AI Exploration: Energy in the Age of AI

Exploring the big questions with the help of AI. Today’s big question: Do we have enough energy for AI?

Mar 19, 2025

Human Thoughts: I want to explore two topics here. First, how good is AI at helping us understand new information and answering some of our big questions? Second, I want the answer to my big question.

Today’s big question: Do we have enough energy for AI?

I didn’t want to read just one article because there is a lot of bias around this topic, and this is a very large topic, so one article won’t give me a very full view, but the energy demand of AI is something I want to know more about. So, how might I string together a summary of dozens of sources to get a mix of perspectives on the topic? I figured Gemini’s Deep Research could probably do this because that’s effectively what you get when you ask it a question: an aggregated response from a ton of sources around the internet. It’s like asking the internet to write a research paper for you, and some of the information is probably wrong, but a lot of it is probably right, and if this technology keeps getting better, I’m interested in what we can learn here as we keep asking the internet to write us research papers.

So, I plan to do one of these each month to track how Deep Research improves and find out what interesting reports the internet can make for us. I’ll make sure to structure it clearly so that you know when I’m talking versus the AI, and if you find this interesting and would like me to do more, please let me know!

My thoughts on this month’s report: The report focuses on the short-term and long-term energy needs of the United States with the new demands of AI. In the long term, advances in battery storage and distribution will likely allow us to effectively leverage renewables. Advancements in nuclear fission and perhaps finally achieving nuclear fusion (this is what the sun does) will allow us to be on a strong long-term path to sustainable energy that exceeds our energy demand. However, our future energy demand is likely to be very high from powering AI, fleets of humanoid robots, drones, electric vehicles (including planes, trains, and automobiles), quantum computers, cryptography, and more; our energy demands are not small.

My main concern is in the short term, can we meet these demands? I think we can, but unfortunately, we can’t do it without fossil fuels because of the intermittent nature of wind and solar. Renewables will play a huge role in the future, but it’s not just about power generation; it’s about distribution, transmission, and storage. Disclaimer: I’m not in the energy industry, I’m just passionate about the future and helping you understand technology today to help you achieve a better future tomorrow.

Now, here is the AI content; it is pretty dense reading, so if you aren’t looking for that right now, you may just want to listen to the audio summary by NotebookLM and only skim the research report. This content is written with Google’s Gemini Deep Research. Google’s NotebookLM is used to generate the audio discussion; the audio discussion is based on the Deep Research report. Because I am one person, I am unable to check all sources used in the report, and AI content can be wrong, so please don’t assume that everything in the report is factual. If you find an error made by the AI, please comment it below so other readers know about it, and remember this technology is likely to only get better from here.

AI Thoughts: AI Podcast Summarizing the Report

1×

0:00

-25:11

Energy in the Age of AI: A Near- and Long-Term Outlook for the United States

1. Executive Summary

The rapid advancement and widespread adoption of artificial intelligence (AI) are poised to exert a significant influence on the energy landscape of the United States. This report examines the critical question of whether the nation possesses sufficient energy resources to meet the escalating demands of the AI era, focusing on both the near-term (next 5-7 years) and long-term (beyond 10 years). The transformative potential of AI is undeniable, yet it is intrinsically linked to substantial energy consumption, particularly within the data centers that power its complex computations. Projections indicate an exponential surge in energy demand from these facilities, with some forecasts suggesting a fourfold increase by the end of the decade ¹. While the United States currently enjoys the status of a net energy exporter ², the sustainability of this position is under immense pressure due to AI-driven consumption. Maintaining sufficient energy will be determined by a complex interplay of factors, including the rate at which AI technologies are adopted, the success of efforts to enhance energy efficiency both within AI operations and across broader sectors, the scalability and deployment of renewable and clean energy sources, and the modernization of the nation's energy grid. This analysis suggests that significant policy adjustments and strategic investments will be essential to ensure a secure and sustainable energy future for the United States in this new age of artificial intelligence.

2. The Current US Energy Landscape

2.1. Historical Trends in Energy Production and Consumption

Over the past half-century, the energy sector in the United States has undergone a remarkable transformation. Initially, the nation consumed more energy than it produced domestically, relying on imports to bridge the gap. However, due to significant advancements in extraction technologies, particularly in the realm of crude oil and natural gas, the US has transitioned to a net energy exporter ². This shift was largely propelled by improvements in drilling techniques such as hydraulic fracturing and horizontal drilling, which gained prominence in the 2000s ². Between January and July 2024, total US energy production surpassed the same period in 1974 by an impressive 68%, representing an additional 24.0 quadrillion British thermal units (quads) produced ². While US energy consumption has also steadily increased since 1974, its growth rate has been less pronounced than that of production. In the first seven months of 2024, consumption was 32% higher than the corresponding period in 1974, marking an increase of 13.2 quads ². This growth in consumption can be attributed to factors such as population expansion and increased economic activity ². Notably, despite the overall increase in consumption, primary energy consumption has generally decreased on both a per capita basis and in terms of energy consumed per dollar of GDP since the 1970s, a trend largely attributed to improved energy efficiency ².

2.2. Composition of Current Energy Sources

In 2023, the United States relied on a diverse portfolio of energy sources, with fossil fuels constituting the majority of the mix. Petroleum accounted for 38% of the nation's energy consumption, followed closely by natural gas at 36%. Coal contributed 9%, while nuclear power and renewable energy sources each supplied 9% ⁴. This breakdown highlights the continued dominance of fossil fuels in meeting the nation's energy needs. When examining electricity generation specifically in 2023, natural gas emerged as the leading source, accounting for 43.1% of the total. Nuclear power contributed 18.6%, coal generated 16.2%, and renewable energy sources collectively provided 21.4% ⁵. This indicates a gradual shift in the electricity generation mix, with natural gas and renewables playing increasingly prominent roles compared to coal ⁶. The total electricity generation capacity in the US reached nearly 1.3 million megawatts as of January 2024, with natural gas facilities comprising the largest share of this capacity at 43.9% ⁷. Notably, solar energy capacity has experienced dramatic growth in recent years, reflecting a significant trend towards renewable energy adoption ⁷.

2.3. US Energy Consumption by Sector

Energy consumption patterns in the United States vary significantly across different sectors of the economy. In 2023, the industrial sector accounted for the largest share of energy consumption at 33%, followed by the transportation sector at 30%. Homes consumed 20% of the total energy, while commercial buildings accounted for the remaining 17% ⁴. Electricity serves as a vital energy carrier across all these sectors, and with the rise of AI, a substantial increase in electricity consumption is anticipated within the commercial sector, primarily due to the proliferation of energy-intensive data centers ¹². To illustrate the scale of individual energy use, the average US per capita daily consumption includes approximately 2.5 gallons of oil, 6.82 pounds of coal, and 260 cubic feet of natural gas. Additionally, residential per capita electricity consumption averages around 12 kilowatt-hours per day ³.

3. The Looming Energy Demand from Artificial Intelligence

3.1. The Energy Intensity of AI

Artificial intelligence, particularly the advanced forms known as generative AI, demands an unprecedented level of computational power, translating directly into substantial energy consumption ¹. The intricate algorithms and vast datasets that underpin AI require massive processing capabilities, far exceeding those of traditional computing tasks. This energy intensity becomes apparent when comparing simple online activities with AI interactions. For instance, a single query on an AI platform like ChatGPT is estimated to consume approximately ten times more electricity than a standard web search ¹³. The training of the sophisticated AI models, such as large language models, is particularly energy-intensive. Estimates suggest that the energy required to train a model like GPT-3 is equivalent to the annual power consumption of around 130 average US households ¹⁵. Furthermore, the energy demand is not limited to the training phase; once these AI models are deployed and actively used for tasks (a process known as "inference"), they continue to require significant and continuous energy to operate ¹⁵.

3.2. Data Centers: The Epicenter of AI Energy Consumption

The immense computational power required by AI is primarily housed within data centers, vast facilities filled with powerful servers and networking equipment. These data centers have become the focal point of energy consumption related to the proliferation of AI technologies ¹. In 2023, data centers in the United States are estimated to have consumed approximately 4.4% of the nation's total electricity ¹⁷. This figure is projected to rise dramatically in the near future as AI adoption accelerates. Estimates from the US Department of Energy indicate that the electricity load from data centers could double or even triple by 2028 ¹⁷. Globally, the combined electricity consumption of data centers, AI operations, and cryptocurrency mining already represented nearly 2% of the world's total in 2022, and this share is anticipated to grow substantially in the coming years ¹⁴.

4. Near-Term Energy Demand and Supply Balance (2025-2030)

4.1. Surging Electricity Demand Projections

A strong consensus is emerging from various energy and technology analysis sources, pointing towards a significant surge in electricity demand in the United States within the next five to seven years. This increase is primarily attributed to the escalating energy needs of AI and the data centers that support it ¹. The US Department of Energy projects that the electricity load from data centers is expected to double or triple by 2028 ¹⁷. McKinsey forecasts that by 2030, US data center energy consumption will reach 606 terawatt-hours (TWh), accounting for a substantial 11.7% of the total US power demand ¹. Goldman Sachs Research anticipates a global increase of 50% in data center power demand by 2027 and a staggering 165% increase by 2030, with North America expected to see the most significant capacity additions ²³. Reflecting this trend, grid planners across the US are revising their overall electricity demand growth forecasts upwards. National power demand is now projected to grow at an average annual rate of 4.7% over the next five years, a significant jump from the earlier projection of 2.6% ¹⁹. The EIA's Short-Term Energy Outlook also anticipates a 3% increase in total US electricity sales in 2025, driven by strong growth in both the residential and commercial sectors, with data center expansion being a key factor in the latter ¹². Furthermore, the EIA expects overall US electricity demand to increase by 2.3% in 2024 and 1.7% in 2025 compared to 2023 levels ²².

4.2. Adequacy of Current and Planned Generation Capacity

As of January 2024, the United States possessed nearly 1.3 million megawatts of electricity generation capacity ⁷. While significant additions to this capacity are planned, particularly in the realm of renewable energy, the adequacy of this growth to meet the rapidly escalating demand from AI remains a critical question. Solar energy is expected to account for the largest share of new capacity additions, with 51% of the most likely projects being solar, followed by wind at 33% ⁷. By the end of 2024, wind capacity is projected to reach 153.8 GW, and solar capacity is expected to climb to 128.2 GW ¹¹. However, the intermittent nature of solar and wind power presents a challenge in meeting the continuous and reliable power requirements of data centers. This may necessitate an increased reliance on natural gas-fired generation to ensure a stable power supply for these facilities ¹². There is also a possibility that planned shutdowns of existing coal-fired power plants could be delayed to further safeguard grid reliability in the face of rising demand ²⁴.

4.3. Looming Infrastructure Bottlenecks

The existing electricity grid infrastructure in the United States, particularly the transmission network, may face significant strain in accommodating the rapid increase in demand driven by AI and the need to distribute power from new generation sources. Regions with a high concentration of data centers, such as the "Data Center Alley" in Northern Virginia, are already experiencing limitations in their power infrastructure's ability to support further expansion ⁹. Expanding transmission capacity and upgrading the grid are complex processes that can face delays due to permitting challenges, environmental reviews, and logistical hurdles ⁹. Additionally, new generation projects often face lengthy interconnection queues, which can postpone their ability to deliver power to the grid ⁹. The regional concentration of data centers further exacerbates these issues, potentially leading to localized power shortages or grid constraints in areas with high AI infrastructure density ¹⁸. These infrastructure limitations could hinder the ability to deliver sufficient power to meet the growing demands of the AI sector, even if adequate generation capacity exists or is planned.

5. Long-Term Energy Demand and Supply Outlook (Beyond 2030)

5.1. Sustained Growth in Energy Demand

Looking beyond 2030, the trajectory of energy demand in the United States remains subject to considerable uncertainty, particularly concerning the long-term growth rate of AI and other emerging technologies ³⁰. However, most projections anticipate a continued significant increase in energy demand, fueled by the ongoing expansion of AI applications, the development of increasingly sophisticated AI models, and the broader trend of electrification across various sectors of the economy, including transportation, heating, and industry ¹. It is also conceivable that entirely new and unforeseen energy demands could arise from future technological breakthroughs ³⁰. While advancements in energy efficiency, both within AI operations and in other sectors, have the potential to moderate this growth to some extent ², the overall outlook suggests a substantial net increase in long-term energy consumption.

5.2. The Scalability of Clean Energy Solutions

5.2.1. Renewable Energy Dominance

In the long term, renewable energy sources, particularly solar and wind power, possess the potential to become the dominant sources of electricity generation in the United States. This shift is driven by their decreasing costs, increasing policy support, and the growing urgency to decarbonize the energy sector ³. However, realizing this potential will require a massive and accelerated deployment of these technologies, coupled with significant advancements and expansion in energy storage solutions to address their inherent intermittency. Robust energy storage is crucial for ensuring grid reliability and meeting the continuous power demands of AI infrastructure ⁹. Ongoing innovation in battery technology, as well as the development and deployment of other storage solutions like pumped hydropower and thermal storage, will be essential in this transition ⁹.

5.2.2. Nuclear Energy's Resurgence

Nuclear energy is also poised for a potential resurgence in the long-term US energy mix. Its capacity to provide reliable, carbon-free baseload electricity makes it particularly valuable for meeting the continuous energy needs of AI data centers and supporting overall grid stability ³⁵. The US Department of Energy has set ambitious targets to potentially triple nuclear energy capacity by 2050 ³⁵. This expansion is expected to involve both traditional large-scale nuclear reactors and the deployment of innovative Small Modular Reactors (SMRs), which offer greater flexibility and potentially lower upfront costs ³⁵. Notably, there is growing interest from technology companies in utilizing nuclear energy as a dedicated power source for their energy-intensive operations ²⁸.

5.2.3. The Promise of Geothermal Energy

Advanced geothermal systems, including Enhanced Geothermal Systems (EGS) and Advanced Geothermal Systems (AGS), represent a vast and largely untapped source of clean energy in the United States. These technologies hold the potential to provide a significant and consistent supply of power across a wider range of geographic locations compared to traditional hydrothermal geothermal resources ⁴². Advancements in drilling technologies, including techniques honed by the oil and gas industry, are making EGS/AGS more accessible and economically viable ². Geothermal energy boasts a high capacity factor, meaning it can generate power consistently, making it an excellent complement to intermittent renewable sources and a reliable option for powering data centers ⁴². Some analyses suggest that geothermal energy could economically meet a substantial portion of the expected data center demand growth ⁴².

6. The Role of Renewable Energy and Sustainability

6.1. Scaling Renewables for a Sustainable Future

The United States has witnessed record growth in the deployment of renewable energy sources, particularly solar and wind power ⁹. To achieve a truly sustainable energy future while meeting the escalating demands of AI, it will be crucial to significantly accelerate the pace at which these renewable technologies are deployed ⁹. This endeavor will necessitate addressing various challenges, including those related to land use, the streamlining of permitting processes, the strengthening of supply chains, and the efficient interconnection of renewable energy projects to the grid ³. The ongoing advancements in renewable energy technologies and their steadily decreasing costs offer a promising outlook for their widespread adoption ¹⁰.

6.2. The Indispensable Role of Energy Storage

The large-scale integration of intermittent renewable energy sources like solar and wind into the power grid hinges on the availability and effectiveness of robust energy storage solutions ⁹. Significant progress has been made in the development and decreasing costs of battery storage technologies ⁹. Additionally, other storage methods, such as pumped hydropower and thermal storage, offer valuable contributions to grid stability ¹⁰. Energy storage plays a vital role in ensuring a reliable power supply, particularly for energy-intensive applications like AI data centers that require continuous operation ⁹.

6.3. Balancing AI's Energy Needs with Decarbonization Goals

The substantial energy demands of AI present a complex challenge to the national goals of decarbonizing the energy sector ³. There is a potential risk of increased reliance on natural gas and a delay in the planned retirement of coal-fired power plants to meet the immediate and growing power needs of AI infrastructure ²⁴. However, many major technology companies are demonstrating a strong commitment to powering their operations with clean energy and are making significant investments in renewable energy projects and other low-carbon technologies ²⁴. Ultimately, achieving a sustainable future will require a concerted effort to simultaneously advance AI innovation and accelerate the transition to a clean energy economy ¹⁵.

7. The Future of Nuclear and Geothermal Energy

7.1. Nuclear Power: A Reliable Low-Carbon Solution

The United States government has outlined an ambitious strategy to significantly expand its nuclear energy capacity, with a goal of potentially tripling it by 2050, aiming for an additional 200 gigawatts of new nuclear power ³⁵. This expansion is envisioned to include both traditional large-scale light-water reactors and the deployment of advanced reactor designs, most notably Small Modular Reactors (SMRs) and microreactors ³⁵. Nuclear energy offers a compelling solution due to its ability to provide dispatchable, carbon-free power, which is essential for ensuring grid stability and meeting the consistent energy demands of data centers and AI operations ³⁶. Existing nuclear power plant sites have the potential to host new reactor builds, and there is also consideration for constructing new nuclear facilities near retiring coal plants, leveraging existing infrastructure and workforce ³⁵. Realizing this ambitious expansion will necessitate advancements in manufacturing techniques, the establishment of robust supply chains (including for high-assay low-enriched uranium, or HALEU, fuel), and the streamlining of regulatory processes to facilitate the timely and cost-effective deployment of new nuclear capacity ³⁵.

7.2. Geothermal Energy: Tapping into Earth's Internal Heat

Geothermal energy represents a substantial and largely untapped resource with the potential to significantly contribute to the future US energy supply. Geothermal systems, such as Enhanced Geothermal Systems (EGS) and Advanced Geothermal Systems (AGS), hold particular promise for providing a consistent and dispatchable source of clean energy across a wider geographic area than traditional geothermal resources ⁴². Innovations in drilling technologies, including those developed by the oil and gas industry, are making EGS/AGS more accessible and economically viable ². Geothermal energy offers a high capacity factor, meaning it can generate power reliably for extended periods, making it an ideal complement to intermittent renewable sources like solar and wind and a dependable option for powering energy-intensive data centers ⁴². The US Department of Energy has set a roadmap with the goal of increasing geothermal capacity twentyfold by 2050, potentially supplying 10% of the nation's electricity ⁴⁵. Furthermore, analyses suggest that geothermal energy has the potential to economically meet a significant portion of the anticipated electricity demand growth from data centers ⁴².

8. AI as a Solution: Enhancing Energy Efficiency and Grid Management

8.1. AI for Optimizing Energy Consumption

Beyond being a significant consumer of energy, artificial intelligence also presents a powerful suite of tools for optimizing energy consumption across various sectors of the economy ¹⁵. In buildings, AI-powered smart control systems can optimize heating, ventilation, and air conditioning (HVAC) systems, automate architectural design processes for energy efficiency, and improve the accuracy and efficiency of energy audits ¹⁵. AI can also play a crucial role in reducing waste and improving efficiency in manufacturing and other industrial processes through predictive maintenance and optimized operations ¹⁵. In the transportation sector, AI can contribute to more energy-efficient logistics, optimized routing, and the design of more efficient vehicles ¹⁵. Studies suggest that AI could potentially reduce energy consumption in buildings by as much as 25% ³¹. By analyzing energy consumption patterns and identifying anomalies, AI can also help detect and prevent unnecessary energy use ³³.

8.2. AI for a Smarter and More Resilient Grid

Artificial intelligence offers transformative potential for modernizing and optimizing the US electricity grid, enhancing its efficiency, reliability, and ability to integrate increasing amounts of renewable energy ¹⁵. AI algorithms can improve grid planning by analyzing historical data to predict future energy demand and optimize the placement of new infrastructure ⁴⁸. Real-time monitoring of grid conditions, automated responses to dynamic changes, and predictive maintenance of grid equipment are also areas where AI can provide significant benefits ⁴⁸. Furthermore, AI can greatly enhance the forecasting of renewable energy production by analyzing vast datasets of weather patterns and historical data, enabling better scheduling and dispatch of power, including for energy-intensive data centers ¹⁵. AI can also optimize the operation of energy storage systems by predicting demand fluctuations and managing charging and discharging cycles more efficiently ²⁴. The implementation of dynamic pricing models and demand-side management strategies can also be facilitated by AI-powered systems ⁴⁸.

8.3. Reducing AI's Own Energy Footprint

The AI research and development community is actively engaged in efforts to mitigate the growing energy footprint of AI technologies themselves ¹⁵. This includes the development of more energy-efficient AI algorithms that can achieve the same or better results with less computational power. Advancements in specialized AI hardware, such as Graphics Processing Units (GPUs) and Tensor Processing Units (TPUs), offer significant performance improvements in AI tasks while often consuming less energy compared to traditional Central Processing Units (CPUs) ¹⁵. The concept of on-device AI processing, where AI tasks are performed locally on individual devices rather than relying on centralized data centers, holds significant promise for drastically reducing the overall energy demands associated with AI ⁵³. Additionally, there is a trend towards developing smaller, more specialized AI models that are tailored for specific tasks and require less computational resources than large, general-purpose models ¹⁵.

9. Policy and Investment Considerations

9.1. Policy Landscape and Its Relevance to AI Energy Needs

The current US energy policy landscape, including landmark legislation like the Inflation Reduction Act ³⁰, provides a foundation for addressing the energy demands of the AI era. These policies aim to incentivize the development and deployment of renewable energy sources, enhance energy storage capabilities, and modernize the nation's electricity grid ³. Policies also exist to support the growth of nuclear energy and geothermal energy, both of which could play crucial roles in meeting the increasing power demands, including those from AI ³⁵. Recognizing the specific challenges posed by the energy needs of the AI sector and data centers, emerging policy discussions are beginning to focus on targeted strategies to ensure a sustainable path forward ²⁶. Environmental regulations and climate change policies also have a significant impact on energy production and consumption patterns related to AI, pushing towards cleaner energy sources ³.

9.2. The Imperative for Strategic Investment

Meeting the escalating energy demands driven by AI will necessitate substantial and strategic investments across the entire energy sector ¹⁸. Significant capital will be required to expand and modernize the electricity grid, including both transmission and distribution infrastructure, to handle the increased load and facilitate the integration of new generation sources ⁹. Investments in new clean energy generation capacity, encompassing large-scale renewable projects, advanced nuclear power (including SMRs), and next-generation geothermal systems, will be crucial ⁹. Furthermore, sustained and increased funding for research and development in key areas such as energy storage technologies, AI energy efficiency, and grid optimization solutions will be essential to ensure long-term energy sufficiency and sustainability ¹⁰.

9.3. Policy Recommendations to Support a Sustainable AI Energy Future

To ensure a sustainable energy future in the age of AI, several policy recommendations warrant consideration. Policies should incentivize the use of renewable energy sources to power data centers and AI infrastructure through mechanisms such as tax credits, renewable energy mandates, and direct subsidies ³. Support for the development and deployment of energy storage solutions should be enhanced through tax incentives, grants, and streamlined permitting processes ⁹. Measures to facilitate the expansion and modernization of the electricity grid are crucial, including streamlining permitting for transmission projects and investing in smart grid technologies ¹⁹. Policies could also encourage energy efficiency in AI hardware and software development through research grants or energy efficiency standards ⁵³. Exploring the implementation of energy credit trading systems specifically for AI could incentivize energy-efficient practices within the sector ⁵³. Finally, continued policy support for the research, development, and deployment of advanced nuclear and geothermal energy technologies is essential ³⁵.

10. Conclusion

The analysis presented in this report underscores that the growing energy demands of artificial intelligence, particularly within the rapidly expanding network of data centers, are poised to significantly reshape the energy landscape of the United States, leading to a substantial increase in electricity consumption. While the nation currently benefits from a position as a net energy exporter, this advantage will be tested by the unprecedented growth in demand from the AI sector. The existing US energy production and consumption profile, characterized by a continued reliance on fossil fuels alongside a growing contribution from renewable sources, will need to evolve to meet these new challenges sustainably. The potential strain on the current energy infrastructure, especially the electricity grid, necessitates significant upgrades and strategic expansion to ensure reliable power delivery. Renewable energy, nuclear power, and advanced geothermal energy each offer crucial pathways for meeting the long-term energy demands of the AI era in a low-carbon manner. Furthermore, artificial intelligence itself presents a powerful toolset for enhancing energy efficiency across various sectors and for optimizing the management of the electricity grid. Ultimately, ensuring energy sufficiency in the age of AI will require proactive and well-informed policy decisions, substantial and sustained investments in energy infrastructure and clean energy technologies, and a continued commitment to innovation across both the energy and technology sectors to navigate this evolving landscape effectively and sustainably.

The Innovator’s Journal

Discussion about this post