As telecom networks evolve toward autonomous 5G‑Advanced and 6G operations, agentic AI workflows—large language models (LLMs) performing multi‑step reasoning, invoking diagnostic tools, retrieving domain knowledge, and coordinating across agent teams—are increasingly distributed across the edge‑cloud continuum. The biological brain accomplishes complex cognition on roughly 20 W, whereas contemporary LLMs are highly energy‑ and memory‑intensive, making sustainable lifecycle orchestration a critical operational priority. Existing AI lifecycle metrics assess only isolated single‑model inference or ignore multi‑agent execution graphs, leaving network operators without a foundation to gauge whether distributed agent communication incurs significant energy costs and where agent teams should reside. To bridge this gap, we introduce agentic‑eCAL, generalizing the Energy Cost of AI Lifecycle (eCAL) metric to directed multi‑agent workflows. The metric couples a closed‑form two‑rate single‑call energy model—compute‑bound prefill and memory‑bound decode—with the 7‑layer OSI data‑transport stack. Grounded in hundreds of GPU benchmark configurations on NVIDIA A100 and H100, 16 open‑weight models, and 8 orchestration topologies, we validate metric components and study placement implications. Results show that inter‑agent text transport across 5G RAN, metro, and optical links accounts for only 0.25% of total workflow energy. Consequently, in edge‑cloud agent placement the dominant energy cost is not the transmission of text itself but the additional inference and context processing induced by that communication.
Review: agentic‑eCAL offers a quantifiable energy framework for multi‑agent systems, revealing that in edge‑cloud settings communication overhead is negligible compared to the extra compute required for shared context, thereby guiding operators toward greener AI deployment strategies.