Double click here to edit new New Header component
Investment Bank
How can we help?
X
  • Contact Us
  • Client Login
    • Search
    Parsys 1
    Double click here to edit Hero Media component
    RESEARCH  |  3 POINT PERSPECTIVE  |  SOLVING SUSTAINABLE

    Artificial Intelligence is hungry for power

    authorImage authorImage 
    Will Thompson & Betty Jiang
    28 Aug 2024

    The rapacious growth of Artificial Intelligence (AI) is expected to cause a dramatic increase in demand for energy and competition for land with access to utilities in the US. Our Research analysts consider these developments, and the knock-on effects on the grid, consumers and corporate sustainability strategies.

    Client log-in for full report
    Series Image
    AI data centre using electricity energy
    Parsys 2
    Double click here to edit Go To Section component

    GO TO SECTION

    Parsys 3

    1. AI's demand for power is growing

    After decades of almost non-existent demand growth for electricity in the US, the AI revolution is expected to more than double data centre electricity needs by 2030 based on current grid capacity, according to Barclays Research1.

    Unlike other industries or energy-consuming activities which place fluctuating requirements on the grid depending on the time of day or year that can be managed to maintain overall reliability and capacity, AI operations are an ‘always on’ demand. Data centres must operate continuously, 24/7/365, to function for users. In effect, AI energy demand can be considered a constant peak that leads to higher overall peak power demand across the grid.

    Based on analysis of available disclosures from technology companies, public data centre providers and utilities, and data from the EIA, our Research team estimates that data centres account for 3.5% of US electricity consumption today, and data centre electricity use could be above 5.5% in 2027 and more than 9% by 2030.

    Related content

    AI revolution: productivity boom and beyond

    Our Research analysts explore how recent breakthroughs in Artificial Intelligence could provide a boost to productivity, similar to past periods of revolutionary technology change.

    While new processing chips may drive more efficient energy consumption, efficiency gains alone cannot offset the energy demand created by the computing power required to run AI’s increasingly complex large language models (LLMs) and training data sets. In addition, as efficiency increases, it is expected that use of AI will grow, in turn resulting in higher electricity demand.

    2. Location, location, location

    Retrofitting conventional data centres for AI is often cost-prohibitive as they typically lack scale, access to sufficient power to handle increasing rack energy density and cannot accommodate advanced cooling technologies. With tech firms adopting a trend for building large campuses, tech firms and developers of new generation data centres are looking beyond mature geographic markets.

    Northern Virginia, which currently accounts for over 20% of US data centre capacity, will continue to dominate as demand for AI grows. However, focus for future development is turning to more remote areas such as the Midwest, which do not have the same land costs, NIMBY (‘Not in my backyard’) pressures and regulatory constraints as more established data centre markets.

    Data centre developers are prioritising land with access to untapped power sources, water, workforces, and favourable regulation. The possibility of hooking data centres up directly to existing nuclear power stations ‘behind the meter’ are being explored, but has raised concerns for other consumers’ power costs, reliability and emissions.

    Established technology hyperscalers are likely to have a first mover advantage when it comes to locating and constructing new data centres in more remote areas. But as reliable access to power and management of supply chain risks are key concerns, a deep knowledge of local markets and utilities infrastructure can be just as important as speed.

    3. Sustainability appears to be a lower priority

    With the current focus on building data centre capacity that prioritises secure access to power over specific fuel type, meeting rising electricity demand while lowering emissions will likely be a monumental challenge for grid operators.  

    The intermittent nature of renewable energy is an obstacle, as AI requires a constant power supply and sources such as solar and wind cannot be consistently tapped around the clock. While there are a range of solar and wind projects in the interconnection queue, based on historical trends, only a fraction of them are likely to come in-service2.

    Nuclear energy offers greater reliability, but it is already working at around 96% capacity nationally3 and new reactors can’t be built fast enough to address the projected power supply shortfall.

    Therefore, a mix of renewable and non-renewable energy will be needed to satisfy AI’s hunger for power, at least in the short-to-medium term. Our Research team expects AI to be a long tailwind for natural gas demand, which is the most likely fuel to fill any gap left by renewable sources as coal continues to be phased out. Access to natural gas pipelines is now a major preference for data centre development4.     

    As a result, tech firms and data centre developers will need to factor this mix of energy sources into their sustainability strategies and be explicit about them in their disclosures. Similarly, investors will need to consider how these disclosures impact the sustainability goals of their portfolios.    

    Related content

    Can global energy evolve to achieve net zero by 2050?

    Adoption of renewables and lower-carbon fuels is rising, but progress to net zero is behind pace. Our Research analysts outline the challenging stages ahead.

    1EIA, Barclays Research
    2PJM Monitoring Analytics, Barclays Research
    3PJM Monitoring Analytics, Barclays Research
    4Barclays Research

    Parsys 4
    Double click here to edit Banner component
    Double click here to edit Experts Panel Card component

    About the experts

    Will Thompson

    Senior Thematic Research Analyst

    Will Thompson is a Director on the Thematic Investing Research team and focuses on the energy and industrial sectors. Will joined Barclays Equity Research in 2014 and was most recently a senior analyst covering Small and Medium Cap Oil & Gas Exploration & Production (E&P) companies. Prior to that, he was part of the Oilfield Services & Equipment research team, where he ranked as an Institutional Investor Rising Star in 2017. Will previously worked in equity research in various other investment banks, covering Oilfield Services & Equipment and Airfreight, Surface, and Marine Transportation. He also spent time as a Senior Research Associate at an investment firm. Will has a Bachelor of Engineering from Dartmouth College and a Bachelor of Arts in Chemistry from Colby College.

    Betty Jiang

    Senior Energy Research Analyst

    Betty Jiang is a Managing Director and senior Equity Research analyst covering the U.S. Integrated Oil & E&P industry, based in New York. Prior to joining Barclays in 2023, she headed U.S. ESG Research for Credit Suisse and was responsible for sustainability research with a focus on energy transition. She was recognized in the Institutional Investor's All-America Research survey in the ESG category in 2021. Before that, Ms. Jiang was an equity research analyst covering the E&P space for over a decade, both on the sell side and the buy side. Her team was consistently ranked in the Top 3 by Institutional Investor. Ms. Jiang holds a B.S. in Chemical Engineering from Cornell University and is a CFA charterholder.

    Important content disclosuresImportant content disclosures
    Parsys 5
    iParsys for Double Pixel component