Green Computing in the Era of Large Models: Highlights from the 2023 Inclusion·Bund Conference

Since the beginning of the year, the large‑model industry has experienced a boom, with over 100 domestic models publicly released, leading to severe GPU shortages and long reservation queues.

The surge in demand for intelligent compute and the accompanying rise in energy consumption prompted the 2023 Inclusion·Bund "Green Computing in the Cloud‑Native Era" forum, where government, industry, academia, and research representatives discussed future development directions.

Shanghai Communications Administration Deputy Director He Feng highlighted three key recommendations: accelerate key green‑compute technology R&D to improve data‑center resource utilization, standardize metrics and definitions for data centers, and deepen industry collaboration to build a new green‑compute ecosystem.

Secretary‑General Xie Ji of the China SME Association emphasized the importance of ESG disclosure and rating standards for small and medium enterprises, aiming to drive green, low‑carbon transformation and carbon‑neutral computing.

Ant Group stressed the need for an end‑to‑end green‑compute system, particularly focusing on the application stage of compute to break efficiency bottlenecks and promote collaborative green development.

End‑to‑end green computing spans power generation, compute production (including intelligent‑compute center builders, hardware vendors, and cloud providers), and the entire application chain. The China Academy of Information and Communications Technology’s 2023 white paper noted that compute‑application stage green computing offers the greatest potential for efficiency gains yet remains under‑addressed.

Du​an Xiaodong, Deputy Director of China Mobile Research Institute, described the construction of a new intelligent‑compute center based on five core aspects—new interconnect, new compute efficiency, new storage, new platform, and new energy saving—to build an open, collaborative, and energy‑efficient compute foundation.

Prof. Liu Yunxin of Tsinghua University and Intel China’s Chief Scientist Zhang Yimin presented AI‑driven techniques for data‑center energy optimization, focusing on high‑efficiency intelligent computing systems and AI‑enabled performance improvements to support a carbon‑neutral smart‑compute platform.

Prof. Jin Hai, a fellow of the IEEE and Vice‑Chair of the China Computer Federation, proposed three near‑data processing models—storage upshift, compute downshift, and distributed expansion—to overcome the “storage wall” and improve data‑centric compute efficiency.

Ant Group CTO He Zhengyu outlined a green, high‑efficiency intelligent computing framework comprising low‑carbon compute selection, global resource scheduling, workload optimization, and green monitoring, which saved 41.4% of machine‑resource consumption in 2022 while supporting 30 billion daily page views.

Industry leaders such as Ren Xiang of the China Electronics Standardization Institute, Wang Chaoyang of Alibaba Cloud IDC, Zhao Shuai of Inspur, and Zhu Youhong of PICC Technology emphasized the necessity of cross‑sector collaboration, standardization, carbon‑aware metrics, and the linkage between compute and carbon emissions to enable effective measurement and green‑computing practices.

Overall, the conference reaffirmed that compute is the foundation for digital transformation and deep intelligence, and that coordinated hardware‑software innovation is essential to address the growing supply‑demand gap and achieve sustainable, high‑performance computing.