By Farhaa Abdelhaq
Saudi Arabia’s mining story is entering a new chapter: from licenses and legislation to the hard work of finding what is actually in the ground. The latest signal comes from an unlikely pairing: a research university and the kingdom’s flagship miner. King Abdullah University of Science and Technology (KAUST) has built a generative AI model that maps the subsurface in multiple plausible scenarios, and Ma’aden has now signed a 30‑month contract to test it on real Saudi geology.
is an early indication of how the kingdom intends to de‑risk exploration, improve capital efficiency, and move up the global mining value chain.
From one “best guess” to many scenarios
Traditional mineral exploration in Saudi Arabia has followed a familiar pattern. Geologists collect borehole data, surface observations, geophysical surveys, and remote sensing imagery, then reconcile them into a single deterministic model of the subsurface. That model becomes the basis for drilling programs, resource estimates, and investment decisions.
The problem is that the underground is inherently uncertain. With limited data, a single “best guess” can hide risk: it may look coherent on paper but still lead to dry holes, delayed timelines, and capital tied up in marginal targets.
KAUST’s approach reframes the question. Instead of forcing all available data into one interpretation, its AI system generates multiple 3D geological scenarios that are all consistent with the evidence. Each scenario carries a measure of uncertainty, so decision‑makers can see not only what is likely, but what else is possible and how probable each outcome is.
Technically, the model is a 78‑million‑parameter 3D attention U‑Net trained on hundreds of thousands of synthetic geological models on KAUST’s Shaheen III supercomputer. The training data comes from StructuralGeo’s geological simulator, which mimics millions of years of tectonic, magmatic, and sedimentary processes to create realistic synthetic geology. The result is a foundation model for subsurface geology that can ingest sparse real‑world data and output an ensemble of 3D models rather than a single interpretation.
As the researchers emphasize, the tool is designed to enhance geologists’ decision-making, not replace them.
The development team includes KAUST researchers and collaborators from the University of British Columbia, with code and technical notes already circulated via arXiv and GitHub. That openness is unusual in a sector known for proprietary black boxes and may encourage other operators and service companies to experiment with similar approaches.
Where this fits in the broader mining push
This is not an isolated tech story. It sits squarely inside Saudi Arabia’s broader push to make mining a third industrial pillar alongside oil and petrochemicals, explicitly backed by digital tools and AI.
Industry and Mineral Resources Minister Bandar Alkhorayef has repeatedly highlighted AI and data analytics as levers to cut exploration costs, improve targeting, and accelerate discovery. National spending on mining exploration has more than doubled in recent years, licensing activity has accelerated, and market analysts project the Saudi AI‑in‑mining market to grow at roughly 38% CAGR through 2032.
What this could mean for capital and operations
Multiple scenarios with uncertainty bands can help prioritize drilling targets, reduce dry holes, and shorten the time from target generation to resource definition. Being able to show a range of plausible resource geometries, not just a single optimistic model, can strengthen internal approvals and external fundraising.
If the 30‑month pilot shows measurable gains, expect other miners, juniors, and service providers to seek similar AI‑enabled workflows, creating demand for data, compute, and specialized talent.
In a sector where projects can take a decade or more from exploration to production, even modest improvements in hit rates or target identification can have a material impact on returns.
Equally important is whether this approach spreads beyond Ma’aden. If other license holders and service companies begin to adopt similar “multiple scenario” workflows, Saudi Arabia could become a live laboratory for AI‑enabled exploration at scale—exactly the kind of capability that supports a $75bn+ mining GDP ambition.
