The company says AI’s next bottleneck isn’t chips or power, it’s finding the metal to wire it all together. Working with Nvidia, Its chief scientist calls Tau 3.8 an “AlphaGo moment for geology”
TORONTO and ADELAIDE, Australia, Nov. 10, 2026 /PRNewswire/ — Fleet Space Technologies (Fleet), the space and AI company backed by In-Q-Tel and Teachers’ Venture Growth (TVG), the late-stage venture and growth investment arm of Ontario Teachers’ Pension Plan, today launched Tau 3.8 in the United States and Canada, an AI “world model” for mineral systems.
Trained on ten million synthetic geological worlds, Tau combines Fleet’s 3D seismic intelligence with drilling, geology, geophysics and mineralogy data. It reconstructs hidden geology, ranks targets for resource expansion and identifies which new measurement would most reduce uncertainty, so mining companies can decide where to drill next with far more confidence.
North America’s AI build-out is running into a physical limit: copper. Every data center, transmission line and grid upgrade needs it. S&P Global forecasts that data-center copper demand will more than double, from 1.1 million metric tons in 2025 to 2.5 million metric tons by 2040, with AI training accounting for 58% of that demand by 2030. Copper is one of six priority minerals in Canada’s Critical Minerals Strategy, yet Canadian copper production fell by nearly 25% between 2014 and 2023, according to Natural Resources Canada. At the same time, the US now relies on imports for 57% of the copper it consumes, up from 45% a year earlier, according to the US Geological Survey.
Most of the debate about AI’s limits has focused on chips and power. Fleet argues the deeper constraint is discovery. Finding and expanding the deposits that must supply the industry is still slow, expensive and uncertain, and billion-dollar decisions rest on drillholes that sample only a fraction of the ground beneath them.
“North America is building an AI economy on a mineral base it hasn’t yet found,” said Fleet Co-Founder and CEO Flavia Tata Nardini. “Hyperscalers can order more GPUs and sign new power deals, but they can’t order a copper deposit into existence. The gap between what these economies need and what the industry can bring online is widening and mineral discovery has become a constraint on growth itself. Companies have spent years collecting drilling and survey data. Tau brings it together in one model so teams understand the orebody sooner and every foot drilled is worth more.”
“Toronto is where much of modern AI was pioneered and it’s where the world finances its mines. Those two worlds haven’t spoken to each other nearly enough,” Flavia said. “Tau is what happens when they do. The capital that funds exploration gets a sharper answer on where to drill before the money goes into the ground. Faster, more capital-efficient exploration is how supply keeps pace with the economy AI is creating.”
Fewer dry holes
Drilling is typically the largest cost in mineral exploration and a single deep diamond drillhole can cost hundreds of thousands of dollars. By predicting copper grades more accurately from the drilling already in hand, Tau shows teams where grade is likely to continue and where it isn’t, so fewer holes end in barren ground. Companies can prioritize their highest-value targets and define and expand resources with less drilling, in less time and at lower cost.
At ATEX Resources’ Valeriano copper-gold project in Chile, Tau combined Fleet’s ambient noise tomography (ANT) and magnetotelluric (MT) surveys with Comet and machine learning. Tested on drillholes held back from it, the model predicted copper grades with 38% less error than a conventional geostatistical estimate (kriging) built from the same drilling. On the deepest step-out holes, the error was up to 67% lower.
Fleet is already being used in the field, from Pakistan to Canada’s Athabasca Basin. Projects include Barrick’s Reko Diq copper-gold project in Pakistan, Lundin Gold’s Fruta del Norte gold mine in Ecuador, Fermi Exploration’s Perch River project in the Athabasca Basin, Southern Cross Gold’s Sunday Creek gold-antimony project in Victoria, Australia and Alkane Resources’ Boda-Kaiser copper-gold project in New South Wales. Fleet’s wider customer base includes Gold Fields, Rio Tinto, Ma’aden, Standard Uranium, IsoEnergy, Q2 Metals and France’s geological survey, BRGM.
World models, applied to the planet
World models, AI systems that learn how the physical world behaves rather than predicting the next word, have become one of the most closely watched frontiers in AI. Most of the attention has gone to robotics, video and autonomous vehicles. Fleet has applied the approach to an environment that is far harder to observe than any road or factory floor: the Earth’s subsurface.
“Putting a wrapper around a large language model does not create a world-class geologist,” said Fleet Chief Scientific Officer Dr. Gerrit Olivier. “Mineral systems are three-dimensional, buried and understood through observations that are sparse and costly to acquire. Fleet’s approach was to build ten million simulated geological environments, model what seismic sensors and drillholes would detect in each, and train Tau to infer the geology between those partial measurements.”
“World models learn the structure, geometry and composition of the physical world itself. We now appear to have crossed a critical threshold: a model trained in synthetic worlds can make predictions and guide exploration decisions using real-world data. I believe this marks an AlphaGo moment for geology.”
Why seismic matters
AI can generate millions of plausible versions of what lies underground, but many of them will fit the same sparse data. Seismic provides the physical evidence to tell them apart, constraining the depth, geometry, orientation and continuity of structures in the vast volumes between drillholes.
“Seismic narrows the geological possibilities between drillholes, so teams can answer the questions that drive capital decisions: where the deposit extends, how deep it runs, and where the next foot of drilling will add the most value,” Olivier said.
This integration is Fleet’s advantage: Tau tests competing predictions against measurable physical properties of the subsurface rather than relying on geological plausibility alone. As new data arrives, the model updates its picture and recommends the next measurement, an approach Fleet calls Agentic Geoscience.
“Tau does not remove the geologist from the decision,” Flavia said. “It gives them a quantified view of uncertainty before the drill bit provides the answer.”
Built with the North American AI ecosystem
Tau was developed by specialists in seismology, geology, geophysics, applied mathematics and machine learning, including geoscientists with more than 200 years of combined experience across 400-plus programs and machine-learning leaders from Caltech and the University of Oxford. Fleet is working with Nvidia on the accelerated computing infrastructure needed to train the next generation of high-resolution subsurface models.
Fleet is an industrial affiliate of Stanford University’s Mineral-X program, which connects researchers and industry on AI, geoscience and decision-making under uncertainty. Tata Nardini serves on the Mineral-X Advisory Council.
*Note: the conventional estimate was an independent exercise and is not related to the 2025 SRK Mineral Resource Estimate.
ENDS
About Fleet:
Fleet Space Technologies is a space and AI company building the technology to find the minerals the global economy depends on. Its flagship platform, ExoSphere, combines Fleet’s satellite network, rapid 3D seismic data acquisition and AI into a single system that images mineral systems in near real time with minimal environmental impact. Tau, Fleet’s world model for mineral systems, extends that platform into Agentic Geoscience. Fleet works with leading exploration and mining companies including Rio Tinto and Barrick, and is backed by investors including In-Q-Tel, Blackbird and TVG. Founded in Australia, Fleet has operations across the United States, Canada, Chile, Australia and Africa.
View original content to download multimedia:https://www.prnewswire.co.uk/news-releases/ai-trained-on-ten-million-synthetic-earths-outperforms-conventional-copper-estimates-at-tsxv-listed-atexs-valeriano-project-302934732.html
