Chevron is wasting no time in scouting locations for natural gas-fired power plants as the US major looks to have its first projects designed to meet soaring demand for artificial intelligence (AI) data centres up and running in the next few years.
"The first site is not yet finalised, but given that timeframe, we'll be making site selections over the next couple of weeks to months," Chevron vice-president of power solutions Daniel Droog says. "The speed to delivery is fast."
US data centre electricity demand is forecast to quadruple by the end of the decade to cater to the AI boom, and Big Oil is eager to grab a slice of the action by tapping into its prior experience in building power plants for its own operational needs. Chevron's proposed plants will be independent of the grid, at least initially. Having already secured turbine contracts in advance is a big plus, given that manufacturers warn of lengthy queues at a time of supply chain bottlenecks.
Chevron has been working on the concept for over a year, and three projects are seen in the first phase, with more likely in the future. The firm is focusing on the US for now, but in time it is likely to look at the international market.
At the start of this year, Chevron said it was looking to serve data centres in the southeastern, midcontinent and western regions after announcing a partnership with investment firm Engine No. 1 and GE Vernova, a spin-off from US electrical engineering giant General Electric, last year. The joint venture aims to deliver up to 4GW of electricity, the equivalent of powering 3mn-3.5mn households for a year.
The potential to twin the projects with carbon capture and renewables will be explored. "Some of them will have carbon capture as an option based on feasibility, and some of them potentially wouldn't, depending on where they're located," Droog said at the recent CERAWeek by S&P Global energy conference in Houston, Texas. That state is "very" attractive as a possible location, he added, given its "regulatory environment from the power and electrical side that potentially allows you to do both speed and scale". An abundant supply of natural gas that makes projects cost-effective in the long term and "robust" infrastructure are also appealing, he said.
Chevron has already lined up seven GE Vernova 7HA gas turbines — despite a tight market fuelled by the boom in AI data centres — under a slot reservation agreement on an accelerated timeline. "The turbine reservations and where we are in the delivery queue bring a significant advantage," Droog said. "That unlocks tremendous value for the customer."
DeepSeek and ye shall find
Forecasts for soaring AI-related power demand in the coming years were recently called into question after a breakthrough by Chinese AI company DeepSeek cast doubt over the energy-intensity of the technology's requirements. Chevron suggests its own assumptions take that uncertainty into account."Our view is very much long term through the cycle, like the rest of our businesses, and that this demand curve — or demand ramp — is significant in the near term, and then over time it will probably adjust," Droog said.
Engine No. 1 is best known for leading an activist shareholder revolt at ExxonMobil four years ago that ended up usurping a quarter of the US major's board. The investment firm brings "speed and hustle, if you will", Droog said, as well as "some additional insight and experience on financing long term". Chevron has built 5GW of power generation capacity to meet its own operational needs in remote locations where options were limited. "The 5GW that we've built has essentially the same redundancy reliability requirement that the hyperscalers have for their data centres," Droog said.
