Back Research Notes Chevron and Exxon Aren’t Oil Giants Anymore. They’re Electron Giants for the AI Age Published on June 17, 2025 By Jordi Visser Over the last month, I have been searching for non-consensus ideas around the next phase of AI driven by inference demand through the rise of AI agents and the embodiment of AI, which to me marks a massive macro regime inflection point for investments. A macro regime shift allows for fresh ideas in areas that investors have not been focused on during the prior regime. As Marc Andreessen recently said at the Reagan Foundation Economic Forum, the emergence of general-purpose robotics what he refers to as the “rise of embodiment” will likely become the largest industry in the history of the world. Embodiment is not just about humanoids; it is about phones, AI agents (digital employees), robots, and self-driving cars. I covered this in my first deep-dive research piece for 22V, From Cloud LLMs to Embodied AI: The Next Hardware Investment Frontier and Macro Regime Shift . Last week was a pivotal week in my opinion for where markets stand on the regime shift. I listened to the Lex Fridman podcast featuring an interview with Google CEO Sundar Pichai. In that interview, he mentioned that at the recent Google I/O Developer event, he said, “The world is adopting AI faster than ever before,” noting that AI usage across Google’s products jumped from 9.7 trillion tokens in April 2024 to 480 trillion in April 2025 a 50× increase year-on-year. Anything growing at a pace of 50x is worth paying attention to as an investor. Later that same day, I listened to Oracle founder Larry Ellison echo the trend on the company’s earnings call, calling AI demand “astronomical” and “insatiable,” and sharing that a customer had said, “We’ll take all the capacity you have, wherever it is. We’ll buy every megawatt you can deliver, anywhere in the world.” This year has left many investors doubting the AI trade. It started with DeepSeek, followed by reports around Microsoft canceling data center leases, and then the tariffs. I highlighted in this week’s video that the token growth Sundar Pichai referenced ironically began to surge in the weeks after the DeepSeek release as AI demand accelerated, inference demand expanded, and we began the next phase of AI agents and embodiment. These eye-opening statements by two of the leaders in AI demand underscore the foundational shift now officially underway: artificial intelligence is catalyzing a new supercycle in electricity demand one that traditional infrastructure is not equipped to serve. I explored this in greater depth in my second deep-dive piece, Electrify and Amplify: How Bits and Bytes Drive Demand for Watts and Volts . To highlight the supply-demand imbalance, Google, Microsoft, Amazon, and Meta have each made aggressive moves to secure energy for their AI expansion. Microsoft signed a first-of-its-kind agreement with Helion for fusion power, a landmark bet on a future energy source. Amazon, through a deal with Talen Energy, acquired a data center campus located directly next to a nuclear power plant, securing a source for 24/7 carbon-free power. Google has also pursued next-generation clean energy, signing a geothermal energy deal with Fervo and actively exploring advanced nuclear solutions. The urgency is clear across the sector as companies work to secure the massive power blocks needed for AI growth. This wave of hyperscaler investment in next-gen energy sources comes with a crucial caveat: these solutions are not arriving fast enough. Most new advanced nuclear projects won’t begin delivering power until after 2030, and the intermittency of wind and solar cannot alone meet the 24/7 power needs of data centers. As AI workloads become increasingly power-intensive, the market requires dispatchable, scalable, low-carbon power and it needs it soon. As AI becomes embodied, moving off the cloud and into physical space through robots, humanoids, autonomous vehicles, smart appliances, and wearable agent, the nature of electricity demand will evolve dramatically. Unlike traditional data centers that are concentrated and centrally planned, embodied AI will generate a vast, decentralized layer of 24/7 compute demand embedded in warehouses, hospitals, homes, vehicles, and factories. These edge applications won’t tolerate latency or downtime, making local, dispatchable power essential. This decentralization of AI compute creates a mosaic of new electricity demand zones, many in areas where solar and nuclear may not be feasible at scale or where grid interconnection lags. While advanced nuclear technologies offer promise for clean baseload power, the timeline for significant deployment remains beyond 2030. In contrast, natural gas offers a ready, scalable, and geographically flexible solution to meet this immediate surge. Chevron and ExxonMobil, with their unmatched natural gas reserves, pipeline infrastructure, and increasingly modular gas-to-power deployments, are uniquely positioned to meet this dispersed, real-time energy need. In this sense, the embodiment of AI doesn’t just increase the quantity of power required, it transforms where, how fast, and how reliably that power must be delivered. This is where I arrive in the long-forgotten historically cyclical world of Chevron and ExxonMobil who have emerged as crucial players in the bridge to this AI transition. Notably, they are now among the two largest power producers in the United States, a fact often overlooked by generalist investors still thinking of them strictly as oil companies. Their pivot over the last decade toward natural gas-fired power plants, integrated with carbon capture and storage (CCS), offers a scalable near-term solution to benefit from the power crunch. If electrons are the essential fuel for the exponential rise in AI-driven electricity demand, and the system faces a structural shortfall, then Chevron and ExxonMobil are uniquely positioned to capture this opportunity. In a world where electricity becomes the limiting factor to AI growth, their role as energy providers becomes central to the entire AI supply chain. Chevron’s efforts are among the most advanced. In partnership with GE Vernova and Engine No.1, it is developing a strategy to build new gas-fired power capacity specifically targeting data centers. Confirming the strategy, Chevron’s Vice President of New Energies, Daniel Droog, stated, “We’re actually developing a business to provide power to data centers… We believe that natural gas with or without carbon capture is a great solution for that.” CEO Mike Wirth said, “These hyperscalers… need off‑grid power to reach the market quickly… Our approach is to build a different kind of value proposition for a different kind of a customer… gigawatt‑scale generation… Speed to market is an important differentiator.” On Bloomberg Surveillance, Wirth said: “There’s a tremendous amount of interest and activity,” and emphasized that “Speed to market is an important differentiator.” CEO Wirth has placed this strategy within a broader context of global energy needs, emphasizing, “The world needs to be focused on a transition that addresses climate change and still delivers affordable, reliable energy.” This view aligns directly with serving the intensive, 24/7 reliability required by the AI industry. During the Gastech conference in September 2024, Wirth highlighted the role that the Permian Basin could play in powering data centers. “Natural gas will help power the rapid growth of artificial intelligence with its insatiable demand for reliable electricity,” Wirth said. “This means AI’s advance will depend not only on the design labs of Silicon Valley, but also on the gas fields of the Permian Basin.” Chevron’s recent remarks highlight the accelerating energy demands of the AI boom and reinforce the broader investment case. Jeff Gustavson, President of Chevron New Energies, told Axios in May 2025 that Chevron’s initiative to