CLAI Ventures Invests in Terra AI: Bringing AI to Critical Mineral Exploration
Every solar panel, electric vehicle battery, grid-scale storage facility, and high-performance AI data center depends fundamentally on a resilient supply of copper, lithium, nickel, and cobalt. These critical resources must be discovered, characterized, and scaled with unprecedented urgency to support global carbon-neutrality targets.
However, the mineral exploration industry remains fundamentally constrained by decades-old operational paradigms. Traditional discovery relies heavily on knowledge-driven, expert-dependent deposit models and manual interpretation of siloed datasets. This methodology leaves exploration teams dependent on qualitative, linear-additive overlays to make high-stakes, multi-million-dollar drilling decisions.
Today, we are pleased to announce our investment in Terra AI, a Silicon Valley-based company building an AI platform for the discovery and development of critical mineral and clean energy resources. Terra AI is backed by Khosla Ventures and BHP Ventures, and we are glad to be investing alongside them.
The Problem: Exploration Is Broken
The numbers behind the mineral supply challenge are sobering. The world will need more than 600 new large mines for lithium, nickel, copper, cobalt, and graphite alone by 2040, according to industry estimates. At the same time, exploration productivity has gone in the wrong direction: the cost per discovery has risen roughly 16x since the 1990s, while new resource additions in copper have fallen 80–90% over the same period.
The result is a widening gap between the pace of exploration and the pace of demand, which has become a strategic concern not just for mining companies, but for governments managing energy security and industrial supply chains.
The Integration Failure
The bottleneck in modern exploration is not an absence of raw physical data; it is the challenge of integrating diverse data sets to analyze the broader picture. Advanced multi-sensor campaigns generate terabytes of highly heterogeneous inputs across distinct scales, from space borne satellite arrays (Landsat, ASTER, Sentinel-2) and centimeter-scale drone imagery to multi-physics geophysics and subterranean geochemical assays.
The friction arises because highly skilled geologists are forced to act as human middleware. They must evaluate these massive, multi-dimensional spatial and spectral data cubes in isolation, attempting to interpret what each independent data layer represents. Afterwards, geologists mentally aggregate these data layers into a cohesive concept. However, each measured input only provides a fragment of information. Much remains open to interpretation from and between these data points, where signals can be easily overfitted — or missed.
The 13-Year Lead Time: Moving an asset from initial greenfield reconnaissance to a defined, developed deposit routinely spans more than 13 years, creating an unsustainable lag relative to the exponential pace of technological demand.
The 90% False Positive Rate: When legacy exploration teams finally cross-reference their data to select a high-stakes, multi-million-dollar drill target, they face a roughly 90% false positive rate.
Imagine a software engineering environment where a single debug loop costs millions of dollars, takes a decade to log results, and compiles with a 90% critical error rate. That is the current operational reality of global resource acquisition.
What was once an operational headache for mining companies has escalated into a core macroeconomic vulnerability for governments managing energy security and industrial computing pipelines.
The Solution: A Data-Universal AI Platform
Terra AI’s approach is to fuse all available modes of exploration data — geophysical, geochemical, geological, and remote sensing — into a unified AI model that produces thousands of geologically plausible, quantitative subsurface models simultaneously. Rather than a single deterministic answer, the platform harnesses geologists’ expertise to test hypotheses and interrogate data, generating sharper probabilistic outputs that reflect where genuine uncertainty in the models remain. These models allow teams to make better-informed decisions about where to survey and where to drill, accelerating the exploration timeline from early discovery to shovel-ready development.
The company describes this as moving beyond the “single qualitative model” paradigm that has defined exploration for generations. Their platform supports end-to-end targeting, from initial screening through drill planning, with the model updating continuously as new data comes in.
The technical differentiation is meaningful. Exploration AI competitors have largely operated in single-data-mode settings, generating 2D hotspot maps or working within one data type at a time. Terra AI’s data-universal approach fuses multiple modalities into 3D probabilistic models, backed by a proprietary IP portfolio including patents in generative geological modeling, neural surrogate simulation, and sequential planning for robust reservoir optimization.
Terra AI has applications for all stages of mining exploration, from target screening, to survey design, to dynamic drilling. At the screening stage, the platform builds full-featured 3D geological models from all available data rather than relying on single-mode inversions, allowing teams to assess economic prospectivity across hundreds of thousands of models rather than just one, and to walk away from poor targets with confidence. For survey design, it optimizes multi-physics programs by simulating deposits before expensive surveys are run, predicting in advance how much each data collection option would reduce uncertainty. And for drilling, it operates in a closed loop: each new drill target is informed by the results of the previous one, with models updating in real time as new core data comes in.
Early Results
The company’s first major pilot result offers a useful proof point. Working with a nickel-copper mine in Latin America, described by the client as one of the most geologically complex they have encountered, using existing data Terra AI identified roughly 5 megatons of high-grade ore mineralization valued at approximately $600M. Early drill testing of the AI-identified targets confirmed the predicted mineralization, a notable early validation that the models reflect real subsurface geology.
The platform has since been deployed with a number of enterprise and mid-sized mining clients across several commodities and continents.
The Team
The founding team brings together rare depth in both AI and geoscience.
John Mern, PhD (Co-Founder & CEO) holds an AI PhD from Stanford and previously worked as an AI lead at KoBold Metals and in AI roles at Boeing and Phantom Works.
Anthony Corso, PhD (Co-Founder & CTO) was the Executive Director of Stanford’s AI Safety Center and also holds an AI PhD from Stanford.
The broader team includes four AI/ML PhDs, two geophysics PhDs, two geoscience professors, two geophysics PhD researchers, five veteran software engineers, and multiple commercial leaders. Their institutional affiliations span Stanford, Berkeley, NASA, Google, Meta, Amazon, BlackRock, Exxon, and SLB.
Why We Invested
At CLAI Ventures, we invest at the intersection of AI and climate. Critical mineral supply is a direct constraint on the pace of the energy transition. Without adequate copper, lithium, and nickel, electrification slows, energy security weakens, and the costs of decarbonization increase. Terra AI is building infrastructure that could make exploration measurably more productive, reducing both the time and capital required to find the resources the transition depends on.
What attracted us to Terra AI specifically was the combination of a technically differentiated platform, a team with genuine depth in both AI and geoscience, strong early customer validation from demanding enterprise clients, and a clear line of sight to the scale of the opportunity. John’s lessons learned with his experience in past AI enabled mineral exploration startups has informed their ability to execute at pace and quality with the modest capital to date.
We are glad to be supporting John, Anthony, and the Terra AI team as they continue building.
If you are an investor or mining professional interested in learning more, we are happy to make an introduction. Reach us at ajay@clai.vc and team@clai.vc.
This post was written by the CLAI Ventures team, a Silicon Valley-based fund investing at the intersection of AI and climate.