Skip to main content

Physna

Category: AI Search / Retrieval

Physna is a geometric deep-learning and 3D search company that converts CAD, scans, meshes, and point clouds into normalized, machine-ready geometry to power "Physical AI" search and 3D parts data training. Physna was founded in 2015. The company is led by Paul Powers. Based in Columbus, Ohio, USA. Team size: 11-50. Total funding raised: $86M. Latest round: Series B. Key investors include Tiger Global Management, Sequoia Capital, GV (Google Ventures), Drive Capital.

Founded
2015
Headquarters
Columbus, Ohio, USA
Team size
11-50
Total funding
$86M

Value proposition

Physna's geometry engine transforms 3D data into canonical, orientation-invariant, topology-consistent representations that capture orders of magnitude more usable training signal than images, text, or raw 3D formats — enabling AI models to understand how real parts are shaped, how they function, and how they fit together.

Products and solutions

Thangs (free-to-use public geometric search engine and 3D model community), enterprise geometric search platform for 3D model search/compare/analyze, Physical AI search and normalization engine (API licensing for AI training on normalized 3D geometry).

Unique value

The "missing training signal for the physical world of AI" — Physna-normalized 3D geometry delivers ~10,000× more effective training signal per sample vs raw 3D, ~1,000,000× vs 2D images, and ~1,000,000,000× vs text, unlocking physical concepts (structural/functional similarity, manufacturability, assembly, supplier/variant equivalence) without manual labels.

Target customer

AI labs, OEMs, industrials, frontier tech companies, and government/defense agencies (e.g., U.S. Department of Defense) needing 3D parts data search, analysis, and AI training.

Industries served

Manufacturing, robotics, aerospace & defense, supply chain, design/engineering, additive manufacturing (3D printing).

Technology advantage

Proprietary geometric deep-learning engine that converts CAD, scans, meshes, and point clouds into canonical, orientation-invariant, topology-consistent geometry; "Object DNA" technology for 3D model comparison and search; patent-pending software for analyzing thousands of 3D models.

How they differentiate

Unlike traditional CAD/PLM vendors, Physna is purpose-built around geometric deep learning and 3D search rather than modeling/simulation, and it now positions its normalized-geometry engine as the foundational training signal for Physical AI (robotics, digital twins, factory automation) — a data-centric approach distinct from incumbent CAD toolmakers.

Main competitors

PTC (Creo/Onshape), Autodesk, Ansys, PhysicsX

Key partnerships

Palantir (strategic partnership to revolutionize 3D data analysis for defense and commercial sectors), U.S. Department of Defense (customer).

Notable customers

U.S. Department of Defense, enterprise OEMs and industrial customers (30+ customers)

Major milestones

Founded 2015, launched Thangs geometric search engine (2020), raised $86M total across Series A (2019) and Series B (2021), announced Palantir strategic partnership, launched Physical AI search and early-access partner program (Jan 2026).

Growth metrics

~$9.8M estimated ARR and ~30 customers (2024, per Latka); Thangs hosts millions of 3D models; U.S. DoD federal contractor (~$6.47M obligations).

Market positioning

Leader in geometric deep-tech / "Physical AI" — the world's leading geometric search engine, with a B2B SaaS model and a free consumer-facing Thangs community that drives adoption and data network effects.

Geographic focus

United States (Columbus, Ohio HQ), with defense and commercial customers globally.

Patents and IP

Patent-pending geometric comparison/analysis software (per Forbes profile).

About Paul Powers

Serial entrepreneur; law degree from Ruprecht-Karls-Universität Heidelberg (Heidelberg University). Forbes 30 Under 30 honoree. Co-founded Physna (geometric deep-learning / 3D search).

Latest news about Physna

More AI Search / Retrieval companies

Official website: