How DeepMind’s AI Cuts Data‑Center Cooling Costs by 40%
Google’s DeepMind AI system autonomously optimizes data‑center cooling, reducing energy consumption by up to 40%, and its success is prompting plans to extend the technology to national power grids and other large‑scale infrastructure, showcasing AI’s potential for massive environmental impact.
Preventing servers in data centers from overheating has become a priority for internet giants to ensure stable services.
Companies such as Facebook, Microsoft, and Google are exploring various methods to cut cooling costs: Facebook uses external air flow, Microsoft experiments with underwater data centers, and Google, together with DeepMind, applies AI to manage data‑center equipment.
A data center is a large facility that houses servers and networked computers storing most of the Internet’s data and providing the compute power for cloud computing.
These facilities generate massive heat, with energy density over a hundred times that of a typical office building; without continuous climate control, equipment would overheat within minutes.
Since 2016, Google has employed an AI system developed by DeepMind to prevent its global data centers from overheating. The AI‑driven recommendation system has reduced cooling energy consumption by about 40%.
DeepMind’s neural network collects data from the cooling system every five minutes to predict how different actions will affect future energy use.
The AI identifies actions that satisfy strong safety constraints while minimizing energy use; these actions are sent to the data center, where local control systems verify and implement them.
Data centers contribute roughly 2 % of global greenhouse‑gas emissions, comparable to aviation. Google operates 15 of the world’s most energy‑efficient data centers and seeks further carbon reductions.
Performance, measured by the cooling‑efficiency metric kW/ton, improved from 12 % to about 30 % over nine months under AI control.
DeepMind believes the technology could be extended to other large‑scale infrastructures, such as the UK’s national power grid, potentially cutting energy use by 10 % without new infrastructure.
Executives from DeepMind and Google emphasize that the algorithms are generic and can be trained on any dataset to predict optimal control actions.
Related links: https://deepmind.com/blog/safety-first-ai-autonomous-data-centre-cooling-and-industrial-control/ https://www.forbes.com/sites/samshead/2018/08/18/google-trusts-deepmind-ai-to-manage-data-centre-cooling/#37454e4168bd
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
21CTO
21CTO (21CTO.com) offers developers community, training, and services, making it your go‑to learning and service platform.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
