Executive Summary: 

• The energy demands of Artificial Intelligence (AI) and data centers are rising rapidly, threatening the achievement of Net Zero 2030 targets, and raising concerns about the environmental impact of AI development and data storage. 

• Data centers will account for 9 percent of the US’s total power consumption by 2030, significantly impacting the power grid and prompting tech companies to invest in sustainable energy sources and innovate solutions for energy-efficient data processing. 

• Policymakers need to support data center creation in areas of abundant sustainable power and offer incentives necessary to maintain American leadership in the tech sector while implementing more AI tools to make power grid management more efficient and secure. 

Introduction: 

The growth of AI and data centers is driving an accelerating rise in energy consumption, posing a serious challenge to global sustainability goals and prompting a shift towards sustainable energy alternatives. For example, a 2024 Google report noted a 48 percent increase in the company’s emissions compared to 2019, a significant barrier to Google’s commitment to being net zero by 2030, largely due to the increasing application and development of Artificial Intelligence (AI) tools. The majority of the increase in emissions lies in the increased energy use of data centers. 

Data centers are set to become one of the most significant contributors to energy consumption in the US, with projections estimating they will account for up to 9 percent of the nation’s total power use by 2030. In comparison, the entire residential sector accounts for 15 percent of US’s total power use in 2023. Given that traditional energy sources are finite, the growing consumption needs will likely lead to more power outages and grid strain. Due to the decentralized nature of data centers, which do not require proximity to affiliated companies, high expenses and an unstable power grid could lead to tech companies building new data centers outside of the US. This would affect citizens and economic prosperity, prompting leaders to find new energy sources to cover the deficit. As the costs of computing increase and the costs of sustainable energy technologies decrease, a number of tech giants turn their attention to renewable energy sources, such as nuclear plants, to power their growing AI needs. Policymakers must respond to this interest by creating a favorable environment for tech companies to find energy-abundant data center build sites within the US. As the intersection between the technology industry and the energy industry creates an unprecedented demand for more energy, this paper aims to analyze the impact of AI use on the power grid, the opportunity to implement more sustainable energy sources to power growing data center energy needs, and how AI tools can make utilities management more efficient. 

Data Center Energy Consumption:  

Data centers, which drive the significant energy demand increases, are facilities that store critical applications and computer systems required to support organizations’ day-to-day computational operations. This, in turn, drives much of the digital economy, from handling high-volume e-commerce transactions to sustaining online gaming platforms to big data analysis, machine learning, and artificial intelligence initiatives.  

These data centers are critical to the digital economy but put a significant strain on the United States energy grid. According to the US Energy Information Administration’s Annual Energy Outlook 2023 document, the largest data centers require the equivalent of 80,000 households of energy to operate. As such, data centers account for approximately 4 percent of the US’s total energy consumption in 2023, with the number increasing to 9 percent by 2030 as AI reliance grows, putting significant strain on the grid. Unfortunately, the average age of regional power grids in North America is 40 years, and updating and generating new capabilities to support data centers will require around $50 billion in investment in the US. 

While leadership in AI is critical to the United States’ geopolitical success, AI development has further increased the need for data center construction. Since incorporating large language models into its service, each Google search inquiry costs ten times more. A single ChatGPT question needs over nine times more electricity than a Google search. As people perform an average of 9 billion searches daily, AI companies consume progressively more power, and the connected price tag is rapidly growing. This exorbitant increase in the price of searches creates a unique motivating factor for tech companies to turn to more affordable and long-term solutions brought by renewable energy. Microsoft announced a 2024 deal with the Three Mile Island nuclear power plant, closed in 2019. The agreement would see the plant reopen in 2028, with Microsoft as the sole purchaser of the entire electricity generated over the next 20 years. Other tech companies will undoubtedly follow in Microsoft’s lead.  

AI’s Role in Energy Consumption:   

Impact of AI Training and Energy Use 

AI is a major application for data centers. AI training, especially for large models like GPT and DALL·E, consumes a tremendous amount of energy. In 2019, researchers at the University of Massachusetts Amherst discovered that training large language models can generate significant carbon emissions, with a single AI model producing over 626,000 pounds of CO2—comparable to the lifetime emissions of five cars. A recent study revealed that training GPT-3, which has 175 billion parameters, consumed 1,287 MWh of electricity and resulted in 502 metric tons of CO2 emissions, equivalent to the annual emissions of 112 gasoline-powered cars.   Regardless of the emission parameters connected to training AI, the number and use of data centers will only increase as AI tools become more widespread, thus increasing CO2 production. Moving to more sustainable energy sources is the only viable response to offset the power consumption and algorithm training emissions. 

Implementing AI tools in Power Management 

AI’s power consumption creates a significant challenge for businesses and policymakers, but AI tools can be implemented to limit demand and increase energy efficiency. For example, AI used in digital twinning, where machine learning and 3D modeling tools are used to make copies of power generation sites to predict corrosion, has decreased inspection and corrosion costs by 10 percent. Other digital twinning operations allow for simulations of real-life wind and water patterns, increasing the accuracy of weather predictions and the ability to calculate and adjust settings. Similarly, machine learning tools can predict downtime based on existing power grid data through simulations, allowing for more efficient power deployment and outage scheduling. Finally, AI tools supplement monitoring and safety assessments. The tools allow for continuous monitoring of thousands of live video feeds without needing a fleet of security guards: cutting down surveillance costs, allowing for more efficient and easier deployment of workers, as well as making intruder identification faster.  

Policymaking: 

Even with these efficiencies, policymakers may need to take action to address issues with energy generation to ensure firms continue to build data centers in the US. Data centers contributed $486 billion to the U.S. GDP in 2021, paid $403.5 billion in local, state, and federal taxes between 2017 and 2021, and accounted for around 460,000 new jobs. Building data centers in the United States allows firms to directly impact the US economy and job market, whereas creating a hostile policy environment could lead to companies turning to other countries to build data centers, stagnating the US economy. Indeed, according to NVIDIA CEO Huang, locating data centers in US areas where energy is plentiful is necessary. 

Without action, however, new data center builds could shift overseas.  In Europe, countries that possess cheap power from sustainable and renewable sources, such as the Nordics or France, have seen a surge in data center builds. Additionally, locations that offer tax reductions and other incentives to companies to set up in large financial and tech hubs, such as Germany and the UK, have also seen an increase in data centers. US policymakers can create the most beneficial environment for AI growth by combining sustainable power-abundant locations with financial incentives.  For example, New York State has announced $275 million in funding for Consortium AI, an initiative to build an AI research computational facility in Upstate New York, a location with abundant hydropower. If federal policymakers wish to increase funding for AI innovation, it may best be focused on building and maintaining computational centers to ensure that the US remains a leader and a first choice among tech companies in which to locate data centers.  

Regarding nuclear energy policy, US policymakers are embracing supportive energy regulations but need to decrease the 3 – 5 year review process time to ensure that plants are built quickly and efficiently. Ensuring that regulations do not stifle data centers’ growth while supporting the sustainability energy partnerships that tech companies are undertaking should remain at the forefront of decision-making. Additionally, the government should work towards educating the public on the benefits of nuclear energy. While public opinion has grown more positive over the past 30 years, it is necessary to alleviate worry among citizens through educational and promotional material. 

Further, early partnerships with tech companies offering AI tools to support utility management are vital. The next few years will be defining in terms of energy use and maintaining power grids, putting pressure on policymakers to ensure the best tools are used to drive power accumulation. Using more technologically advanced power infrastructures requires AI tools to ensure efficient management and decrease downtime while allowing simulation of decision-making without sunken cost worry. The more effective the power grids, the more energy is produced and used to benefit the American economy and consumers. 

Conclusion: 

The energy costs associated with AI and data centers are set to rise drastically, which can jeopardize the Net Zero 2030 commitments. Tech companies are turning to sustainable energy to minimize costs and guarantee long-term solutions for energy procurement. The US currently leads the data center market. To maintain this performance, policymakers need to create incentives in the form of grants, financing plans, and AI-growth-centric regulations for technological companies to set up data centers in areas with a high abundance of sustainable energy. Future-centric power generation requires advanced AI tools to support efficiency, surveillance, and digital twinning to ensure that American citizens are not the primary victims of power shortages caused by data center-driven increases in energy consumption during peak hours.