Nvidia teams with Fervo Energy and PNNL on AI “digital twin” for geothermal systems

Nvidia is expanding its push to apply artificial intelligence beyond software and semiconductor products, partnering with geothermal developer Fervo Energy and the Pacific Northwest National Laboratory (PNNL) on an AI-based “digital twin” aimed at enhancing geothermal power systems.

According to a report carried by Yahoo Finance, the companies are working together on EGS-Twin, described as an AI-powered digital twin for Enhanced Geothermal Systems (EGS). EGS refers to projects that create or improve underground heat extraction by engineering subsurface reservoirs rather than relying only on naturally occurring geothermal fields. In practice, a digital twin is a data-driven virtual representation meant to mirror real-world physical systems so teams can simulate scenarios, test changes, and improve decisions without repeated field trial-and-error.

The project is built around Nvidia Omniverse, Nvidia’s platform for building and running simulation and 3D “world” models. Nvidia has positioned Omniverse as an enabling layer for connecting simulation, data, and collaboration across complex physical systems, and this geothermal initiative is presented as one more domain where those capabilities could be applied.

Yahoo Finance’s account ties the geothermal effort to the broader idea of bringing advanced computing and AI workflows to energy infrastructure. EGS-Twin is framed as a co-development effort, with Nvidia, Fervo Energy, and PNNL combining domain expertise and technical tooling. The implication is that geothermal operators can use high-fidelity digital models to accelerate design iterations, refine operations, and reduce uncertainty during development.

While Nvidia and its partners have not, in the reported materials provided for this story, disclosed timelines, performance targets, or how the system will be validated against operating geothermal assets, the collaboration suggests a structured approach to simulation-heavy energy work. Digital twins in energy typically rely on a combination of engineering models, sensor or operational data, and machine-learning components. The exact data sources and model training approach for EGS-Twin are not described in the provided report.

For Nvidia, the geothermal project reflects a strategy that goes beyond selling chips for training and inference. By embedding its simulation and platform software into energy-industry workflows, the company is also positioning itself as a supplier of “digital infrastructure” for applied AI, including industrial modeling and computational science.

For Fervo Energy and PNNL, the value proposition likely centers on faster learning cycles. Geothermal development involves costly drilling and long development windows, and uncertainty around subsurface behavior is a persistent challenge. A digital twin approach, if paired with reliable data, can help operators test well configurations, stimulation strategies, and operational changes virtually before committing resources in the field.

Still, key specifics remain unreported. The available description does not include any contract size, licensing terms, or which geothermal sites will supply data for the digital twin. It also does not state measurable outcomes such as cycle-time reduction, accuracy improvements, or expected cost impacts. Without those details, it is difficult to assess how quickly EGS-Twin could move from a pilot concept to a production tool used across geothermal projects.

Looking ahead, what to watch is whether Nvidia, Fervo Energy, and PNNL publish evaluation results, technical documentation, or milestones demonstrating EGS-Twin’s performance. Industry observers will likely focus on how the digital twin is calibrated, how frequently it is updated as field data arrives, and whether it meaningfully improves decision-making during development or operations.