Google uses AI to reduce energy to cool data centres by 15pc

20 Jul 2016

A combination of machine learning and deep neural networks has cut Google’s data centre energy bills by 15pc

Google’s artificial intelligence (AI) division DeepMind claims to have cut the energy used to cool the internet giant’s data centres by 40pc, resulting in an overall PUE reduction of 15pc.

Across the world, data centres are built in places to keep the servers cool and cut down on energy usage. For example, Apple is building an €850m data centre in Athenry near Ireland’s west coast.

Internet giant Google also has data centres all over the world, including two in Dublin.

Future Human

It is estimated that data centres are responsible for 2pc of global greenhouse gas emissions. Google alone is estimated to have used 4,402,836 MWh of electricity in 2014, equivalent to the average yearly consumption of about 366,903 US family homes.

Can machine learning be used to cut energy bills?

Google’s AI division DeepMind claims that, by applying its machine learning to Google data centres, it has reduced the amount of energy used for cooling by up to 40pc.

“The implications are significant for Google’s data centres, given its potential to greatly improve energy efficiency and reduce emissions overall,” said Jim Gao, data centre engineer at Google.

“This will also help other companies who run on Google’s cloud to improve their own energy efficiency. While Google is only one of many data centre operators in the world, many are not powered by renewable energy as we are.

“Every improvement in data centre efficiency reduces total emissions into our environment and, with technology like DeepMind’s, we can use machine learning to consume less energy and help address one of the biggest challenges of all – climate change.”

‘We accomplished this by taking the historical data that had already been collected by thousands of sensors within the data centre and using it to train an ensemble of deep neural networks’


Google said that every data centre it operates is different with a unique architecture and environment, operating in different climates and geographies.

To address the energy consumption problem, Google began applying machine learning two years ago in order to operate data centres more efficiently. Google acquired British company DeepMind in 2014 for around £400m and, over the past few months, its researchers began working with the company’s data centre team.

“Using a system of neural networks trained on different operating scenarios and parameters within our data centres, we created a more efficient and adaptive framework to understand data centre dynamics and optimise efficiency,” said DeepMind research engineer Rich Evans.

“We accomplished this by taking the historical data that had already been collected by thousands of sensors within the data centre – data such as temperatures, power, pump speeds, setpoints, etc – and using it to train an ensemble of deep neural networks.”

Testing the model on a live data centre, the machine learning system was able to consistently achieve a 40pc reduction in the amount of energy used for cooling. This equates to a 15pc reduction in overall PUE (power usage effectiveness) after accounting for electrical losses and other non-cooling inefficiencies.

Google is planning to use the algorithm to find other applications such as getting more energy from the same unit, reducing semiconductor energy and water usage.

“We are planning to roll out this system more broadly and will share how we did it in an upcoming publication, so that other data centre and industrial system operators – and ultimately the environment – can benefit from this major step forward,” Gao said.

Data centre image via Shutterstock

John Kennedy is a journalist who served as editor of Silicon Republic for 17 years