AI Takes Aim at the CO₂-EOR Challenge

New studies and rising CCS activity push operators to explore AI tools for future CO₂-EOR and carbon strategy planning

10 Oct 2025

Artificial intelligence is starting to shape how operators weigh the future of CO₂-enhanced oil recovery. Fresh research and a burst of carbon-storage activity have encouraged firms to test digital models for both production and decarbonisation, even as commercial use remains slight.

A March 2025 study from the University of Kansas, published in Frontiers in Energy Research, has added momentum. It shows that an AI-driven proxy model can speed the assessment of CO₂-injection behaviour, oil-recovery potential and mitigation effects across varied reservoir settings. The work is academic, not industrial. But it reinforces hopes that such tools can shrink the time needed to sift through scenarios before engineers launch full simulations.

Meanwhile America’s carbon-storage business is gathering pace. In April 2025 ExxonMobil struck a deal with Calpine to transport and permanently store as much as 2m tonnes of CO₂ a year from the Baytown Energy Center in Texas. The venture is aimed at transport and burial rather than CO₂-EOR, yet it signals that large-scale carbon management is shifting from promise to practice. The growing flow of data and infrastructure may later support enhanced-recovery schemes.

Chevron, too, has stressed digital capacity. Its expanded Engineering and Innovation Excellence Center in Bengaluru now supports global work through real-time geological modelling and advanced analytics. There is no confirmed commercial use of AI dedicated solely to CO₂-EOR. Even so, observers expect these investments to shape recovery and storage strategies over time.

Service firms and software houses spy an opening as operators probe how new tools might refine reservoir characterisation, screen fields and strengthen long-term planning. For now, though, this is exploration, not deployment. Many companies are running internal pilots to see where AI can complement conventional simulation, risk analysis and storage-verification methods.

Researchers warn that algorithms are only as sound as the data behind them. Patchy records from older fields and uncertainty over CO₂-plume movement can unsettle even polished models. Hence growing demands for transparent tests, open benchmarks and integrated workflows that spell out how AI systems are trained, checked and updated.

Still, incentives for carbon storage and pressure for credible climate strategies are nudging producers toward tools that sharpen forecasts and monitoring. As CCUS projects expand and sensing improves, AI-assisted assessment is likely to gain ground. The field is in its early days, but the drift is clear: deeper digitalism in future recovery and storage plans.