Mike Kai Chen/Bloomberg/Getty Images/File
Tuesday, November 28, 2023, at Google's Bay View Campus in Mountain View, California, USA.
new york
CNN
—
As Google races to incorporate artificial intelligence into its core products (with sometimes disappointing results), a problem is brewing behind the scenes: The systems needed to run its AI tools are significantly increasing the company's greenhouse gas emissions.
AI systems require a lot of computers to run, and the data centers that run them are like warehouses filled with powerful computing equipment and consume a lot of energy. It processes the data and manages the heat that all those computers generate.
The end result is that Google's greenhouse gas emissions have surged 48% since 2019, according to the company's annual environmental report. The company attributed the increase primarily to “increased data center energy consumption and supply chain emissions.”
Now, Google calls its goal of achieving net-zero emissions by 2030 “very ambitious,” and says the pledge may be subject to “uncertainties about the future environmental impacts of AI that are complex and difficult to predict.” In other words, sustainability. Company push — This is Long time ago The code of conduct slogan “Don't be evil” has become more complicated thanks to AI.
Google, like other tech rivals, is committed to investing in AI, which is widely seen as the next big tech revolution that will change how we live, work, and consume information. The company has integrated Gemini-generated AI technology into several of its key products, including Search and Google Assistant, and CEO Sundar Pichai has called Google an “AI-first company.”
But AI has a major drawback. Google and other big tech companies are now spending tens of billions of dollars each quarter to expand their power-hungry data centers to fuel their AI ambitions.
As an example of how much more demanding AI models are than traditional computing systems, the International Energy Agency estimates that a Google search query requires an average of 0.3 watt-hours of power, while a ChatGPT request typically consumes around 2.9 watt-hours of power. A study published in October by Dutch researcher Alex de Vries estimated that in a “worst-case scenario,” Google's AI systems could eventually consume as much electricity as the country of Ireland per year, assuming a full deployment of AI on current hardware and software.
“As we further integrate AI into our products, we may find it more difficult to reduce our emissions due to increased energy demand from the centralization of AI computing, as well as emissions from our expected increased investments in technology infrastructure,” Google said in a report released on Monday. The company added that electricity consumption for its data centers is currently growing faster than it can bring zero-carbon sources online.
Google said it is considering investing in clean energy sources such as wind and geothermal to power its data centers and expects its total greenhouse gas emissions to continue to rise before falling.
The large amounts of water used as a coolant to keep data centers from overheating also pose a sustainability challenge. Google has said it aims to replenish 120% of the freshwater it consumes in its offices and data centers by 2030. Just 18% of water was replenished last year, up significantly from 6% the year before.
Google is among the companies experimenting with ways to use AI to fight climate change. For example, a 2019 Google DeepMind project trained an AI model using weather forecasts and historical wind turbine data to predict wind power availability, helping make renewable energy sources more valuable to wind farmers. The company is also using AI to suggest more fuel-efficient routes for drivers using Google Maps.
“We recognize that scaling and using AI to accelerate climate action is just as important as addressing the associated environmental impacts,” Google said in the report.
