US oil and gas majors remain bullish on their move to power the artificial intelligence (AI) data center market despite the launch of large language models from Chinese lab DeepSeek.
In the last three months, ExxonMobil and Chevron, the two largest oil and gas companies in the US, announced plans to develop gas-fired power plants to serve the data center market, citing AI power demand as the primary driving force.
However, the launch of DeepSeek's technology has had significant reverberations across the market. The lab's AI models claim to offer similar performance to US models - such as OpenAI's GPT-4-based systems and Google's Gemini - but with much lower energy consumption levels, potentially reducing the need for expensive, energy-hungry data centers.
Despite this, Chevron CEO Mike Wirth believes that the launch of DeepSeek doesn't threaten the US sector as many claim.
“DeepSeek actually underscores how competitive global and urgent the race for AI leadership is,” Wirth said in an interview with Bloomberg. “We will see the use of these models proliferate across the economy. Demand for AI, the demand for power will grow and reflect that.”
Late last month, Chevron partnered with Engine No.1 to develop up to 4GW of gas-fired generation to serve co-located data centers in the Southeast, Midwest, and West US.
In Exxon's Q4 results call, CEO Darren Woods claimed the company is “well positioned to meet surging demand from data centers for low-carbon power and on a timetable that alternatives such as nuclear simply can't match.”
At the end of last year, Exxon revealed it was developing a 1.5GW natural gas-fired power plant to supply the data center market. Exxon plans to leverage its pipeline portfolio to supply natural gas to the plant and use carbon capture and storage technology to store emissions from the plant.
Exxon and Chevron clarified that their foray into the power sector won’t transform them into traditional utilities. Instead, they intend to offer more bespoke power solutions. Chevron plans to foster partnerships tailored to the specific demands of hyperscalers, such as Meta and Amazon Web Services, while Exxon seeks to provide “low-carbon” power, utilizing its CCS operations in tandem with natural gas-fired generation.
Woods revealed that Exxon is in talks with several data center firms to implement CCS, claiming the first data center deployed with CCS could go live in 2028.
"Our view is we'll bring this on faster than anybody else in the industry, and we'll certainly bring it on faster than any other opportunities for decarbonization," said Woods.
CCS is a controversial technology. While many—especially within the oil and gas world—endorse it as an effective means of decarbonizing fossil fuel production, critics contend that it is expensive, poses environmental risks, and has a limited impact on carbon reduction.
Late last year, Meta became one of the first hyperscalers to announce a data center powered directly by onsite natural gas generation. The company inked an agreement with Entergy Louisiana to construct three combined-cycle combustion turbines with a total capacity of 2,260MW to power the 2GW data center.
The growth of the AI sector led global market intelligence agency S&P Global to project last year that demand for natural gas to support data centers could reach up to three billion cubic feet per day (bcf/d) by 2030 on the low end and as high as six bcf/d on the high end.
Speaking with DCD, Michael Grande, midstream energy and refining managing director, S&P Global Ratings, said: "AI has driven demand for data centers, with corporations increasingly aware of the risks of relying on hyperscalers and centralized infrastructure."
This has led to increasing numbers of companies to consider on-site natural gas generation as a potential solution. "It’s faster to power data centers with gas than to wait for emerging clean energy technologies. Timing is key," Grande stated.
