How Chevron Leverages Generative AI to Enhance Oil Exploration and Production

Harnessing Data in Oil and Gas Operations

Oil and gas operations generate vast amounts of data, with a seismic survey in New Mexico producing a file size that can reach a petabyte.

Bill Braun, CIO of Chevron, highlighted the scale of data processing required, stating, “To turn that into an image for decision-making is a 100 exaflop operation. It’s an incredible amount of compute.”

Chevron has been relying on GPUs since 2008, long before many industries recognized the need for such processing power. Now, the company is leveraging cutting-edge generative AI tools to extract deeper insights and greater value from its extensive datasets.

“AI is a perfect match for large-scale enterprises with substantial datasets — it’s exactly the tool we need,” Braun remarked.

Insights from the Permian Basin

The challenge isn’t unique to Chevron; large data repositories are prevalent across the industry. Braun mentioned the Permian Basin, located in West Texas and southeastern New Mexico, where Chevron holds significant land. This expansive area spans approximately 250 miles wide and 300 miles long, housing an estimated 20 billion barrels of oil, contributing to 40% of U.S. oil production and 15% of natural gas output.

“They’ve been a huge part of the U.S. production story over the last decade,” Braun noted. One unique advantage is a requirement from the Railroad Commission of Texas for all operators to publicly disclose site activities. “Everything’s a public record,” Braun said, emphasizing the strategic advantage of this transparency: “It presents a chance to learn from your competition, and if you’re not doing that, they’re learning from you. It’s an enormous accelerant for industry learning.”

Promoting Proactive Collaboration and Safety

Chevron’s operations span extensive areas, where data quality can vary. Braun pointed out that generative AI can be essential for filling in geological gaps between data points. “It’s the perfect application to complete the model,” he said.

For instance, with well lengths stretching for miles, AI can alert teams to potential interferences with nearby operations, allowing for proactive communication to prevent disruptions. Additionally, Chevron employs large language models (LLMs) to develop engineering standards, specifications, safety bulletins, and alerts, continually refining these models for optimal precision.

“When exact constructions are needed, we don’t want our generative AI to get creative,” Braun explained. “Those must be tuned exceptionally tightly.”

The company is also exploring robotic models to enhance safety. “The goal is for robots to tackle dangerous tasks while humans oversee the operations from a safe distance,” he stated, adding that this approach can reduce costs and liabilities.

Fostering Collaboration Across Teams

Traditionally, energy sector teams have operated in silos, both physically and digitally. Chevron has focused on bridging this gap by embedding teams together. “The highest performing teams emerge when machine learning engineers collaborate with mechanical engineers on shared challenges,” Braun explained.

Chevron has also invested in sending engineers back to school to earn advanced degrees in data science and systems engineering, integrating data scientists — or “digital scholars” — with operational teams to encourage innovative approaches. “We’ve matured by starting with small wins and continuously building on them,” said Braun.

Addressing Environmental Concerns with Technology

As with every industry, environmental impact is a significant concern in energy. Carbon sequestration — capturing and permanently storing CO2 — is gaining importance. Braun asserted that Chevron operates some of the largest carbon sequestration facilities globally, though uncertainties remain about reservoir performance over time.

To address these uncertainties, Chevron employs digital twin simulations to ensure carbon remains contained and generates synthetic data for predictive analytics. Braun also emphasized the importance of managing the substantial energy consumption of data centers and AI applications to maintain clean operations, stating, “How to manage those often-remote locations as cleanly as possible is always where the conversation starts.”