By Haley Zaremba
- The energy consumption of AI is on a rapid upward trajectory, surpassing that of renewable energy sources and exceeding global greenhouse gas emissions.
- Tech firms are expressing concerns about securing an adequate power supply, particularly clean energy, to fuel their IoT-driven initiatives.
- Industry insiders suggest that nuclear power could present a viable solution to meet the substantial energy demands of the tech sector without compromising decarbonization goals.
In recent years, major tech companies have been at the forefront of advocating for sustainable energy practices. Silicon Valley behemoths have invested significant resources into making their operations more eco-friendly and advancing the adoption of solar power. However, the focus on energy usage within the tech industry has shifted from optimism to apprehension, with Artificial Intelligence (AI) emerging as a key concern.
Despite the rapid expansion of renewable energy production and improved incentives for capacity growth in the United States, the soaring energy requirements of AI are driving this surge. For instance, the training process for GPT-3, the predecessor of ChatGPT, consumed approximately 1,287 watt-hours of electricity and 10,000 PC chips. This energy expenditure is equivalent to powering 121 U.S. households for a year and generating 550 tonnes of carbon dioxide.
The energy consumption and carbon footprint of AI and its associated data centers are escalating rapidly, a trend that will intensify as AI becomes more pervasive, reshaping our interactions and workflows with technology. Presently, AI alone consumes as much energy annually as entire developed nations and rivals the energy footprint of Bitcoin. If current trends persist, Gartner predicts that AI will account for a staggering 3.5% of global energy consumption by 2030.
However, this substantial energy usage is accompanied by significant greenhouse gas emissions. While Big Tech utilizes some renewable energy sources, it still heavily relies on fossil fuels, which are insufficient to meet its energy demands. Data centers and distribution networks collectively emit as much carbon dioxide annually as the nation of Brazil, as per data from the International Energy Agency.
The projected energy requirements of the tech sector, primarily driven by the rapid expansion of AI, are so substantial that industry leaders are not only concerned about obtaining sufficient clean energy to manage their carbon footprint but also about securing an adequate energy supply overall. The Wall Street Journal recently highlighted that “there is already a supply shortage, with new data centers being constructed more rapidly than new power generation facilities.”
According to Bill Vass, an executive vice president at Amazon Web Services, a new data center is commissioned somewhere globally every three weeks. However, this pace is unfeasible if there is an insufficient electricity supply to power these new facilities. Due to energy sourcing challenges, construction timelines for certain data centers have been delayed multiple times.
Consequently, the tech industry finds itself in a predicament. Waiting for clean energy to power their operations is not a viable option. As quoted by the Wall Street Journal, Toby Rice, CEO of EQT, a major U.S. organic-oil producer, emphasized, “Tech is not going to wait seven to ten years to get this system built.” This leaves reliance on fossil fuels as a probable course of action.
Nevertheless, some industry insiders are placing their bets on nuclear power to fulfill the substantial energy demands of the tech sector without impeding decarbonization efforts. Others are banking on technological innovations to rescue the industry from this dilemma. At a Bloomberg event following the World Economic Forum’s annual meeting earlier this year, OpenAI CEO Sam Altman emphasized, “There is no way to get there without a breakthrough,” alluding to nuclear fusion as a potential solution.
In 2021, Altman personally injected $375 million into the U.S.-based private nuclear fusion company Helion Energy, which subsequently secured an energy supply agreement with Microsoft. Altman also serves on the board of directors for Oklo, a nuclear microreactor company that is poised for a potential public offering this year. According to Altman, the energy demands of future AI systems can be met through fission and fusion technologies.
