Neural Concept
Lausanne-based engineering AI platform that connects CAD, CAE and PLM and predicts physics in near real time from your own simulation data
Quick Overview
| Company | Neural Concept SA |
|---|---|
| Category | Engineering Simulation (CAE/FEA) |
| Headquarters | Lausanne, Switzerland |
| Founded | 2019 |
| EU Presence | Yes - Switzerland |
| Data Location | Customer choice: AWS, Azure, on-premise or own HPC |
| Open Source | No |
| GDPR Compliant | Yes |
| Self-Hosting | Yes - on-premise or own HPC |
| Main Features | Geometry Intelligence, Physics Intelligence, Program Intelligence, Design Copilot, Design Studio, connectors for CAD, CAE and PLM, campaign manager |
| Pricing | Pricing on request |
| Best For | Large engineering organisations with years of simulation data that want faster design iteration |
| Replaces | Ansys design exploration and optimisation workflows |
Detailed Review
Neural Concept does not replace a solver. It sits on top of CAD, CAE and PLM systems, trains physics-aware models on a company's past simulations, and predicts performance of new designs in near real time, with the customer's trusted solvers used for validation.
The privacy policy names Neural Concept SA at the EPFL Innovation Park in Lausanne as the data controller. The company page says it was born in 2019 from an AI research lab in Switzerland, with headquarters in Lausanne and offices in Paris, Munich, Zurich, Detroit, San Francisco and elsewhere.
Three Intelligences
Geometry Intelligence covers CAD understanding, variant generation and simulation-ready conversion. Physics Intelligence provides fast performance prediction, design comparison and optimisation. Program Intelligence stores design rules, constraints and past decisions so agents follow engineering intent.
These sit behind four interfaces: an AI IDE extension for automation experts, a Design Copilot, a Design Studio and apps for the wider organisation.
Deployment and Data Ownership
The platform can be deployed on AWS, Azure, on-premise or the customer's own HPC, with GPU and Kubernetes support. The vendor states that models built from customer data remain the customer's.
Neural Concept lists ISO/IEC 27001:2022, audited by a third party.
Where It Fits
Case studies on the site name automotive, aerospace, energy, consumer electronics and semiconductor customers. The value comes when many similar designs have already been simulated.
Teams with little historical data gain less, because the models need examples to learn from.
Limitations to Consider
Neural Concept is not a simulation solver: it depends on existing simulation data and your own solvers, which in many cases means Ansys itself. Pricing is on request and aimed at enterprises, and the company also has a large US presence (Detroit, San Francisco, New York), so confirm the contracting entity is the Swiss SA. Consider PhysicsX as a comparable approach from the UK.
Who Should Use Neural Concept
Automotive, aerospace and industrial engineering organisations that want to accelerate design exploration and keep data under their control.
Pros and Cons
Pros
- Works with existing CAD, CAE and PLM tools
- On-premise or own-HPC deployment
- ISO/IEC 27001:2022 audited
- Customer owns trained models
- Swiss entity named in privacy policy
Cons
- Not a solver, needs prior simulation data
- Pricing on request
- Enterprise focus
- Significant US operations
Alternatives to Neural Concept
Other European engineering simulation (cae/fea) worth comparing before you decide:
SimScale
Germany · Cloud simulation for CFD, structural and thermal analysis in a browser
Quanscient
Finland · Cloud multiphysics solver with all physics and unlimited users
COMSOL
Sweden · Multiphysics platform with one interface and add-on modules
code_aster
France · Open-source structural and thermal solver from EDF R&D
See all European engineering simulation (cae/fea) or compare them directly against Ansys design exploration and optimisation workflows on our alternatives page.
Frequently Asked Questions
No. It predicts performance from existing simulation data and uses your solvers for validation, so it complements a solver such as Ansys, COMSOL or Abaqus rather than replacing it.
Yes. The platform page lists AWS, Azure, on-premise and customer HPC deployment.
The vendor states that what the platform learns from your data, and every model built from it, stays yours.