For about a decade, the upstream industry has anticipated the rise of artificial intelligence (AI) and its potential to reshape oil and gas operations. The big promise was a new era of field optimization. Far less discussed was the possibility that AI could also affect the upstream business from the demand side. That is clearly no longer the case as the buildout of data centers needed to further the commercialization of AI takes shape. The US is widely considered to represent the center of this activity, which many have dubbed the “AI revolution. ” In particular, the impact of this technological shift is becoming visible in Texas and the US northeast, where data centers are being developed at a rapid pace. The change is easiest to measure through the electricity demand that these facilities are projected to generate in the coming years. In March, the US Energy Information Administration (EIA) noted that since 2020, annual US electric demand growth averaged about 1. 7%, up from just 0. 1% annually from 2005 to 2019. A key inflection point appears to be the introduction and widespread adoption of large language models (LLMs) such as ChatGPT and Claude, which debuted in 2022 and 2023, respectively. To this point, the EIA cites electricity use by data centers needed to run and develop LLM technology as the main driver of this demand surge. But this year, the pace will pick up even more. The EIA said growth is now expected to accelerate to 1. 9% in 2026 and 2. 5% by 2027. The growth becomes more pronounced at the regional level. In Texas, annual electricity load demand is projected to increase by about 10% from 2025 levels by 2027. That’s the EIA’s baseline prediction. In a high-demand scenario, it said demand growth in Texas could reach 15% by 2027. In the regional grid covering Washington, DC, and all or parts of 13 states (Delaware, Illinois, Indiana, Kentucky, Maryland, Michigan, New Jersey, North Carolina, Ohio, Pennsylvania, Tennessee, Virginia, and West Virginia), demand is expected to grow by about 3. 2%. Modest compared with Texas, but still nearly double the national average seen in recent years. The EIA has issued a caution, too. If demand rises too fast, outstripping supply, the US government agency said, “the stresses on the grid would be evident in spikes in wholesale power prices or even periods of rolling blackouts. ” With this topic moving further into the public consciousness, in early March, US President Donald Trump went so far as to get the commitment of seven of the country’s largest tech firms (Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI) to sign a commitment that they will cover the costs of all the new power generation to protect consumers from rising electricity prices. The White House added that it wants to ensure AI companies also become power companies. So, where does this leave oil and gas companies? The answer appears to favor producers with large natural gas portfolios. The EIA predicts that much of the additional power required to support the AI boom will come from greater use of natural gas-fired power plants, which already account for about 40% of US electricity generation. It added that all this growth will lead to an estimated 1. 7% increase in gas-fired power generation from 2025 to 2027. In its higher-demand scenario, the EIA said this figure rises to 7. 3%. Among the upstream companies responding to this trend is Permian Basin supermajor Chevron, which last year announced a partnership with GE Vernova and investment group Engine No. 1 to build “behind-the-meter” power plants. The dedicated facilities, which will operate outside the grid, are designed to supply up to 4 GW of electricity to US data centers. Chevron said the GE-built gas turbines are expected to begin operating in 2027. Chevron also notes on its website that data centers could consume as much as 9% of all US electricity by 2030, which would be a 350% increase from their share in 2024. Various analyses peg data center-driven natural gas demand by 2030 in the range of 3 to 6 Bcf/D, equivalent to roughly 3 to 6% of total US gas demand today. US utility provider NextEra Energy announced in December that it was building a 1. 2 GW power plant dedicated for data center use and that it is working with ExxonMobil to secure carbon capture and storage capabilities for the 2, 500-acre site. Also announced at the end of last year was an agreement between Halliburton and VoltaGrid, in which the service company holds a 20% stake, to develop power-generation systems for data centers on a global scale, beginning with facilities in the Middle East. The project would involve building and operating gas turbines and reciprocating engines to support large-scale data center expansion. In March, the largest natural gas producer in the US, EQT Corp. , participated in a private-equity-led deal to acquire power company AES Corp. for 33. 4 billion. The deal was widely viewed by markets as a major bet on rising electricity demand from US data centers. EQT’s involvement in the sector does not end there. One of the Pittsburgh-based company’s equity investments, Scale, acquired Reload in February. Reload specializes in acquiring and developing land for data centers. The capital EQT is bringing to the venture is expected to support multi-gigawatt power generation for Scale’s off-grid data center business. However, the rush to enter the data center business is also creating potential supply chain bottlenecks. A recent research note from Rystad Energy said the surge in demand has in some cases inflated prices for power-generation equipment by as much as 300%, while lead times for high-efficiency gas turbines are stretching to 5 years. Nevertheless, with AI on the fast track to becoming a central element of the modern economy and global business landscape, oil and gas companies appear to be joining the same race as AI firms to become power companies. And perhaps this development could also offer oil companies in the Permian another pathway to address flaring, which is one of the unconventional sector’s most stubborn and lingering challenges.
Trent Jacobs (Wed,) studied this question.