# Towards Net-Zero Carbon Emissions in Network AI for 6G and Beyond

arXiv: https://arxiv.org/abs/2401.01007 · HTML: https://mlsanity.com/paper/2401.01007/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 8 of its 16 references that are in the map.
- First posted on arXiv 18 Sep 2023.

## Builds on (most cited first, 8 of 8)

- [Green AI](https://mlsanity.com/paper/1907.10597/) (2019)
- [Carbon Emissions and Large Neural Network Training](https://mlsanity.com/paper/2104.10350/) (2021)
- [Imitation Learning-based Implicit Semantic-aware Communication Networks: Multi-layer Representation and Collaborative Reasoning](https://mlsanity.com/paper/2210.16118/) (2022)
- [Distributed Optimization for Energy-efficient Fog Computing in the Tactile Internet](https://mlsanity.com/paper/2001.10199/) (2020)
- [Towards Self-learning Edge Intelligence in 6G](https://mlsanity.com/paper/2010.00176/) (2020)
- [6G Network AI Architecture for Everyone-Centric Customized Services](https://mlsanity.com/paper/2205.09944/) (2022)
- [Time-sensitive Learning for Heterogeneous Federated Edge Intelligence](https://mlsanity.com/paper/2301.10977/) (2023)
- [Dynamic Network Slicing for Scalable Fog Computing Systems with Energy Harvesting](https://mlsanity.com/paper/2001.10300/) (2020)

## Built on by (most cited first, 3 of 3)

- [Towards Agentic AI Networking in 6G: A Generative Foundation Model-as-Agent Approach](https://mlsanity.com/paper/2503.15764/) (2025)
- [The Energy Cost of Artificial Intelligence Lifecycle in Communication Networks](https://mlsanity.com/paper/2408.00540/) (2024)
- [Sustainable LSTM-Based Precoding for RIS-Aided mmWave MIMO Systems with Implicit CSI](https://mlsanity.com/paper/2509.12658/) (2025)

## Abstract (arXiv)

A global effort has been initiated to reduce the worldwide greenhouse gas (GHG) emissions, primarily carbon emissions, by half by 2030 and reach net-zero by 2050. The development of 6G must also be compliant with this goal. Unfortunately, developing a sustainable and net-zero emission systems to meet the users' fast growing demands on mobile services, especially smart services and applications, may be much more challenging than expected. Particularly, despite the energy efficiency improvement in both hardware and software designs, the overall energy consumption and carbon emission of mobile networks are still increasing at a tremendous speed. The growing penetration of resource-demanding AI algorithms and solutions further exacerbate this challenge. In this article, we identify the major emission sources and introduce an evaluation framework for analyzing the lifecycle of network AI implementations. A novel joint dynamic energy trading and task allocation optimization framework, called DETA, has been introduced to reduce the overall carbon emissions. We consider a federated edge intelligence-based network AI system as a case study to verify the effectiveness of our proposed solution. Experimental results based on a hardware prototype suggest that our proposed solution can reduce carbon emissions of network AI systems by up to 74.9%. Finally, open problems and future directions are discussed.