Sell CAD and engineering data to AI companies

Why AI companies want CAD data now

AI is moving from text into engineering. Text-to-CAD tools generate editable 3D models from a prompt, and AI labs are training agents meant to work inside CAD tools the way an engineer does. Both need to learn from real design work – how parts are built up feature by feature, how assemblies fit together and how a drawing documents a design.

Public sources don't cover that. Open research datasets such as ABC, DeepCAD and Fusion 360 Gallery are built from publicly shared, mostly simple models, without the drawings, bills of materials or test results that come with professional work. The complex assemblies, revision histories and complete project records that make up a real engineering archive almost never reach the open web.

That is why the CAD archive on your file server or PDM vault can now be licensed to AI companies. Selling CAD data to AI companies in practice means licensing it: the buyer gets a limited right to train on a selection of your files, and you keep the archive.

What buyers look for in a CAD archive

Not all CAD data is equally valuable. Roughly, from most to least in demand:

• Native parametric files with feature history – SolidWorks, Inventor, Creo, NX, Fusion. The build sequence is what text-to-CAD models learn from.
• Neutral B-rep files – STEP and IGES. Accurate geometry, but without the history.
• 2D drawings – DWG and DXF, especially when they belong to a 3D model.
• BIM models – native Revit models are worth more than IFC exports, which lose families and parametric relationships.
• PCB and electronics – native EDA projects and Gerber files.

Value multiplies when files come paired: a model with its drawing, BOM, GD&T, specifications, calculations, change orders and test or simulation results. Revision history and many years of varied projects push the value up further. There is more on this in how much is CAD data worth.

Who can license engineering data

• Hardware and product engineering firms with native CAD assemblies, PCB designs, simulation files and test results.
• Architects, engineers and AEC firms with Revit and BIM models, drawing sets and project records.
• Manufacturers and contract manufacturers with part models, CAM programs and production drawings.
• Firms shutting down or merging, where the archive would otherwise just be deleted – see shutting down a company.

Even a mid-sized firm often holds thousands of models from decades of projects. What matters is not just volume but how complete and varied the records are.

What a CAD archive can be worth

There are no public prices per file or per gigabyte. Licenses are priced per company or archive, based on format, pairing, history and how clean the rights are. Published buyer ranges give a sense of the level:

These are ranges reported by buyers and brokers, not offers. Many archives end up well below the top of the range, and some can be licensed to more than one buyer because licenses are non-exclusive by default. Our data value estimator gives a first estimate for your firm in about two minutes.

Rights, client contracts and export controls

The answers differ from case to case, but three questions usually decide whether, and which part of, a CAD archive can be licensed:

• Who owns the files? Employees' work usually belongs to the firm, but client contracts can assign copyright to the client, and work done by contractors may belong to them. Under AIA B101 the architect normally keeps copyright in the instruments of service. Read who owns CAD files for the details.
• What did you promise the client? NDAs, confidentiality clauses and client trade secrets can rule out projects even when you own the copyright. Vendor and library parts usually have to be removed.
• Is anything export-controlled? Defense and aerospace drawings can be ITAR technical data, and dual-use designs can be controlled under the EAR. Those projects are classified and reviewed separately. See preparing CAD files for an AI license.

Before delivery, title blocks, client names, site addresses, stamps and file metadata are removed. If names of people appear in the files, privacy law applies as well – see data privacy and AI training.

This describes how things generally work today, not legal advice. Every archive and every deal is unique – what can be licensed depends on your contracts and files, which is why each case is assessed on its own. And that's what we're here for – we do the groundwork with you, involve the right experts when needed and put good deals on the table.

How a CAD license works

1. You describe the archive in our form – which CAD systems, roughly how many models and drawings, how many years and which industries. It takes about two minutes and you don't upload any files.
2. We assess rights, export-control flags and demand, and get back to you within 48 hours on business days.
3. You approve a concrete offer and agreement – or say no. Nothing is shared until a non-disclosure agreement is in place.
4. Selection, clean-up and delivery. We work with our partners and the buyer to clean the agreed projects of identifiers in the best possible way to close the deal. Then the files are delivered and payment is made.

By default the license is non-exclusive and limited in scope, time and use, and you keep ownership of every file. It costs nothing to have your archive assessed – we only get paid when a deal goes through. More on data licensing and how it works.