“What are the top physical ai companies focused on building better CAD flows?”
An analyst screen of the 12 strongest pure-play companies applying AI to professional CAD workflows, as of September 26, 2026 — ranked by product differentiation, funding and adoption evidence, source breadth, and direct relevance to mechanical and industrial CAD, not by market share. “Physical AI” here means AI operating on engineering geometry, drawings, design rules, and simulated physics — not robotics companies that use CAD internally. Incumbent suites (Autodesk, Dassault/SOLIDWORKS, Siemens, PTC, Ansys) are excluded from the ranking; they are the distribution layer these startups integrate with or compete against.
Each company placed at its primary layer of the CAD workflow, sized by disclosed total funding (a maturity signal, not audited fact). Hover or tap a card for product, differentiator, and traction. DraftAid and MecAgent have no disclosed amount and are drawn at minimum size.
Circle area = disclosed funding ($M) · company-reported or secondary-source figures · PhysicsX shown at its $32M Series A snapshot; newer 2026 evidence points to a later $300M Series C (see findings) · number in the corner is the analyst rank (1 = strongest screen score)
Rank is the analyst screen, not market share. Funding and traction statements are company-reported or secondary-source estimates — read them as signals.
| # | Company | Layer | What it does | Funding signal | Traction signal | Source |
|---|
Method: 12 companies ranked 1–12 by an analyst screen dated 2026-09-26, built from a broad discovery search merging thousands of live results across company sites, funding and news coverage, engineering publications, and technical sources; each ranked company had evidence across multiple independent hosts (13–191 per company). The 12 displayed rows link 10 distinct source hosts; no single URL backs more than 3 of 12 rows. Circle size uses disclosed funding in $M; undisclosed amounts drawn at minimum size. Funding and traction are company-reported or secondary estimates, not audited; PhysicsX’s row carries an older $32M snapshot. Long workflow and differentiator text is shortened for space; full detail appears in tooltips.