Artificial intelligence and machine learning technologies will play a significant role in Google's efforts to make its massive data centers around the world more energy-efficient than they are already.
Tests conducted by the company over the past two years on the use of its DeepMind AI technology in data center operations have already helped Google reduce the energy used for data center cooling by 40 percent and overall energy usage by 15 percent.
The success of the tests is now prompting the company to explore the use of AI to deal with other data center challenges, Google DeepMind research engineer Rich Evans and data center engineer Jim Gao wrote in the company's Green Blog this week.
Examples include figuring out how to squeeze more energy from the same unit of input, reduce the energy required for semiconductor manufacturing, cut down on water usage and improve overall throughput in a manufacturing facility.
"Given how sophisticated Google's data centers are already, it's a phenomenal step forward," the two engineers said referring to the 40 percent reduction in cooling energy consumption that Google has achieved with DeepMind. "The implications are significant for Google's data centers, given its potential to greatly improve energy efficiency and reduce emissions overall."
Google has spent hundreds of millions of dollars in recent years trying to reduce the carbon footprint of its data centers. The company has said it eventually wants all of its data centers to be powered exclusively by renewable energy sources, and it has been acquiring new facilities and repurposing existing ones to meet that goal. Google has also spent heavily on building ultra-efficient data center servers and in exploring a range of alternate approaches for cooling its data centers, like water-based cooling, evaporative cooling and seawater cooling.
The company's goal with DeepMind is to use all the historical data it has gathered over the years from temperature sensors, power sensors, pumps and other data center equipment to create predictive models for energy usage for its data centers. The purpose of the predictive models is to simulate recommended actions for ensuring maximum power usage effectiveness within data centers, the two Google engineers said.
"Each data center has a unique architecture and environment," Evans and Gao wrote. "A custom-tuned model for one system may not be applicable to another. Therefore, a general intelligence framework is needed to understand the data center's interactions."
According to the two engineers, tests of the models have shown that the AI system is able to consistently reduce cooling energy costs by up to 40 percent. Google will release details of its experiments using DeepMind in an upcoming technical publication, they added.
Google claims that its data centers are already among the most energy-efficient in the world. The company has claimed that its data centers use barely 50 percent of the energy consumed by most other data centers of comparable size.
In recent years, Google has committed to purchasing some 2.2GW of renewable energy, which it says has made it the largest purchaser of renewable energy in the world outside of energy companies. Google has also committed to investing about $2.5 billion in various renewable energy projects in the United States and around the world.