Evan Lee Salim
This startling forecast, released in a new report by BloombergNEF, underscores the profound energy implications of the rapidly accelerating artificial intelligence revolution and the broader expansion of digital infrastructure. Over the coming decade, data centers are poised to become the second-strongest driver of natural gas demand growth in the United States, surpassed only by the burgeoning liquefied natural gas (LNG) export sector. The report projects that these digital behemoths could consume approximately 18 billion cubic feet per day (Bcf/d) of natural gas by 2035, a figure that nearly doubles BloombergNEF’s prediction from just nine months prior. This upward revision highlights the unprecedented pace at which AI development and data processing needs are escalating, demanding a re-evaluation of energy infrastructure, climate targets, and economic stability.
The Unprecedented Surge in Demand
The revised BloombergNEF forecast takes into account a more realistic assessment of data center project completion rates, acknowledging that not all announced initiatives will materialize. Despite this tempered outlook, the sheer scale of the projected consumption remains staggering. To put 18 Bcf/d into perspective, the U.S. Energy Information Administration (EIA) reported total U.S. natural gas consumption averaged around 88 Bcf/d in 2023 across all sectors (residential, commercial, industrial, and electric power). This means data centers alone could account for over 20% of the nation’s total natural gas demand within a little over a decade, a dramatic shift in the energy landscape.
The primary catalyst for this surge is the exponential growth in demand for high-performance computing, driven largely by the training and inference of large language models (LLMs) and other advanced AI applications. These processes require immense computational power, which translates directly into massive electricity consumption. Modern AI servers, often equipped with powerful graphics processing units (GPUs), can draw significantly more power than traditional central processing units (CPUs), increasing the power density within data centers and, consequently, their overall energy footprint.
Onsite Power Generation: A Growing Trend
A notable trend highlighted in the report is the increasing move by hyperscale tech companies towards developing data centers with their own dedicated, onsite natural gas power plants. Giants such as Meta, Microsoft, Google, and Amazon have all announced plans for such facilities, signaling a strategic shift away from relying solely on the existing grid infrastructure. These projects are motivated by a combination of factors: ensuring uninterrupted power supply, hedging against volatile grid electricity prices, and circumventing potential delays or limitations in grid expansion.
These onsite-powered data centers are projected to consume between 2.9 billion and 3.4 billion cubic feet per day of natural gas by 2035. This amount alone is roughly equivalent to the total natural gas consumed by all data centers today, a figure that includes both onsite generation and the natural gas used to generate power for grid-connected facilities. While this represents a significant portion of the new demand, BloombergNEF indicates that it will ultimately constitute only a fraction of the overall demand growth attributable to data centers. The bulk of the additional consumption will still originate from data centers connected to the broader electricity grid.
Grid-Connected Data Centers: The Larger Share
By the middle of the next decade, grid-connected data centers are predicted to drive an additional 15 Bcf/d of natural gas consumption by the power sector. This figure dwarfs the demand from onsite generation and underscores the immense pressure this growth will place on the national electricity infrastructure. To contextualize this impact, this 15 Bcf/d represents five times more demand growth through 2035 than from all other grid-connected sectors combined. This unparalleled demand surge will necessitate substantial investments in new power generation capacity, transmission lines, and distribution networks, predominantly relying on natural gas as the most readily available and scalable fuel source to meet such rapid load increases.
The implications for existing utility providers are profound. Many grids, particularly in regions experiencing rapid data center development, were not designed to accommodate such concentrated and fast-growing loads. Utilities face the challenge of rapidly expanding infrastructure, often navigating complex permitting processes and public opposition, all while trying to maintain grid reliability and manage costs for their broader customer base.
Economic Ripples: Price Volatility and Ratepayer Burden
If this stunning demand growth materializes as projected, it carries significant economic ramifications, particularly for natural gas prices. Much of the current data center buildout strategy, especially the reliance on natural gas for onsite power or grid supply, has been predicated on the relatively stable and historically low natural gas prices observed in recent years. However, analysts, such as those at Noreva, suggest this reliance might be a false hope.
The combined impact of the unprecedented data center boom and the escalating demand from U.S. LNG exports could create a perfect storm, pushing natural gas prices significantly higher. A substantial increase in gas prices would ripple through the energy sector, affecting not only the tech companies directly consuming gas but also every utility and industrial user reliant on natural gas. While hyperscale tech companies, with their robust balance sheets, might be able to absorb such a surge in energy costs, the burden on residential and commercial utility ratepayers could be severe. Increased natural gas prices translate directly into higher electricity bills, potentially exacerbating energy poverty and imposing financial strain on households and small businesses across the country.
Moreover, the capital expenditure required to build out new power generation and transmission infrastructure to support this demand will likely be passed on to ratepayers, further contributing to rising electricity costs. This creates a challenging scenario where the benefits of technological advancement are weighed against the economic viability for everyday consumers.
Environmental Crossroads: A Setback for Climate Goals
Beyond the economic concerns, the environmental implications of this projected natural gas consumption are arguably the most critical. The additional demand from data centers will generate an estimated 1 million metric tons more greenhouse gas (GHG) pollution daily. This figure alone is equivalent to approximately 12% of total U.S. greenhouse gas emissions today, a substantial increase that threatens to derail national and international climate targets.
Burning one cubic foot of natural gas releases the equivalent of 60 grams of carbon dioxide into the atmosphere, a calculation that includes emissions from extraction, processing, and distribution, according to the International Energy Agency (IEA). The scale of additional emissions from data centers presents a significant challenge to the U.S.’s commitments under the Paris Agreement and its ambitious goals for decarbonization.
Environmental advocacy groups have swiftly voiced concerns, arguing that the rapid expansion of fossil fuel-dependent data centers undermines efforts to transition to a clean energy economy. They point out that while many tech companies have public commitments to 100% renewable energy, the immediate and projected reliance on natural gas for new infrastructure risks locking in fossil fuel dependence for decades. This situation highlights a tension between the pursuit of technological advancement and the imperative of climate action.
Industry Responses and Challenges
The tech industry’s response to its escalating energy footprint is complex. While companies like Meta, Microsoft, Google, and Amazon are investing in onsite natural gas plants for reliability and cost control, they are also simultaneously investing in renewable energy projects and advocating for grid modernization. However, the pace of AI growth appears to be outstripping the current ability of renewable energy deployment and grid upgrades to keep up.
Utilities, on their part, are grappling with the unprecedented speed and scale of data center development requests. Planning and building new power plants and transmission lines typically take years, if not a decade, due to regulatory approvals, environmental assessments, and construction timelines. The accelerated timelines of data center deployments create immense pressure, often forcing utilities to prioritize readily available, dispatchable power sources like natural gas over slower-to-deploy renewables.
There is also an increasing focus within the industry on energy efficiency. Innovations in chip design, server cooling technologies (such as liquid immersion cooling), and data center architecture are being explored to reduce power consumption per unit of computation. However, these efficiency gains are often outpaced by the sheer increase in the volume and complexity of AI workloads, leading to a net increase in overall energy demand.
Navigating the Future: Policy, Innovation, and Sustainability
The BloombergNEF report serves as a critical warning, necessitating a multi-faceted approach to address the looming energy challenge posed by AI data centers.
Policy and Regulation: Policymakers may need to consider new regulations or incentives to guide data center development towards more sustainable energy sources. This could include mandates for a certain percentage of renewable energy for new facilities, streamlined permitting for renewable energy projects that directly serve data centers, or energy efficiency standards for data center operations. Local governments are also starting to scrutinize data center proposals more closely, balancing economic development benefits against environmental and grid impacts.
Technological Innovation: Continued investment in energy-efficient AI hardware and software is crucial. Research into novel computing paradigms, such as neuromorphic computing, and alternative energy sources, including advanced nuclear technologies like Small Modular Reactors (SMRs), could offer longer-term solutions. Improved grid-scale energy storage solutions are also vital to integrate more intermittent renewable energy sources effectively.
Sustainable Siting and Design: Strategic siting of data centers in regions with abundant renewable energy potential or robust, underutilized grid capacity can mitigate some impacts. Designing data centers to capture and reuse waste heat for district heating or other industrial processes could also improve overall energy efficiency.
The projected energy consumption of U.S. data centers by 2035 represents a significant inflection point, highlighting the tension between rapid technological advancement and sustainable energy practices. As the AI race continues its frenetic pace, the decisions made today regarding energy infrastructure and policy will profoundly shape not only the future of technology but also the trajectory of global climate efforts and economic stability. Addressing this challenge requires urgent, coordinated action from tech giants, energy providers, policymakers, and communities to ensure that the digital future is both powerful and sustainable.
