Terra AI raises $20M to scale generative subsurface modeling

Terra AI sits at the intersection of generative modeling and the physical world, using AI to build probabilistic 3D models of the Earth’s subsurface. That’s important because many of the hardest bottlenecks in energy transition and critical minerals are geological, not purely financial. If AI can compress the time and capital needed to find and de-risk new deposits, it changes the speed at which large-scale infrastructure can come online.

Strategically, this is an example of “agentic AI” moving into capital-intensive, slow-feedback domains. Terra AI’s geology reasoning agent collaborates with human experts to propose drill plans and exploration strategies, an early template for how domain specialists and agents might co-manage complex physical systems. Backers like Khosla Ventures and BHP Ventures are effectively betting that whoever owns the AI layer for subsurface decisions will have leverage across mining, geothermal and carbon storage.

For AGI timelines, this is more about breadth than peak intelligence. The models Terra AI builds are highly specialized but demand multi-modal reasoning over noisy, sparse data. Pushing AI to operate reliably under those constraints will surface techniques — from uncertainty quantification to human-AI decision protocols — that are directly relevant to making more general systems trustworthy in high-stakes environments.