While AI data centers consume electricity comparable to entire cities, driving demand for various energy sources and straining power grids while raising public concerns about rising electricity bills, the massive computing power's energy consumption is merely a secondary effect. This conclusion comes from a recent research paper that aims to quantify how artificial intelligence is reshaping the energy system. The article argues that AI's most profound impact is not in the energy required to power the technology itself, but in its application scenarios. The core prediction: artificial intelligence may further solidify the dominant position of fossil fuels.
The study built a model to simulate scenarios where AI simultaneously enhances the production efficiency of both renewable energy and fossil fuels. The calculations found that after improving the economic viability of oil and gas extraction, global annual carbon dioxide emissions would increase by 470 million to 1.8 billion tons. Paper co-author Holly Alpine stated, "The lower-end increment is roughly equivalent to Mexico's total annual carbon emissions." The research was published in the journal Nature. Holly Alpine and her husband Will Alpine, also a co-author, both previously worked at Microsoft on sustainability-related projects. They said they left the company due to disappointment over Microsoft's business involvement in the oil and gas sector, and now run a climate advocacy organization calling for restrictions on using artificial intelligence to expand oil and gas extraction. A Microsoft spokesperson responded that the company believes AI can accelerate the clean energy transition and that these employees are central to its sustainability mission.
The debate around AI and climate change typically contrasts the massive energy consumption of data centers with AI's potential to drive green technology development. Data on computing power consumption is constantly being updated. Recently, Amazon disclosed that its data center project in Texas was approved to draw 7.65 gigawatts of natural gas power; climate scientist Zeke Hausfather calculated that intelligent AI agents can consume up to 600 times more energy than standard command interactions. In contrast, major tech companies are heavily promoting AI's green potential. A 2023 report co-authored by Google and the Boston Consulting Group suggested that through grid optimization and other means, AI could help reduce global emissions by 10%. In 2024, OpenAI's Sam Altman stated that artificial intelligence could help "solve the climate puzzle."
The Alpines sought to systematically assess the productivity improvements brought by AI. To understand their perspective, one can look at the statements of oil and gas company executives. Now, in their financial reports, major oil and gas companies often disclose their progress in using artificial intelligence to analyze geological data and optimize extraction operations. For a long time, oil and gas engineers have been pushing the boundaries of exploration and enhancing oilfield recovery rates, and artificial intelligence is the latest tool in this process. Consulting firm Wood Mackenzie estimates that AI could help extract an additional one trillion barrels of crude oil; Goldman Sachs predicts that large-scale adoption of AI could accelerate project timelines and reduce capital expenditures, potentially lowering the price per barrel of crude oil by up to $11.
Not everyone agrees that AI will solidify the position of fossil fuels. Artem Abramov, Deputy Head of Rystad Energy, believes the Alpines underestimate AI's potential to boost the iteration of low-carbon technologies. He does not deny that AI can help oil and gas companies reduce costs, but he is uncertain about the final outcome. "Most of the cost savings are converted into investor returns, rather than extracting more crude oil," Abramov said. The Alpines argue that there is increasing evidence that falling extraction costs will stimulate energy consumption growth. For example, ExxonMobil mentioned in a recent earnings report that, with the help of AI models, four potential drilling areas were identified in Guyana that traditional exploration methods had missed. Will Alpine stated, "We should take these industry reports seriously. AI is changing the cost structure and commercial viability of the oil and gas industry, thereby delaying the energy transition process."
Comments