“The narrative is a net impact equation that only includes the positive use cases of AI as compared to the operational impacts, which we believe is apples to oranges,” Holly Alpine, co-founder of the Enabled Emissions Campaign, told me. “We need to expand that conversation and include the negative applications in that scoreboard.”
Alpine founded the campaign alongside her partner, Will Alpine, in February of last year, with the goal of holding tech giants accountable for the ways users leverage their products to accelerate fossil fuel production. Both formerly worked for Microsoft on sustainability initiatives related to data centers and AI, but quit after what they told me amounted to a string of unfulfilled promises by the company and a realization that internal pressure alone couldn’t move the needle as far as they’d hoped.
While at Microsoft, they were dismayed to learn that the company had contracts for its cloud services and suite of AI tools with some of the largest fossil fuel corporations in the world — including ExxonMobil, Chevron, and Shell — and that the partnerships were formed with the explicit intent to expand oil and gas production. Other hyperscalers such as Google and Amazon have also formed similar cloud and AI service partnerships with oil and gas giants, though Google burnished its sustainability bona fides in 2020 by announcing that it would no longer build custom AI tools for the fossil fuel industry. (In response to my request for comment, Microsoft directed me to its energy principles, which were written in 2022, while the Alpines were still with the company, and to its 2025 sustainability report. Neither addresses the Alpines’ concerns directly, which is perhaps telling in its own right.)
AI can help fossil fuel companies accelerate and expand fossil fuel production throughout all stages of the process, from exploration and reservoir modeling to predictive maintenance, transport and logistics optimization, demand forecasting, and revenue modeling. And while partnerships with AI hyperscalers can be extremely beneficial, oil and gas companies are also building out their own AI-focused teams and capabilities in-house.
“As a lot of the low-hanging fruit in the oil reserve space has been plucked, companies have been increasingly relying on things like fracking and offshore drilling to stay competitive,” Will told me. “So using AI is now allowing those operations to continue in a way that they previously could not.”
Exxon, for example, boasts on its website that it’s “the first in our industry to leverage autonomous drilling in deep water,” thanks to its AI-powered systems that can determine drilling parameters and control the whole process sans human intervention. Likewise, BP notes that its "Optimization Genie” AI tool has helped it increase production by about 2,000 oil-equivalent barrels per day in the Gulf of Mexico, and that between 2022 and 2024, AI and advanced analytics allowed the company to increase production by 4% overall.
In general, however, the degree to which AI-enabled systems help expand production is not something companies speak about publicly. For instance, when Microsoft inked a contract with Exxon six years ago, it predicted that its suite of digital products would enable the oil giant to grow production in the Permian Basin by up to 50,000 barrels by 2025. And while output in the Permian has boomed, it’s unclear how much Microsoft is to thank for that as neither company has released any figures.
Either way, many of the climate impacts of using AI for oil and gas production are likely to go unquantified. That’s because the so-called “enabled emissions” from the tech sector are not captured by the standard emissions accounting framework, which categorizes direct emissions from a company’s operations as scope 1, indirect emissions from the generation of purchased energy as scope 2, and all other emissions across the value chain as scope 3. So while tailpipe emissions, for example, would fall into Exxon’s scope 3 bucket — thus requiring disclosure — they’re outside Microsoft’s reporting boundaries.
According to the Alpines’ calculations, though, Microsoft’s deal with Exxon plus another contract with Chevron totalled “over 300% of Microsoft’s entire carbon footprint, including data centers.” So it’s really no surprise that hyperscalers have largely fallen silent when it comes to citing specific numbers, given the history of employee blowback and media furor over the friction between tech companies’ sustainability targets and their fossil fuel contracts.
As such, the tech industry often ends up wrapping these deals in broad language highlighting operational efficiency, digital transformation, and even sustainability benefits —- think waste reduction and decreasing methane leakage rates — while glossing over the fact that at their core, these partnerships are primarily designed to increase oil and gas output.
While none of the fossil fuel companies I contacted — Chevron, Exxon, Shell, and BP — replied to my inquiries about the ways they’re leveraging AI, earnings calls and published corporate materials make it clear that the industry is ready to utilize the technology to its fullest extent.
“We’re looking to leverage knowledge in a different way than we have in the past,” Shell CEO Wael Sawan said on the company’s Q2 earnings call last year, citing AI as one of the tools that he sees as integral to “transform the culture of the company to one that is able to outcompete in the coming years.”
Shell has partnered since 2018 with the enterprise software company C3.ai on AI applications such as predictive maintenance, equipment monitoring, and asset optimization, the latter of which has helped the company increase liquid natural gas production by 1% to 2%. C3.ai CEO Tom Siebel was vague on the company’s 2025 Q1 earnings call, but said that Shell estimates that the partnership has “generated annual benefit to Shell of $2 billion.”
In terms of AI’s ability to get more oil and gas out of the ground, “it’s like getting a Kuwait online,” Rakesh Jaggi, who leads the digital efforts at the oil-services giant SLB, told Barron’s magazine. Kuwait is the third largest crude oil producer in OPEC, producing about 2.9 million barrels per day.
Some oil and gas giants were initially reluctant to get fully aboard the AI hype train — even Exxon CEO Darren Woods noted on the company’s 2024 Q3 earnings call that the oil giant doesn’t “like jumping on bandwagons.” Yet he still sees “good potential” for AI to be a “part of the equation” when it comes to the company’s ambition to slash $15 billion in costs by 2027.
Chevron is similarly looking to AI to cut costs. As the company’s Chief Financial Officer Eimear Bonner explained during its 2024 Q4 earnings call, AI could help Chevron save $2 to $3 billion over the next few years as the company looks towards “using technology to do work completely differently.” Meanwhile, Saudi Aramco’s CEO Amin Nasser told Bloomberg that AI is a core reason it’s been able to keep production costs at $3 per barrel for the past 20 years, despite inflation and other headwinds in the sector.
Of course, it should come as no surprise that fossil fuel companies are taking advantage of the vast opportunities that AI provides. After all, the investors and shareholders these companies are ultimately beholden to would likely revolt if they thought their fiduciaries had failed to capitalize on such an enormous technological breakthrough.
The Alpines are well aware that this is the world we live in, and that we’re not going to overthrow capitalism anytime soon. Right now, they told me they’re primarily running a two-person “awareness campaign,” as the general public and sometimes even former colleagues are largely in the dark when it comes to how AI is being used to boost oil and gas production. While Will said they’re “staying small and lean” for now while they fundraise, the campaign has support from a number of allies including the consumer rights group Public Citizen, the tech worker group Amazon Employees for Climate Justice, and the NGO Friends of the Earth.
In the medium term, they’re looking toward policy shifts that would require more disclosure and regulation around AI’s potential for harm in the energy sector. “The only way we believe to really achieve deep change is to raise the floor at an international or national policy level,” Will told me. As an example, he pointed to the EU’s comprehensive regulations that categorize AI use cases by risk level, which then determines the rules these systems are subject to. Police use of facial recognition is considered high risk, for example, while AI spam filters are low risk. Right now, energy sector applications are not categorized as risky at all.
“What we would advocate for would be that AI use in the energy sector falls under a high risk classification system due to its risk for human harm. And then it would go through a governance process, ideally that would align with climate science targets,” Will told me. “So you could use that to uplift positive applications like AI for methane leak detection, but AI for upstream scenarios should be subject to additional scrutiny.”
And realistically, there’s no chance of something like this being implemented in the U.S. under Trump, let alone somewhere like Saudi Arabia. And even if such regulations were eventually enacted in some countries, energy markets are global, meaning governments around the world would ultimately need to align on risk mitigation strategies for reigning in AI’s potential for climate harm.
As Will told me, “that would be a massive uphill battle, but we think it’s one that’s worth fighting.”
Proportion of species classified in risk areas
