ExxonMobil

XOM · Q2 2026 · reported 2026-07-31 · revenue $116.02bn

Assessment

ExxonMobil names AI for the first time in coverage in Q2 2026, in the prepared remarks and slides published with the call, on the call and in cautionary language, none of it with a dollar figure. Exploration is the clearest channel: the prepared remarks say a model trained on Guyana discovery data identified of known discoveries in validation testing and has found new Guyana prospects not found by traditional methods (claims c12 and c13), and the CEO confirms it on the call, adding that much more work is needed to confirm them (claims c2 and c3). A count of unconfirmed prospects sizes nothing and no revenue has started, so the channel is read directional and unsized, tagged relabelled because the anchor describes the same reservoir-model work without AI.

Two further channels name AI beside other causes and are read described and unsized. The CFO names AI machine learning beside extended reach laterals and surfactants as contributing to Permian performance (claim c5), and a Guyana slide credits the developments' pace and cost advantage to the project model, AI-enhanced drilling performance and partnerships together (claim c15); the prepared remarks credit the same Permian and Guyana results to technology and the project model without naming AI (claims c17 and c16). Both are tagged relabelled. The other AI mentions are kept as context: the CEO says the enterprise-wide process and data platform transformation will help accelerate the adoption and value of AI (claim c1), an aim with no tool or measure, and the 10-Q and the release add AI-enhanced technologies to the cautionary statement's list of factors for drilling performance and recovery (claims c6 and c7). Power for data centers, discussed with hyperscalers in Q1, is not mentioned in Q2.

The cost story of the quarter is not AI. Cumulative structural cost savings relative to 2019 reached , with added in the first half; the CEO credits the transformation and centralized organizations (claims c9 and c10) and the 10-Q defines the savings as operational efficiencies, workforce reductions and divestments (claim c11). A disruption cost avoided through advanced modeling and fleet changes is not attributed to AI either (claim c8). These are confounds, not savings. Revenue grew to on higher prices; selling, general and administrative expenses moved to , and exploration expenses changed by .

Cost displaced by AI1 channel · 1 not sized

operations · expensive to verify

Drilling pace and cost on Guyana developments with AI-enhanced drilling

Not sized

AI-enhanced drilling performance is named beside the design one, build many model and strategic partnerships, and nothing separates its part (the methodology rule for AI named beside another cause); the saving would also fall on capitalized development drilling, which stays unsized. The benchmark advantage, measured over projects from 2017 to 2025, is quoted as metrics, not as a size.

described, no sizedisclosure: described· motive: exploratory· before LLMs: relabelled

Opened this quarter, in the slides only. The Guyana slide credits the developments' pace and cost advantage to the design one, build many model, AI-enhanced drilling performance and strategic partnerships together; the prepared remarks give the advantage as projects faster and up to cheaper than the industry average and credit it to the project model without naming AI. Registered as a described, unsized joint-cause channel, by the same reading as the Permian remark. Read relabelled: the anchor describes drilling technology and project execution without AI.

Evidence: 2 quotes, 2 figures, 3 confounds, 2 from before coverage

Development cost and time avoided on the Guyana projects through AI-enhanced drilling performance. A Q2 2026 slide credits the Guyana developments' pace and cost advantage over a benchmark of industry projects to the design one, build many model, AI-enhanced drilling performance and strategic partnerships together; the prepared remarks credit the same advantage to the design one, build many approach and the Global Projects organization without naming AI. Nothing separates AI's part, and the saving falls on development drilling, which the company capitalizes under the successful efforts method, so the channel is described and unsized on both counts (joint cause; saving on capitalized work). In Q1 2026 the CEO credited a fully autonomous deepwater well section in Guyana to rig automation and automated downhole steering tools, with no AI named (claim xom-2026-cq1-c3). The cost would have gone to drilling contractors, service companies and the company's own staff, and it is shared with the Stabroek co-venturers, so counterparty mixed.

Why this motive

A slide names AI-enhanced drilling performance as one of three enablers of the Guyana pace and cost advantage (claim c15), and the measured advantage is credited in the prepared remarks to the project model without AI (claim c16); a result credited to several changes together does not meet the efficiency tell, and AI named as a cause with no measure of its own is the exploratory tell.

Before LLMs: relabelled

At the anchor the FY2024 10-K said the company must apply its project management expertise to bring discovered resources online as scheduled and within budget (anchor claim xom-anchor-c2) and described leading capabilities and technology in drilling and completions (anchor claim xom-anchor-c4), with no AI named. In coverage the drilling performance is called AI-enhanced beside the same project model, with no line, rate or cost shown to move because of AI, so the activity is read as renamed (tie-break).

“Our ability to maintain and grow our oil and gas production depends on the success of our exploration and development efforts. Among other factors, we must continuously improve our ability to identify the most promising resource prospects and apply our project management expertise to bring discovered resources online as scheduled and within budget.”

Filing, risk factors, 10-K periodic report, 2025-02-19

“ExxonMobil operations continue to deliver industry-leading capital efficiency and cost performance by leveraging scale, integration, and technology. Examples include deploying ExxonMobil cube design and proprietary proppant as well as leading capabilities and technology in drilling and completions.”

Filing, business, 10-K periodic report, 2025-02-19

Figures

Guyana and Global Projects delivery speed against the industry average, faster by more than, projects benchmarked 2017 to 2025 · 2017-01-01..2025-12-31
Guyana and Global Projects cost against the industry average, lower by up to, projects benchmarked 2017 to 2025 · 2017-01-01..2025-12-31

What else could explain it

other: The design one, build many model, the Global Projects organization and strategic partnerships are credited with the same advantage (claims c15 and c16); the prepared remarks give the speed of and the cost of against the industry average for the project model without naming AI.
other: In Q1 2026 the CEO credited an autonomous deepwater well section to rig automation and automated downhole steering tools, with no AI named (claim c3).
line composition: Development drilling is capitalized under the successful efforts method and shared with the Stabroek co-venturers; any saving would show in capital spending and later depreciation, not in a quarter's operating cost line.

Quotes

“Unprecedented pace and cost advantage enabled by 'design one, build many' model, AI-enhanced drilling performance, and strategic partnerships”

c15 · Filing, presentation, earnings release, 2026-07-31

“A key enabler has been our "design one, build many" approach, supported by the scale and expertise of our Global Projects organization. We standardize designs and carry lessons from one project to the next. That has helped us deliver projects more than 30% faster and at costs up to 20% below the industry average.”

c16 · Filing, presentation, earnings release, 2026-07-31

Revenue arriving through AI2 channels · 2 not sized

research · expensive to verify

Exploration prospects found by AI models trained on subsurface data

Not sized

The only measure of output is a count of unconfirmed prospects, which sizes nothing and makes the channel directional (methodology, counts); the validation share is a test of the model, not a share of any line. No prospect has been drilled and no date is given for drilling or production, so the money has not started; under the ballpark rules such a channel stays unsized with no estimate.

described, no sizedisclosure: direction only· motive: exploratory· before LLMs: relabelled

Opened this quarter. The prepared remarks published with the call say the company has built exploration models powered by AI and its seismic database; one model, trained on Guyana discovery data, identified of known discoveries in validation testing and has since identified new Guyana prospects not found by traditional methods, and the slides list the same opportunities as found through AI-powered exploration. This is the disclosure the analyst cites on the call. Answering him, the CEO relays the count from an earlier conference remark by a senior executive (), says much more work is needed to confirm the prospects, and repeats the point later in the call. Read relabelled: the anchor shows the same reservoir-model work, and no reserve, production, revenue or expense line moves because of AI.

Evidence: 6 quotes, 6 figures, 2 confounds, 3 from before coverage

Revenue the company could earn from oil and gas resources found because AI models, trained on what earlier drilling and subsurface characterization found, identify exploration prospects in the Guyana block that the company had not identified before. The Q2 2026 prepared remarks say the company has created exploration models powered by AI and its seismic database, that one model trained on Guyana discovery data identified most known discoveries in validation testing, and that it has identified new Guyana prospects not found by traditional methods; the slides repeat the point and the CEO confirms it on the call, adding that much more work is needed to confirm them. The only measure is a count of unconfirmed prospects, which sizes nothing (a count makes the channel directional), and no prospect has been drilled, so the money has not started and no date is given. The exploration work itself sits in exploration expenses, including dry holes, which the 10-Q does not explain by AI. Any revenue would come from crude and gas sold to outside refiners, traders and gas buyers, so counterparty enterprise; sales to the company's own refineries are eliminated in consolidation and are never a size, and the Government of Guyana's share of production is not company revenue. Read as relabelled under the tie-break: the anchor already describes the same work (updating reservoir models from drilling and seismic data to look for new opportunities in the block), and no reserve, production, revenue or expense line moves because of AI.

Why this motive

The tool is in use and the company gives results: a validation test against known discoveries (claim c12) and a count of new prospects (claim c13), which point toward efficiency; but the prospects are unconfirmed and undrilled, the remarks call the results promising, and the CEO says much more work is needed to confirm them (claim c3), the exploratory tell for pre-revenue work. A validation score is a test of the model, not an operating result. With the tells in conflict the less durable motive is kept.

Before LLMs: relabelled

At the anchor the CEO described collecting information with every well, shooting seismic and updating reservoir models to look for new opportunities in the Guyana block (anchor claim xom-anchor-c1), with no AI named. The FY2024 10-K said the company must continuously improve its ability to identify the most promising resource prospects (anchor claim xom-anchor-c2), and, in its section on proved reserves additions, named seismic processing, reservoir modeling and simulation software and data analysis packages as the tools used to interpret subsurface data for reserves (anchor claim xom-anchor-c3); neither passage names AI. Exploration expenses, including dry holes, were for FY2024. In coverage the work is called AI and yields a count of unconfirmed prospects, with no line, reserve or revenue shown to move, so the activity is read as renamed rather than expanded (tie-break).

“I think every time we drill, we're collecting information that allows us to better characterize that whole block and focus in on potential new areas of opportunity. And that's basically the work that our teams are very engaged in, is continuing to collect information, continuing to do seismic, continuing to drill, and through that work, update our reservoir models, update our understanding of that block, and then look for new opportunities.”

CEO, qa, earnings call, 2024-04-26

“Our ability to maintain and grow our oil and gas production depends on the success of our exploration and development efforts. Among other factors, we must continuously improve our ability to identify the most promising resource prospects and apply our project management expertise to bring discovered resources online as scheduled and within budget.”

Filing, risk factors, 10-K periodic report, 2025-02-19

“The tools used to interpret the data included seismic processing software, reservoir modeling and simulation software, and data analysis packages.”

Filing, business, 10-K periodic report, 2025-02-19

Figures

New Guyana prospects identified by the AI exploration model, unconfirmed, in the prepared remarks · as-of 2026-07-31
New discovery opportunities in the Guyana block identified with AI-trained models, unconfirmed, as relayed on the call by the CEO from an earlier conference remark · as-of 2026-07-31
Share of known Guyana discoveries the AI exploration model identified in validation testing (nearly) · as-of 2026-07-31
Exploration expenses, including dry holes · 2026-CQ2
Exploration expenses, including dry holes, prior year · 2025-CQ2
Change in exploration expenses, year over year · 2026-CQ2

What else could explain it

other: The anchor describes the same work without AI: each well and each seismic survey updates the reservoir models used to find new opportunities in the block (anchor claim xom-anchor-c1). New prospects would also come from the additional drilling and data the company describes, so the count is not separable into an AI part.
line composition: Exploration expenses of against a year earlier hold seismic, geological and geophysical work and dry holes across the world; the prior year includes a write-off of of previously capitalized exploratory well costs. The 10-Q does not explain the line by AI, and it is shown as the line that holds the work, not as a size.

Quotes

“Technology is also improving our ability to find additional opportunities. We have created sophisticated exploration models powered by both AI and the world's largest seismic database. For example, we trained one model using our Guyana discovery data, and validation testing showed it could identify nearly 90% of known discoveries.”

c12 · Filing, presentation, earnings release, 2026-07-31

“It's already delivering promising results. To date, this model identified four new Guyana prospects ones previously not identified through traditional methods. Even more, this technology can be applied to other basins and plays with similar subsurface characteristics.”

c13 · Filing, presentation, earnings release, 2026-07-31

“4 additional opportunities identified through AI-powered exploration”

c14 · Filing, presentation, earnings release, 2026-07-31

“I think Neil Chapman had mentioned at a prior conference this year that we've really put a lot of effort into artificial intelligence and training models based on what we've found already, all the drilling that we've done, the characterization of that subsurface, and have unleashed that in the rest of the block and have four new discovery opportunities above and beyond what we thought were opportunities.”

c2 · CEO, qa, earnings call, 2026-07-31

“We're optimistic there. Obviously, a lot more work to do to confirm those.”

c3 · CEO, qa, earnings call, 2026-07-31

“I mentioned that with some of the AI tools that we've trained with what we've already found in the drilling we've done, we've seen some new opportunities to explore that we hadn't previously identified.”

c4 · CEO, qa, earnings call, 2026-07-31

operations · expensive to verify

Permian well performance and recovery with AI machine learning

Not sized

AI machine learning is named beside extended reach laterals and surfactants as contributing to Permian performance, and nothing separates its part (the methodology rule for AI named beside another cause); the Permian production it sits inside is quoted as the ceiling, and no ballpark is built on a judgment share.

described, no sizedisclosure: described· motive: exploratory· before LLMs: relabelled

Opened this quarter, on the call only. Asked about Permian technology, capital efficiency and recovery, the CFO says extended reach laterals, surfactants and AI machine learning are all contributing to very strong performance. Registered as a described, unsized joint-cause channel, as UPS's network planning remark was; the prepared remarks and slides do not name AI for the Permian, and Permian production reached a record of more than oil-equivalent barrels per day, the ceiling the AI part sits inside. Read relabelled: the anchor credits the same performance to drilling and completions technology without AI.

Evidence: 3 quotes, 1 figure, 3 confounds, 2 from before coverage

Revenue from Permian oil and gas production that AI machine learning helps raise, through well performance and recovery. On the Q2 2026 call the CFO names AI machine learning beside extended reach laterals and surfactants as contributing to very strong Permian performance, and the CEO describes a portfolio of technology developments being trialed and deployed to raise recovery with fewer wells. Registered as a joint-cause channel, as at UPS (network-planning-ai): AI is named as a present cause of a result, beside other causes, and nothing separates its part, so the channel is described and unsized, with the Permian production it sits inside quoted as the ceiling. The Q2 prepared remarks and slides do not change the reading: they credit Permian recovery and capital efficiency to technology without naming AI. Any revenue would come from crude and gas sold to outside refiners, traders and gas buyers, so counterparty enterprise; sales to the company's own refineries are eliminated in consolidation.

Why this motive

The CFO names AI machine learning as contributing to Permian performance (claim c5) with no measure or line of its own, beside other technologies; a result credited to several changes together does not meet the efficiency tell for any one of them, and AI named as a cause with no measure and no line moving is the exploratory tell (the UPS reading).

Before LLMs: relabelled

At the anchor the FY2024 10-K credited Permian capital efficiency and cost performance to scale, integration and technology, including cube design, proprietary proppant and leading capabilities and technology in drilling and completions (anchor claim xom-anchor-c4), and its cautionary statement named the success of new unconventional technologies (anchor claim xom-anchor-c5); neither names AI. The same cautionary sentence in the Q2 2026 10-Q adds AI-enhanced technologies, and no line, price or volume is shown to move because of AI as distinct from the other technologies, so the activity is read as renamed (tie-break).

“ExxonMobil operations continue to deliver industry-leading capital efficiency and cost performance by leveraging scale, integration, and technology. Examples include deploying ExxonMobil cube design and proprietary proppant as well as leading capabilities and technology in drilling and completions.”

Filing, business, 10-K periodic report, 2025-02-19

“reservoir performance, including variability and timing factors applicable to unconventional resources and the success of new unconventional technologies”

Filing, mdna, 10-K periodic report, 2025-02-19

Figures

Permian production, record, oil-equivalent barrels per day (more than); the ceiling the AI part sits inside · 2026-CQ2

What else could explain it

other: Extended reach laterals and surfactants are named in the same sentence as contributing to the same performance (claim c5), and the CEO credits a portfolio of technology developments with more recovery from fewer wells (claim c18); none of them is AI.
acquisition: Permian growth also reflects the Pioneer acquisition and continued development, as the Q1 prepared remarks say (claim c7).
other: The prepared remarks credit recovery and capital efficiency to technology in general, without naming AI (claim c17).

Quotes

“You look at the extended reach laterals, surfactants, AI machine learning, all of that is contributing to very strong performance even before we start to deploy some of these other technologies.”

c5 · CFO, qa, earnings call, 2026-07-31

“Our Permian operations reached record production of more than 1.8 million oil-equivalent barrels per day. Year versus year, our Permian production growth outpaced every IOC and large independent. Technology is improving recovery and capital efficiency and will contribute to a ~9% production CAGR from 2025 through 2030.”

c17 · Filing, presentation, earnings release, 2026-07-31

“We've been talking for some time now that we've got 40+ technology developments that we're working and have been going out and trialing in the field. The value of those technologies are, most of them are stackable so that you keep building on the success and drive more and more recovery, fewer wells, so less capital.”

c18 · CEO, qa, earnings call, 2026-07-31

Reported lines, year-over-year growth

Revenue +42.3%

Q2 2026. Growing slower than revenue: selling, general and administrative expenses (−1.8%), exploration expenses, including dry holes (−38.2%), total costs and other deductions (+36.4%). A displaced cost shows up as a line that stays under the dashed revenue line. These are the audited lines, as first reported; nothing here is attributed to AI by the filing. Not drawn: exploration expenses, including dry holes, which one-time items move by more than 60% in a quarter; the values are in the table below.

Selling, general and administrative expensesTotal costs and other deductionsRevenue

Reported values and filings

Line	Q1 2025	Q2 2025	Q3 2025	Q4 2025	Q1 2026	Q2 2026
Total revenues and other income	$83.13bn	$81.51bn	$85.29bn	$82.31bn	$85.14bn	$116.02bn
Selling, general and administrative expenses	$2.54bn	$2.53bn	$3.03bn	$3.03bn	$2.68bn	$2.48bn
Exploration expenses, including dry holes	$64mn	$251mn	$149mn	$543mn	$126mn	$155mn
Total costs and other deductions	$71.53bn	$70.80bn	$74.36bn	$74.28bn	$78.17bn	$96.59bn