4Point AI

Market position

4Point AI describes itself as a frontier technology lab in the AI mineral discovery and spatial intelligence space, listing industries including artificial intelligence, construction, energy, environmental engineering, geospatial, geothermal energy, government, machine learning and mining technology. Its Techstars profile records the company as pre-seed with 1-10 employees. It cites incubation through Techstars Space Accelerator, Google for Startups Canada and Creative Destruction Lab's Minerals stream, investment associations with J.P. Morgan, Spatial Capital and ID3 Ventures, and procurement relationships with the U.S. Department of Energy, the Saudi Arabia Geological Survey and the KSA Ministry of Industry and Mineral Resources.

The company positions its approach as moving uncertainty management upstream in the exploration sequence rather than treating it as a late-stage step: SIIM is said to separate structural signal from sampling history at the target-generation stage, RL-trained campaign ranking prioritizes information gain before capital commitment, GNN-hybrid modeling preserves spatial and depth coherence across faults and anisotropy instead of smoothing over discontinuities, a single intelligence stack links geology to downstream metallurgical and processing behavior, and physics-informed realizations replace a single deterministic map with inspectable scenario sets and uncertainty volumes.

Technology

Core components include SIIM, a proprietary Spatially Informed Intelligence Model that separates geological signal from drill-density bias; a GNN-hybrid model stack pairing graph reasoning over spatial dependencies with sequential depth modeling for vertical coherence; reinforcement learning trained in simulated 3D terrains using remote sensing and surface geophysics for targeting in data-sparse areas; voxel-native 3D geological graph intelligence for fault-aware boundary refinement; and physics-informed inversion generating multiple probabilistic geological realizations with uncertainty volumes. Published benchmark figures include roughly 88-90% variance capture, clay-model R² above 0.97 internally and 0.875 externally, cyanide-leach prediction R² of 0.882 for gold and 0.932 for copper, RMSE of 0.0130-0.0140, MAPE under 7%, more than 89% of predictions within 10%, and 50-70% average error reduction versus baselines rising to 90% in sparse and boundary settings.

Go-to-market

The company markets directly to enterprise and government buyers through briefing requests and sector-specific deployment contacts on its website, supported by published benchmark results and a public benchmark page. Accelerator and ecosystem programs — Techstars Space Accelerator, Techstars Los Angeles, Google for Startups Canada and Creative Destruction Lab's Minerals stream — form part of its visibility and network channel. The joint venture arm adds a project-origination channel aimed at junior explorers, prospect generators and holders of inactive or non-core mineral properties.

Investors, mining operators and governments needing decision-grade geological insight, with named deployment segments in mining, oil and gas, construction and geotechnical work, government strategic resource intelligence, and earth science research. Its joint venture program targets junior explorers, prospect generators, distressed or underfunded projects, and holders of abandoned or non-core mineral assets.

Geography

Headquartered in Toronto, Canada, with locations listed as West University, Austin, Texas; Toronto, Ontario; and Culver City, California. Techstars listed the company as Los Angeles, CA / Toronto, Canada during its 2024 accelerator cohort. Named customers and procurement relationships span the United States and Saudi Arabia, and the joint venture screening program targets Nevada, Utah and Arizona.

History

The company was founded in 2024 according to its Techstars profile, with a technical roadmap that begins in late 2023 with baseline geostatistical and classical machine-learning workflows. In September 2024 it was named to the Fall 2024 class of the Techstars Los Angeles Accelerator powered by JP Morgan, listed with locations in Los Angeles and Toronto. Successive model generations followed: SSAI 2.0 with edge-weighted GNN attention and sequential RNN modeling in mid-2024, reinforcement-learning integration in early 2025, voxel-native Lithology Expansion AI in mid-2025, unified surface-plus-subsurface targeting in late 2025, and physics-informed multi-realization modeling in early 2026. The company has also been referenced in a 2025 Google Economic Impact Report and BetaKit coverage, and in a 2026 StartUs-Insights emerging AI solutions listing.