See floating-point operation.
Floating-point operation A single arithmetic step in which a computer’s work is counted. An EFLOPS is a billion billion of them a second. Chapter 2.
From The AI Compute Race. Read the book
Sources that use it
- Ministry of Industry and Information Technology and five other departments of the People’s Republic of China, 算力基础设施高质量发展行动计划, the Action Plan for the High-Quality Development of Computing Power Infrastructure, document 工信部联通信〔2023〕180号, 8 October 2023 (the 2025 national targets, p. 2 of the plan — computing-power scale above 300 EFLOPS, intelligent computing 35 percent of it, and balanced, coordinated development of eastern and western computing power, in the book’s translation of the Chinese, and targets set on the day of issue rather than measurements: the plan cannot answer whether they were met; the plan’s own definition, p. 1, that computing power is “a new type of productive force” integrating information computing, network carrying and data storage power, provided to society mainly through computing power infrastructure; the glossary at Annex 1, pp. 13-14, measuring computational power in FLOPS and estimating 1 EFLOPS at about five Tianhe-2A systems or 500,000 mainstream server CPUs, by the plan’s own reckoning with no method given; the supply-chain clause, p. 10, to “gradually form independent and controllable solutions” and to build computing infrastructure on secure and trusted basic hardware and software; the plan never uses the term 算力经济, the computing power economy, which is CAICT’s; issued jointly with the Office of the Central Cyberspace Affairs Commission, the Ministry of Education, the National Health Commission, the People’s Bank of China and SASAC) In the Sources
- China Academy of Information and Communications Technology, 算力经济发展研究报告, the Computing Power Economy Development Research Report, September 2025 (the layout commonly called Eastern Data, Western Computing, a name the report itself does not print, in the institute’s own count, p. 20 of the PDF, printed p. 14 — eight national hub nodes, Beijing-Tianjin-Hebei, the Yangtze River Delta, the Guangdong-Hong Kong-Macao Greater Bay Area, Chengdu-Chongqing, Guizhou, Inner Mongolia, Gansu and Ningxia, and thirteen data-center clusters, built to guide the growing eastern demand for computing power to the west in an orderly way, in the book’s translation of the Chinese, and from a state think tank rather than a ministry issuing policy: CAICT is the research institute directly under China’s Ministry of Industry and Information Technology; the thirteen clusters named at the same page are Zhangjiakou, Shanghai, Suzhou, Jiaxing, Wuhu, Shaoguan, Chengdu, Chongqing, Qingyang, Zhongwei, Gui’an, Horinger and Jining; the report’s own statement at p. 39 that the position of the computing power economy in national and regional development strategies “is not yet clear”, and its recommendation at p. 40 that the state make that position clear, so the definition is a proposal it wants adopted; its other figures are CAICT’s own estimates with no method given on the page — intelligent computing scale 748 EFLOPS FP16 at end-March 2025, p. 27, and compute-center electricity about 166 TWh in 2024, 1.68 percent of national consumption, p. 25; a GDP elasticity of 0.0426 from a fixed-effects panel of 336 prefecture-level cities 2017-2022 whose own annex gives the low-GDP group a coefficient of -0.3609, pp. 42-43) In the Sources
Related terms
From the glossary of The AI Compute Race.
How to cite this page
Dryburgh, Lee S. “EFLOPS.” In The AI Compute Race. DRYBURGH. https://dryburgh.com/vol1/glossary/eflops (accessed 2026-11-01).
BibTeX
@misc{dryburgh-eflops-2026,
author = {Dryburgh, Lee S.},
title = {{EFLOPS}},
howpublished = {In \emph{The AI Compute Race}, DRYBURGH},
year = {2026},
url = {https://dryburgh.com/vol1/glossary/eflops},
note = {Accessed 2026-11-01}
}