Industry View · Oil & Gas
AI inside oil and gas is a modest software market — roughly $3.8 billion in 2025 — riding on a far larger story: the same hydrocarbon firms using machine learning to squeeze barrels are now selling gigawatts of gas-fired power to the data centers training the models. The line item is small; the demand AI creates is not.
Real wins (Exxon gas-lift >5%) but 86% of energy AI projects never leave pilot
Deep-learning models read subsurface data faster and cheaper than crews, compressing exploration cycles and cutting field acquisition.
AI trims well planning time and steers real-time drilling; ExxonMobil cut well planning and design from 9 to 7 months.
Gas-lift and artificial-lift optimization plus predictive maintenance deliver the clearest, sourced production and uptime gains today.
Digital twins and emerging agentic AI orchestrate multi-step operational decisions, though most deployments remain advisory.
Majors and midstream pivot to supplying behind-the-meter gas power to data centers, the fastest-moving and largest-dollar shift.
01 · The thesis
The first clock is operational and quiet. AI works inside the well: ExxonMobil's automated gas-lift optimization in the Bakken lifts output more than 5% across hundreds of wells; Shell's deep-learning seismic work with SparkCognition cut required shots by ~99%; McKinsey pegs integrated AI at up to 20% lower opex and 5-8% higher production efficiency. Yet the same firm finds 86% of energy AI projects never leave pilot — the value is proven, the diffusion is not. The second clock is structural and far louder. AI's electricity hunger has turned oil majors into power developers. Chevron is building 2.5 GW (scalable to 5 GW) in West Texas; ExxonMobil a 1.5+ GW gas-plus-carbon-capture plant; developers have announced roughly 101 GW of on-site gas generation. The disruptive force here is not AI optimizing barrels — it is AI buying the molecules.
Deep-learning models read subsurface data faster and cheaper than crews, compressing exploration cycles and cutting field acquisition.
AI trims well planning time and steers real-time drilling; ExxonMobil cut well planning and design from 9 to 7 months.
Gas-lift and artificial-lift optimization plus predictive maintenance deliver the clearest, sourced production and uptime gains today.
Digital twins and emerging agentic AI orchestrate multi-step operational decisions, though most deployments remain advisory.
Majors and midstream pivot to supplying behind-the-meter gas power to data centers, the fastest-moving and largest-dollar shift.
02 · The two clocks
Operational clock. The wins are real and measurable — ExxonMobil's gas-lift ML adds more than 5% to well output, McKinsey models up to 20% opex reduction — but 86% of energy AI pilots never scale. Inside the well, AI is a steady compounding efficiency story, not a sudden rupture. Demand clock. Data-center load is rewriting gas markets in real time. Meeting AI power demand could require US gas production to rise 10-15% by the early 2030s, and developers have already announced about 101 GW of on-site gas generation to bypass grid interconnection queues. Labor clock. Efficiency cuts both ways. Chevron has said it will reduce its workforce 20% by end-2026, ConocoPhillips up to 25%, and ExxonMobil roughly 2,000 jobs — driven by prices, consolidation, new technology and workflow change, with automation accelerating the trend.
03 · Public players & exposure
We plot the listed players on two editorial axes — how exposed each is to AI disruption, against how ready its data, brand and position are to be the answer. The figures in the table are sourced; the placement is our read.
04 · Private flagships
The companies attacking this industry AI-first, with disclosed funding where available:
Runs production-grade ML in the Bakken (gas-lift, >5% uplift) while pivoting to supply 1.5+ GW of gas-plus-carbon-capture power to a data center.
Leveraging decades of behind-the-meter power experience (Gorgon, Tengiz) to build 2.5-5 GW of West Texas generation for AI campuses.
Its Lumi platform unifies subsurface data and embeds generative AI across exploration, drilling and production workflows.
Deployed enterprise AI to Shell, Eni, QatarEnergy LNG, Petronas and ExxonMobil for production and reliability optimization.
AI-data-center power demand turned an enhanced-geothermal startup into clean energy's biggest IPO.
Machine-health AI applied to rotating equipment underpins the predictive-maintenance layer across energy assets.
05 · Signals
06 · The exposure read
AI rewards clean, structured advantage and punishes friction. The line runs through who owns the data, the brand and the customer — and who is merely a step the technology can route around.
Sources
Energy-for-AI: gas power the majors are building
Announced gas-fired generation capacity tied to AI data centers (Source: JPT/SPE, DCD, Energy News Beat, 2025) (GW)
ExxonMobil unveils 1.5+ GW data-center plant
A gas-plus-CCS facility designed to capture >90% of CO2, signaling majors entering behind-the-meter power generation.
Augury raises $75M at $1B+ valuation
Industrial predictive-maintenance AI keeps attracting capital even as broader funding tightens.
Chevron commits to 2.5-5 GW West Texas plant
With Engine No. 1 and GE Vernova, Chevron targets up to 4 GW of generation for the data-center market, online ~2027.
C3.ai and Baker Hughes renew JV through 2028
The energy AI alliance, >$500M revenue since 2019, is extended and expanded amid pure-play vendor uncertainty.
Fervo's IPO tops $10B on AI power demand
The largest clean-energy IPO ever, explicitly fueled by data-center electricity needs, validates energy-for-AI as a category.
Runs production-grade ML in the Bakken (gas-lift, >5% uplift) while pivoting to supply 1.5+ GW of gas-plus-carbon-capture power to a data center.
Leveraging decades of behind-the-meter power experience (Gorgon, Tengiz) to build 2.5-5 GW of West Texas generation for AI campuses.
Its Lumi platform unifies subsurface data and embeds generative AI across exploration, drilling and production workflows.
Deployed enterprise AI to Shell, Eni, QatarEnergy LNG, Petronas and ExxonMobil for production and reliability optimization.
AI-data-center power demand turned an enhanced-geothermal startup into clean energy's biggest IPO.
Machine-health AI applied to rotating equipment underpins the predictive-maintenance layer across energy assets.
| Company | Stance | The sourced fact |
|---|---|---|
| ExxonMobil XOM | Power pivot leader | Bakken automated gas-lift ML raises output >5% across hundreds of wells; planning a 1.5+ GW gas-plus-CCS plant for a data center. |
| Chevron CVX | Gigawatt developer | Building a 2.5 GW West Texas plant (scalable to 5 GW) for AI data centers, online expected 2027; up to 4 GW total with Engine No. 1 and GE Vernova. |
| SLB SLB | Digital incumbent | Launched the Lumi data and AI platform embedding LLMs and domain models across the energy value chain; Digital & Integration decouples revenue from oil-price cycles. |
| Baker Hughes BKR | AI JV partner | BakerHughesC3 joint venture with C3.ai has generated over $500M revenue since 2019 and was renewed and expanded through June 2028. |
| Halliburton HAL | Services under pressure | Competes in digital suites against SLB and Baker Hughes but began 2025 layoffs amid lower prices and consolidation. |
| Williams WMB | Midstream-as-power | Partnered to build and operate a dedicated gas-fired generation facility for a Meta data center campus in Ohio, defining midstream as integrated power provider. |
| C3.ai AI | Pure-play AI vendor | FY2025 revenue up 25% to $389.1M with 84% recurring subscriptions; energy JV with Baker Hughes anchors its oil and gas exposure. |
| Fervo Energy FERV | Energy-for-AI winner | Raised $1.89B in a 2026 IPO valuing it above $10B, fueled by data-center demand; earlier $462M round included Google and Devon Energy. |
| Cognite COGNITE | Industrial data layer | Industrial data platform valued at $1.6B in a $150M round; supplies the contextualized data foundation many oil and gas AI workflows depend on. |
| SparkCognition SPARK | Predictive-AI unicorn | Reached a $1.4B unicorn valuation; its deep-learning seismic work helped Shell cut required seismic shots by roughly 99%. |