TotalEnergies and Mistral Launch €100 Million Partnership to Develop Energy AI Models

The partnership builds on an earlier collaboration that began in June 2025, rather than starting from scratch.
The joint scientific laboratory will have TotalEnergies and Mistral AI teams working side by side, combining domain specialists with AI developers in a dedicated program.
The AI tools are intended to shorten reservoir-planning timelines: tasks that traditionally require engineers to spend months modeling development options could be accelerated so teams can assess many more scenarios involving factors such as pressure, porosity, production rates and economic assumptions.
TotalEnergies CEO Patrick Pouyanné said exploration and reservoir engineering are areas where AI can create “the greatest value” for the company’s activities.
TotalEnergies and French AI firm Mistral AI are launching a three-year partnership worth more than €100 million to build custom artificial-intelligence models for oil and gas exploration and reservoir engineering. Neoteo The joint effort combines nearly a century of TotalEnergies' geoscience expertise with Mistral's AI capabilities and close to 10 petaflops of subsurface data — roughly 10 quadrillion bytes of geological information stored in Europe.
The initiative expands an earlier collaboration that started in June 2025. CEO Patrick Pouyanné said exploration and reservoir engineering are areas where AI can create "the greatest value" for the company's operations. Yahoo Finance The models aim to speed up tasks that normally take engineers months, allowing teams to test many more development scenarios involving pressure, porosity, production rates, and economic assumptions.
The partnership builds on momentum from an earlier deal announced in June 2025, when TotalEnergies and Mistral formed an initial strategic lab focused on multi-energy solutions and emissions reductions. Petroleum Australia The new €100 million commitment signals a serious expansion into domain-specific AI development — moving beyond general-purpose tools to build models tailored specifically to oil and gas subsurface science.
DataCenterDynamics reports that the two companies will establish a dedicated joint scientific laboratory with teams working side by side. Domain specialists from TotalEnergies and AI developers from Mistral will collaborate in a structured program designed to merge reservoir expertise with cutting-edge machine-learning engineering.
Traditionally, engineers spend months building subsurface models to evaluate different drilling and development options. The new AI tools aim to compress that process dramatically. Rather than testing a handful of scenarios, teams could assess dozens or hundreds of complex development plans — each factoring in variables like rock pressure, fluid flow, production forecasts, and cost assumptions.
This speed improvement cuts exploration risk and uncertainty. A single misjudged well can cost tens of millions of dollars, so better decision-making tools deliver enormous economic value. BeBeez The partnership also supports TotalEnergies' goal of extending the productive lives of existing mature fields across its global portfolio, including African operations.
Both companies emphasize a commitment to European technological sovereignty. The deal ensures that TotalEnergies' sensitive subsurface data and intellectual property remain within Europe rather than flowing to U.S. cloud providers or AI giants. Mistral AI co-founder Arthur Mensch said the partnership "highlights the strength of Europe's industrial ecosystem" and the need for large enterprises to adopt customizable AI solutions that protect their proprietary information.
The initiative provides Mistral AI with a high-profile industrial proving ground for European AI models in complex scientific workflows. It also validates the case for building bespoke, domain-specific AI rather than relying solely on off-the-shelf commercial tools — a shift in strategy for many major European corporations seeking to reduce dependence on American tech infrastructure.
Publishers
36
Articles
68
Reach
104