Advanced Computing in the Age of AI | Saturday, July 20, 2024

Google Struggles to Meet Sustainability Goals Amid AI Expansion 

Google has an ambitious goal to reduce its carbon emissions to net zero by the year 2030. This would require cutting its carbon emissions in half compared to its 2019 baseline. However, the energy demands of artificial intelligence pose a significant threat to this goal.

In its 2024 Environmental Report, the search engine giant admitted that it produced 13% more carbon emissions in 2023 compared to the prior year. This translates to 14.3 million metric tons of carbon dioxide pollution. Google attributes the rise in emissions to its supply chain operations and higher energy consumption of its data centers. 

Several other tech companies are facing challenges in meeting energy demands for AI systems. Data centers are notoriously energy-hungry, especially those used for training AI models. 

Amazon Web Services and several other big tech companies are turning to nuclear power to fuel their projects. According to a report by the Wall Street Journal, roughly a third of nuclear power plants in the U.S. are discussing deals with tech companies to provide electricity for their data centers. 

While graphics processing units (GPUs) are designed to be energy efficient, the high-end models used in AI workloads tend to consume more power than the standard center processing units (CPUs). 

Google disclosed that 37% of the jump in its market-based emissions is a result of its data centers’ energy use. The company used 25,910 gigawatt-hours of electricity in 2023. To put that into perspective, a single gigawatt-hour can power several hundred thousand households for one hour.

“As we further integrate AI into our products, reducing emissions may be challenging due to increasing energy demands from the greater intensity of AI compute, and the emissions associated with the expected increases in our technical infrastructure investment,” Google said in its report. It added that data center consumption is expected to continue increasing. 

To overcome the challenge of increasing energy demands, Google says it plans to make its data centers, hardware, and AI models more energy-efficient. According to Google, 67% of its energy consumption in 2023 was from carbon-free sources, and it has already signed contracts to purchase 4 gigawatts of clean energy from Texas, Australia, Belgium, and other sources.

(Gorodenkoff/Shutterstock)

Last year, Google also operationalized a geothermal project in Nevada to contribute carbon-free energy to the local grid. The company’s sustainability initiatives also include the preservation of water used in data centers for cooling purposes. Google's Water Stewardship program was successful in replenishing about 1 billion gallons of water in 2023. This equates to about 18% of its freshwater consumption. 

Google seems to be committed to achieving its sustainability goals by 2030. However, it is going to be challenging. According to the International Energy Agency (IEA), the world’s data centers are going to consume 10 times as much electricity in 2026 as they did last year.

While Google has a long way to go, the company is heading in the right direction. It is establishing strategic partnerships around the globe, contracting for carbon removal credits, and implementing initiatives to enhance the energy efficiency of its operations.

Related Items 

Permutable AI Offers an Initial Glimpse of GreenProof Framework Exposing Sustainability Claims in the Financial Sector 

Going Green in the AI Era: an Imperative for Performance-Intensive Applications 

Green IT: Burden or Opportunity?

EnterpriseAI