Green AI Benchmarking of Domestic Accelerator–Driven Foundation Models: Performance, Efficiency, and Carbon Footprint Analysis
Keywords:
Green AI, foundation models, domestic AI accelerators, carbon footprint, edge-cloud computing, benchmarking, sustainabilityAbstract
The rapid proliferation of large-scale foundation models has brought the environmental sustainability of artificial intelligence into sharp focus, prompting the emergence of Green AI as a rigorous paradigm for evaluating performance jointly with energy consumption and carbon emissions. This paper presents a comprehensive benchmarking framework for domestic accelerator–driven foundation models, with an emphasis on edge-native architectures trained entirely on non–GPU hardware such as China-developed AI accelerators. The study systematically examines the interplay among model throughput, inference latency, energy efficiency, training dynamics, and cradle-to-grave carbon footprint across heterogeneous hardware ecosystems. We adopt a system-level perspective that integrates infrastructure considerations, software–hardware co-design, data center energy mix, and supply chain factors into a unified analytical lens. By comparing domestic accelerator–based systems against conventional GPU–centric deployments, the paper reveals structural trade-offs that transcend simple floating-point operations per second comparisons, highlighting the importance of memory bandwidth utilization, compiler maturity, and workload scheduling. The analysis extends to governance and policy dimensions, discussing how national AI sovereignty, semiconductor supply chain resilience, and carbon accounting standards interact with system design choices. Through this multi-layered evaluation, the paper argues for an expanded definition of computational efficiency that incorporates geopolitical, infrastructural, and lifecycle carbon costs, while proposing benchmarking methodologies that align with emerging regulatory frameworks for sustainable computing. The findings indicate that domestic accelerator ecosystems, despite immature software stacks, can yield significant carbon efficiency advantages when coupled with regionally optimized energy grids and tailored model architectures, underscoring the need for context-aware Green AI metrics.
References
1. Strubell, E., Ganesh, A., & McCallum, A. (2019). Energy and policy considerations for deep learning in NLP. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 3645–3650.
2. Patterson, D., Gonzalez, J., Le, Q., Liang, C., Munguia, L.-M., Rothchild, D., So, D., Texier, M., & Dean, J. (2021). Carbon emissions and large neural network training. arXiv preprint arXiv:2104.10350.
3. Lacoste, A., Luccioni, A., Schmidt, V., & Dandres, T. (2019). Quantifying the carbon emissions of machine learning. arXiv preprint arXiv:1910.09700.
4. Schwartz, R., Dodge, J., Smith, N. A., & Etzioni, O. (2020). Green AI. Communications of the ACM, 63(12), 54–63.
5. Wu, C.-J., Raghavendra, R., Gupta, U., Acun, B., Ardalani, N., Maeng, K., Chang, G., Behram, F., Huang, J., Bai, C., Gschwind, M., Gupta, A., Ott, M., Melnikov, A., Candido, S., Brooks, D., Chauhan, G., Lee, B., Lee, H.-H. S., ... Hazelwood, K. (2022). Sustainable AI: Environmental implications, challenges and opportunities. Proceedings of Machine Learning and Systems, 4, 795–813.
6. Anthony, L. F. W., Kanding, B., & Selvan, R. (2020). Carbontracker: Tracking and predicting the carbon footprint of training deep learning models. arXiv preprint arXiv:2007.03051.
7. Faiz, A., Kaneda, S., Wang, R., Osi, R., Sharma, P., Chen, F., & Jiang, L. (2023). LLMCarbon: Modeling the end-to-end carbon footprint of large language models. arXiv preprint arXiv:2309.14393.
8. Li, Z., Zhuang, S., Guo, S., Zhuo, D., Zhang, H., Ng, A. Y., & Stoica, I. (2023). Alpa: Automating inter- and intra-operator parallelism for distributed deep learning. Proceedings of the 17th USENIX Symposium on Operating Systems Design and Implementation, 559–578.
9. Miao, X., Oliaro, G., Zhang, Z., Cheng, X., Wang, Z., Wong, R. Y. Y., Chen, Z., Arfeen, D., Abhyankar, R., & Jia, Z. (2022). Towards efficient generative large language model serving: A survey from algorithms to systems. arXiv preprint arXiv:2312.15234.
10. Dodge, J., Prewitt, T., Tachet des Combes, R., Odmark, E., Schwartz, R., Strubell, E., Luccioni, A., Smith, N. A., DeCario, N., & Buchanan, W. (2022). Measuring the carbon intensity of AI in cloud instances. Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, 1877–1894.
11. Chen, C., Wang, C., Li, Y., Wan, Z., Geng, M., Xiao, J., ... & Peng, Y. (2026). JuZhou 1.0 Technical Report: The First Edge-Native Text-to-Image Foundation Model Trained Entirely on China-Developed AI Accelerators. arXiv preprint arXiv:2606.28421.
12. Luccioni, A. S., Viguier, S., & Ligozat, A.-L. (2023). Estimating the carbon footprint of BLOOM, a 176B parameter language model. Journal of Machine Learning Research, 24(253), 1–15.
13. Henderson, P., Hu, J., Romoff, J., Brunskill, E., Jurafsky, D., & Pineau, J. (2020). Towards the systematic reporting of the energy and carbon footprints of machine learning. Journal of Machine Learning Research, 21(248), 1–43.
14. Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the dangers of stochastic parrots: Can language models be too big? Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610–623.
15. Reddi, V. J., Cheng, C., Kanter, D., Mattson, P., Schmuelling, G., Wu, C.-J., Anderson, B., Breughe, M., Charlebois, M., Chou, W., Chukka, R., Coleman, C., Davis, S., Deng, P., Diamos, G., Duke, J., Fick, D., Gardner, J. S., Hubara, I., ... Zhou, Y. (2020). MLPerf inference benchmark. Proceedings of the 2020 ACM/IEEE 47th Annual International Symposium on Computer Architecture, 446–459.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Journal of Engineering Systems and Digital Innovation
This work is licensed under a Creative Commons Attribution 4.0 International License.
