Recent data from Google revealed the “side effect” of the growth in the application of artificial intelligence (AI) technology to Silicon Valley giants: over the past five years, Google's greenhouse gas emissions have soared 48% due to the expansion of data centers supporting AI systems.
On Tuesday, July 2, EST, Google's annual environmental report shows that in 2023, that is, last year, Google's pollution emissions reached 14.3 million tons of carbon equivalent, an increase of 48% over the 2019 baseline level, and an increase of 13% last year. Last year, Google's energy-related emissions, mainly from data center electricity consumption, increased 37%, accounting for a quarter of the company's total greenhouse gas emissions. Supply chain emissions, which account for the largest share and account for 75%, increased 8% last year.
According to the report, the increase in carbon emissions last year highlights the challenge of reducing carbon emissions while investing in the Big Language Model (LLM), or related applications and infrastructure construction. The report admits that “the future impact of AI on the environment” is “complex and difficult to predict.” The report predicts that supply chain emissions will continue to grow in the near future due to increased demand for infrastructure to run AI systems.
The above report on Tuesday means that the demand for electricity generated by AI systems makes Google face the threat that it will be difficult to achieve carbon reduction goals. Google previously promised to reduce all direct and indirect net greenhouse gas emissions to zero by 2030, and that the grid it operates at that time will be entirely powered by carbon-free energy 24 hours a day.
Meanwhile, Tuesday's report warned that last year, some clean energy projects “terminated”, reducing the amount of renewable energy that Google received.
Kate Brandt, Google's chief sustainability officer, said in the report that Google is still committed to achieving the above 2030 goals, but emphasized that the goal is “extremely ambitious.” “We do expect our emissions to continue to rise until we fall back to our targets,” Brandt said. She also said that Google is working “very hard” to reduce emissions, including signing a clean energy agreement, and that AI also provides “huge opportunities” to address climate issues.
The media pointed out that Silicon Valley giants such as Google, Amazon, and Microsoft previously announced plans to invest tens of billions of dollars in the AI field. This has already caused climate experts to worry that energy-intensive tools and systems have a negative impact on the environment.
Wall Street News mentioned earlier this week that with the rapid development of AI technology, especially the rise of LLM and generative AI, technology companies' demand for electricity has exploded, and they have begun to compete for nuclear power resources. About one-third of US nuclear power plant owners are in talks with technology companies to supply electricity for new data centers. Among them, Amazon Cloud Service AWS is close to reaching a direct power supply agreement with Constellation Energy, the largest nuclear power plant operator in the US. In March of this year, AWS also bought a nuclear-powered data center in Pennsylvania for $650 million.
This marriage between nuclear power and technology giants has sparked controversy in many parts of the US. On the one hand, the combination of nuclear power and data centers can match the most reliable electricity in the power grid with the richest customers. Furthermore, this new “factory network integration” arrangement means that the construction speed of data centers may be greatly accelerated because there is almost no need for new grid infrastructure. Data centers may also avoid paying transmission and distribution costs, which account for most of the electricity bill.
On the other hand, such power supply agreements also mean that technology companies are not adding new green energy to meet their soaring electricity demand; they are actually transferring existing electricity resources. Data centers may remove stable power generation resources from power grids, exacerbate increasingly serious electricity reliability problems in many parts of the US, and push up electricity prices.