Faster Fields: AI Speeds Up Subsurface Decisions

SLB and Rescale use AI models to reduce subsurface planning cycles from days to hours

23 Mar 2025

SLB has partnered with cloud simulation company Rescale to accelerate reservoir modelling by using artificial intelligence to shorten simulation times from days to hours, in a move aimed at speeding up planning decisions in oil and gas projects.

Reservoir simulations are used to predict how oil, gas and water move through underground rock formations and are central to field development plans. The models help engineers decide where to drill, how to manage production and how to respond to new data. But conventional physics-based simulations are computationally heavy and can take several days to run, slowing decision-making.

The companies say the new approach relies on so-called AI surrogate models, which are trained on large libraries of past simulations. Once trained, the models can reproduce results that are close to those produced by traditional methods, but at much higher speed.

SLB said its upgraded modelling platform can now complete simulation cycles within hours, allowing engineers to test more scenarios within the same planning window. This, it argues, improves the ability of teams to respond to uncertainty and adjust plans as reservoir conditions change.

“AI isn’t replacing the physics, it’s accelerating it,” said Edward Hsu, chief technology officer at Rescale. “This lets engineers evaluate more scenarios, make faster decisions, and stay ahead of the curve.”

Faster modelling could have financial implications for operators, which face rising pressure to improve efficiency and control costs. By compressing planning cycles, companies can assess alternative development strategies more quickly and reduce the risk of expensive design changes later in a project’s life.

The collaboration also reflects a wider shift in the energy industry towards digital tools that combine cloud computing, data analytics and automation. Static, long-range planning models are increasingly being replaced by systems that can be updated more frequently as new data becomes available.

However, the use of AI in subsurface engineering remains sensitive. Industry specialists have warned that surrogate models can obscure how results are produced and may drift from physical reality if not properly monitored. In response, developers say the AI outputs are checked against conventional simulations and include controls designed to flag unreliable results.

The pace at which such tools are adopted will depend on how confident operators remain in their accuracy, particularly for high-cost, long-lived assets.