Startup profile · funding coverage
Physicl funding, valuation and investors
Synthetic 3D training data for robotics and world models, focused on the rare scenes where models fail.
Verified company research · funding, investors and latest developments
Keep track of Physicl
Save this profile to your VCT watchlist for a quick return.
Funding, valuation & investors
Answer-first snapshotLatest funding
Seed (announced)
$10 million · November 6, 2026
Latest known valuation
Undisclosed
Seed (announced) · November 6, 2026
Total disclosed equity funding
$10M
Excludes debt, grants, acquisitions, secondaries, and IPO proceeds.
Current status
private
seed · Paris, France
Investors in latest funding
Lead: Inovia Capital
Other: U.S. Venture Partners
Overview
Physicl is a Paris-based physical-AI data company led by founder and CEO Alex de Vigan, who previously founded the 3D-commerce technology company nfinite. On November 6, 2026, Physicl announced a $10 million seed financing led by Inovia Capital with participation from U.S. Venture Partners. The original company-issued GlobeNewswire release does not give a valuation or identify a legal completion date separate from its announcement.
It also calls Insight Partners a backer, without specifically identifying a contribution to this seed round.
Physicl's first product, Studio, is in private beta. A robotics or world-model research team selects or describes a 3D environment that exposes a failure (for example a cluttered warehouse aisle), varies controlled factors such as light, objects or camera viewpoint, chooses annotations and render settings, approves the run, and receives image frames and geometry-derived ground-truth data in its connected storage bucket.
Output can include metric depth maps, surface normals, instance segmentation, camera poses and labeled bounding boxes. Its stated library exceeds 10,000 scenes and 100,000 objects; these are catalog claims, not numbers of paying customers.
The distinction is synthetic rendered training imagery versus a complete robot physics simulator. The company's own documentation notes that rendering images from a 3D scene does not simulate physical interactions forward in time.
More varied images and precise scene-derived labels can help with perception training, but do not alone establish that a real robot can manipulate an object safely, that a scene recreates real physics, or that generated distributions transfer to deployment conditions.
The company cites SpAItial as a collaborating world-model developer and says synthetic data more than halved its room-reconstruction error in a selected case. The methodology, baseline, test split and independent validation are not disclosed; this should not be interpreted as a universal improvement. The company also lists leading AI organizations as groups it works with, which does not establish that every logo is a paying customer.
Studio availability is limited to private-beta researchers/design partners, and public subscription or API prices, revenue and margin figures were not located.
Source documents: Physicl issuer release dated November 6, 2026 https://www.globenewswire.com/news-release/2026/11/06/3397485/0/en/physicl-raises-10-million-to-solve-physical-ai-s-edge-case-gap.html ; official Studio guide https://www.physicl.ai/studio ; independent original interview https://techcrunch.com/2026/11/06/physicl-raises-10m-to-make-the-digital-world-real-enough-for-robots/ .
Why Physicl is interesting
Its $10M November 2026 seed backs targeted post-training data rather than another general-purpose robotics foundation model. Studio lets teams control scene variation and export geometry-derived labels; physical-world transfer is not proven by a rendered demo.
Product & use cases
Buyer/first-visit tour (illustrative, based on official documentation): (1) identify a perception failure such as a robot missing boxes under low light; (2) describe/select a matching scene from Studio's 3D library; (3) choose controlled factors like camera height, clutter, lighting while holding others fixed; (4) select ground-truth layers and rendering volume; (5) approve a run and export labeled frames plus metadata into the connected cloud bucket; (6) compare model errors on held-out real and synthetic cases. Studio is private beta; this is not a hands-on product test. Pricing and contract terms are undisclosed.
- Robotics perception evaluation on rare warehouse, household and industrial edge cases.
- World-model spatial reconstruction training with labeled synthetic scenes.
- Repeatable model regression benchmarks with controlled changes to camera, layout and lighting.
Key facts
- $10M seed announced November 6, 2026: Inovia Capital lead; U.S. Venture Partners participated; valuation undisclosed.
- Studio private beta, with company-advertised 10,000+ 3D scenes and 100,000+ objects.
- Geometry-derived per-frame labels: depth, normals, segmentation, bounding boxes and camera pose.
- SpAItial's more-than-halved reconstruction error is a selected company/customer case, not independently audited proof.
Funding history (newest first)
Seed (announced)
2026-11-06 $10 million Valuation: Undisclosed- Inovia Capital (lead)
- U.S. Venture Partners (participant)
Source: Read source · globenewswire.com
Competitive landscape
Edge: Potential advantage: a large curated 3D asset library and repeatable, parameter-controlled renders whose depth/normals/segmentation labels come directly from geometry, reducing manual annotation. Claim of physical fidelity and generalization requires independent real-world validation. Unlike a full simulator, its documented Studio focuses on rendering labeled frames, not forward physics.
-
NVIDIA Omniverse/Isaac Sim adjacent
Full robotics simulation and synthetic-data workflows with physical simulation; may be more technically involved but addresses motion and dynamics beyond rendered frames.
-
Parallel Domain direct
Synthetic data and scenario generation for computer-vision/perception teams; compare environment fidelity and annotations.
-
Synthesis AI direct
Synthetic labeled data for computer vision with overlapping pipeline goals; specific modalities and access differ.
-
Real-world teleoperation/data collection alternative
Captures real conditions and interactions but rare events are expensive and challenging to label.
Market / IPO context
Editorial / static context — not a live quote.
FAQs about Physicl
Practical answers founders, operators, and investors typically search for.
Last updated:
Editorial note: AI tools assisted with research, structure, or drafting. Venture Capital Tracker retains human editorial responsibility for factual accuracy, relevance, and source quality before publication.