use. AI itself can support these objectives through predictive maintenance, asset optimization, and more informed decision-making across the network lifecycle.

The guidance provided also reinforced a principle that remains especially relevant for telecom networks: sustainability cannot be considered in isolation from performance. Telecom networks exist to deliver reliable connectivity and quality of service, and sustainable outcomes must be achieved alongside these core objectives. The use of KPIs and KVIs to support informed trade-offs remains a practical approach for aligning environmental goals with operational requirements.

What to watch in the next 12–24 months

Building on the foundation established by Sustainable AI in Telecom, the potential from these advancements is becoming clearer. Several developments expected to shape how sustainable AI in telecom evolves over the next one to two years:

• Carbon- and grid-aware networking: AI-driven network functions that account not only for energy consumption but also for carbon intensity and grid conditions are likely to mature, extending energy awareness into sustainability-aware decision-making.
• Efficiency-first AI models: Purpose-built, smaller models optimized for specific network tasks are expected to gain prominence, particularly for edge and RAN applications where energy efficiency and latency are critical.
• Sustainability as an explicit network objective: Intent-based networking approaches are likely to evolve to include sustainability goals alongside traditional performance metrics, reinforcing the need for transparency and explainability.
• Expanded sustainability metrics: In addition to established energy efficiency measures, metrics related to water usage, embodied emissions, and lifecycle impacts are expected to play a larger role in network planning and reporting.

Building on the shared principles and practical guidance developed by the Next G Alliance

In writing this report, the Next G Alliance did not aim to anticipate every technological shift. Instead, it established a clear, shared understanding of how AI and sustainability intersect across telecom systems. As AI becomes more deeply embedded in network operations and sustainability constraints continue to tighten, that foundational understanding remains a valuable guide.

Looking ahead, the challenge is not whether sustainable AI matters, but how effectively its principles are translated into scalable, measurable, and operational practices. The work outlined in Sustainable AI in Telecom provides a strong basis for that next step, supporting the industry as sustainable AI moves from concept to operational priority.