COMPANY PROFILE / MATERIALS INFORMATICS
Dassault Systèmes
— Bringing public machine-learning force fields into Materials Studio, a long-established materials simulation package
Dassault Systèmes is a large French software company known for CATIA and SOLIDWORKS. Its materials work sits in BIOVIA, the brand for life sciences and materials research, whose flagship is Materials Studio, software for calculating materials at the atomic and molecular level.
The latest release (2026), published in December 2025, adds the externally developed machine-learning force field MACE, and a way to fine-tune it on your own data. This profile ignores the company as a whole and focuses on its materials software: what it provides, where it takes technology and data from, and what we could not confirm. Revenue for BIOVIA or for materials is not disclosed.
- Dassault Systèmes in 30 seconds, and where it stands in materials informatics
- The logo and how the name is written
- Where the company sits in the materials informatics field
- Scale (the whole company in brief)
- Resources committed to materials informatics
- What it has done so far
- What comes next
- What the company is aiming for in materials informatics
- Map of materials informatics partnerships
- Technology, products and services
- The risks this company carries
- Glossary
- References
- Claim-to-source audit
1. Dassault Systèmes in 30 seconds, and where it stands in materials informatics
France10 rue Marcel Dassault
(primary)Materials work sits under the BIOVIA brand
Materials informatics (MI) is a way of finding new materials by collecting materials data, using AI or computation to predict and design properties, and then checking the result by experiment. Dassault Systèmes sits on the “predict and design” step with software that calculates at the atomic and molecular level Sourced. Unlike an AI company building a new model, it takes machine learning in by building externally published machine-learning force fields into long-established calculation software Sourced.
But the company as a whole covers design, manufacturing and simulation broadly, and materials is a part of it. The URD says only that the Life Sciences & Healthcare sector (which includes Medidata and BIOVIA, among others) is 21% of non-IFRS software revenue; the amount for BIOVIA or materials is not disclosed Not yet confirmed.
2. The logo and how the name is written
A note on names: the company is written Dassault Systèmes (with a grave accent on the è), and its legal form is SE (a European company). BIOVIA is the brand for the company’s life sciences and materials work, not a separate company, and Materials Studio is a BIOVIA product. This article writes “Dassault Systèmes” for the whole company and “BIOVIA” for the materials-facing brand.
Official site: https://www.3ds.com/3. Where the company sits in the materials informatics field
Even when two companies both say “AI”, they differ in where they sit along the flow of collect data → predict and design → check by experiment. Here is BIOVIA placed on that flow.
For quantum calculation, see our explainer “Fundamentals of Density Functional Theory”; for molecular dynamics, “Molecular Dynamics”; and for machine-learning force fields, “Machine-Learning Interatomic Potentials” and “Multiscale Simulation and Digital Twins”.
4. Scale (the whole company in brief)
Only the company-wide scale is given here Sourced.
| Item (IFRS, € million) | FY2024 (to Dec 2024) | FY2025 (to Dec 2025) |
|---|---|---|
| Revenue | 6,213.6 | 6,235.8 |
| of which software revenue | 5,613.3 | 5,641.0 |
| Operating income | 1,359.6 | 1,354.3 |
| R&D expense | 1,286.2 | 1,323.3 |
| Item | Detail |
|---|---|
| Employees | About 25,000 (31 December 2025). R&D staff number 9,968 (about 40%), and more than 1,000 hold PhDs |
| Sites and customers | 188 sites worldwide; more than 390,000 customers (URD) |
| Headquarters | Vélizy-Villacoublay, France |
| Listing | Euronext Paris; a CAC 40 member |
| Software revenue by sector (non-IFRS) | Manufacturing Industries 71%, Life Sciences & Healthcare 21%, Infrastructure & Cities 8% (2025) |
These figures are for the whole company. They do not show the scale of its materials business. The 21% for Life Sciences & Healthcare includes clinical-trial businesses such as Medidata, and no breakdown for BIOVIA or materials is given. The €1,323.3 million of R&D is also company-wide. Revenue, R&D expense and headcount for materials are all undisclosed, and this article does not estimate them Not yet confirmed.
5. Resources committed to materials informatics
Materials headcount and budget are not disclosed. What the official sources show is set out below, including items we could not confirm.
| Type of resource | What we found | Type |
|---|---|---|
| People | Company-wide R&D staff number 9,968 (about 40%), with more than 1,000 PhDs. Dedicated BIOVIA or materials headcount is not disclosed | Sourced (materials not disclosed) |
| Money | Company-wide R&D expense was €1,323.3 million in 2025 (the company received €41.8 million in government grants). No materials breakdown is given | Sourced (no breakdown) |
| Equity stakes, acquisitions, joint ventures, divestitures | Acquisition: it bought Accelrys in 2014 and created the BIOVIA brand (URD history). We found no investment in materials start-ups or new acquisition in the URD | Sourced (others not found) |
| Universities, research institutes and public funding | The software takes in academic versions of codes (GULP 6.4, ONETEP 7.3.95 and others) and uses an API for searching CCDC databases (release note). The URD describes a research project in which a glass packaging maker and a consultancy used BIOVIA to lighten packaging (no figures). Public funding was not found | Sourced (public funding not found) |
Status of the main products
This is what the company’s official material lets us confirm. Statements about performance are the company’s description and were not reproduced or verified for this article.
| Name | Kind | Content (published specifications, target, conditions) | Release form and license | Stage | Date | Source |
|---|---|---|---|---|---|---|
| BIOVIA Materials Studio | An integrated environment for atomic and molecular simulation of materials | For researchers on catalysts, polymers, composites, metals and alloys, pharmaceuticals, batteries and more. Takes an “in silico first” approach and says it reduces physical experiments (company description). More than 25 years of history | Commercial. Price and license terms not confirmed | Offered | 2026 release, 1 Dec 2025 | Reference 3, 4 |
| New in Materials Studio 2026 | Version update | Python added to MaterialsScript, MACE fine-tuning, reaction-path search with FlexTS, a new viscosity workflow, faster CASTEP (ACE), GPU support for DFTB+ and more | Commercial | Offered | 1 Dec 2025 | Reference 5, 4 |
| MACE-family machine-learning force fields | Machine-learning interatomic potentials | MACE-OFF23 (for molecules) and MACE-MP-0 (for solids) were added in the 2025 release. In 2026, MACE-MP-0 was replaced by MACE-MP-0b3, extended to a full foundation model, and MACE-MATPES-r2SCAN-0 was added. The datasheet describes the latter as trained first on Meta’s public OMat24 data and fine-tuned on MatPES (a recalculation of part of the Materials Project with the r2SCAN method). We did not check the models’ licenses | Built into the product. The models were released outside the company | Offered | 2025 and 2026 releases | Reference 4 |
| Pipeline Pilot Materials Studio Collection | Workflow foundation | The “MACE Fine Tuning” protocol fine-tunes MACE on your own data (a Study Table) | Commercial | Offered | 1 Dec 2025 | Reference 4, 5 |
| Materials Studio Simulation on 3DEXPERIENCE | Cloud edition | Builds, simulates and analyzes atomistic and mesoscale models in a secure, collaborative cloud environment | Commercial | Offered | No date stated | Reference 3 |
6. What it has done so far
| When | What happened | Why it matters |
|---|---|---|
| 1981 | Dassault Systèmes was founded (per the datasheet) | Began as a design-software company, known for CATIA and the like |
| 2014 | It acquired Accelrys and created the BIOVIA brand (URD) | Brought in software for computation and data management in materials, chemistry and biology |
| 2015 | Changed its legal form from SA to SE, a European company (URD) | A change of corporate form |
| 2025 release | Materials Studio introduced the MACE machine-learning force fields (MACE-OFF23 and MACE-MP-0) (datasheet) | The first release to bring machine-learning force fields into the product |
| 2025 | The URD says it introduced three “AI-native” categories (Virtual Companions, Generative Experiences, Virtual Twins as a Service) | Company-wide AI strategy. For BIOVIA it says they can also be used to search for new molecules and materials (a statement of strategy) |
| 1 Dec 2025 | Materials Studio 2026 released | Python support, MACE fine-tuning, reaction-path search and more |
| 31 Mar 2026 | The 2025 Universal Registration Document was filed | The primary source for company scale and strategy |
7. What comes next
tangible value”A strategic statement in the URD; not a plan specific to materials
The URD says that in 2026 the company will turn AI into tangible value for all industries Not yet confirmed, but it gives no specific schedule or numerical target for materials.
None of the following is stated in the official sources we checked Not yet confirmed.
The timing of the next Materials Studio release; revenue, headcount and investment for BIOVIA or materials; any plan to develop machine-learning force fields in-house; and the license terms of the external models it has built in. Press-based talk of BIOVIA’s materials business growing by some percentage is not used here.
8. What the company is aiming for in materials informatics
This part includes things the company does not say directly. It is our reading of the primary sources, written with the evidence for each point.
| What we read from it | Primary evidence | Type |
|---|---|---|
| For machine-learning force fields the policy is to take in published models and data rather than build from scratch | The datasheet builds its description of the product’s core around MACE (citing outside papers), Meta’s OMat24 and MatPES | Inference |
| Machine-learning force fields are added as one more tool alongside the existing quantum and classical calculations | Forcite can use MACE, while existing modules such as CASTEP were also improved in parallel (GPU support, speed-ups) | Inference |
| Materials research is framed as a “virtual twin” | The URD describes BIOVIA as simulating biology and materials R&D virtually before physical trials | Sourced (the company’s stated positioning) |
| It emphasizes users growing machine-learning force fields themselves | Python support and the MACE Fine Tuning protocol are headline features of this release | Inference |
The datasheet describes MACE-MATPES-r2SCAN-0 as a model aiming to be a general force field for solid-state systems, trained on Meta’s public data Sourced. From this we read Dassault Systèmes’ strength as lying not in developing machine-learning force fields itself but in putting the latest outside models into the product in a usable form (interface, workflows, fine-tuning) Not yet confirmed. In that case, progress in the machine-learning part depends on updates to outside models.
9. Map of materials informatics partnerships
10. Technology, products and services
How machine-learning force fields are used in Materials Studio
Materials Studio is an atomic-level simulation environment that brings together quantum-mechanical calculation (CASTEP, DMol3, ONETEP and others) and classical force-field calculation (Forcite, COMPASS III, GULP and others). A machine-learning force field (MLIP) learns from quantum data and aims to treat bond breaking and chemical reactions much as quantum calculations do, but in far less time (the datasheet’s description).
The datasheet says the new release can do the following Sourced (all are the company’s descriptions).
| Feature | Content | How to read it |
|---|---|---|
| Automation with Python | Python joins MaterialsScript, and Python machine-learning libraries can be used | The company calls it the first Python support in over 25 years |
| MACE fine-tuning | Start from a foundation model and train further on your own data | The datasheet gives no figure for how much accuracy improves |
| Reaction paths and transition states | Forcite’s FlexTS with MACE computes barriers for polymerization, catalysis and reactions in a battery cell | Only an example of lithium diffusion in silicon (no values given) |
| Dispersion (van der Waals) correction | Models such as MACE-MP-0b3 were trained without dispersion information, so a D3 correction is applied alongside | The effect is shown for the volume contraction of NMC811 on charging (no values given) |
| Viscosity calculation | A new workflow averages over several starting states and fits a double exponential, moving from an estimate to a “prediction” | No figures for accuracy or run time |
Three points in the Materials Studio approach matter to a materials engineer Sourced. (1) A public foundation model is a starting point: MACE-MATPES-r2SCAN-0 aims to be general, but for your own system (a particular alloy, polymer or interface) the premise is to fine-tune it or to check with quantum calculation. (2) Dispersion correction: according to the datasheet, some solid-state models were trained without dispersion forces, so layered materials and molecular crystals need a correction such as D3. Results change with or without it, so check the settings.
(3) Calculation is not a substitute for experiment: the company says to “try it by calculation first”, but the examples shown are few and the datasheet gives no figures for accuracy. Use results by comparing them with your own measurements. Outside the chemistry covered by the training data (part of the Materials Project, Meta’s inorganic materials data and so on), accuracy may fall (a general property, not individually confirmed in company material).
11. The risks this company carries
| Risk | Detail | Type |
|---|---|---|
| The scale of materials is invisible | Sales, headcount and investment for BIOVIA and materials are undisclosed. The Life Sciences & Healthcare sector also includes Medidata and others | Sourced |
| The ML core relies on outside models | MACE and similar models are built outside. Updates and changes of terms are up to others. We did not check license terms | Inference |
| Performance claims are only illustrated | The datasheet’s examples (lithium diffusion in silicon, volume contraction of NMC811) have no values or conditions, and we found no independent verification | Sourced |
| Rivals exist | Schrödinger’s 10-K names BIOVIA as a rival in materials science. This article does not compare the two | Sourced |
| Depends on versions of academic codes | It takes in versions of academic codes such as CASTEP, ONETEP and GULP. The pace of updates is affected by outside development | Inference |
| Gap between strategy and results | The URD’s “AI-native” categories are statements of strategy, and there are no results figures for materials | Inference |
| Price and terms are not visible | License terms and prices could not be confirmed on official pages | Not yet confirmed |
12. Glossary
- Materials informatics (MI)
- Finding new materials with materials data plus AI and computation, in the flow of data, then predict and design, then experiment.
- BIOVIA
- Dassault Systèmes’ brand for life sciences and materials research. Created in 2014 with the Accelrys acquisition.
- Materials Studio
- BIOVIA’s materials simulation software. It brings calculations from quantum to classical into one environment.
- Machine-learning force field (MLIP)
- A machine-learning model trained on quantum calculations that gives forces and energies between atoms quickly.
- MACE
- A machine-learning force-field method developed outside the company. MACE-OFF23 is for molecules, MACE-MP-0b3 a foundation model for solids.
- OMat24 / MatPES
- Public datasets used to train machine-learning force fields. OMat24 is from Meta; MatPES recomputes part of the Materials Project with the r2SCAN method.
- CASTEP, DMol3, ONETEP
- Calculation modules for quantum mechanics (density functional theory).
- Forcite
- A module that runs molecular dynamics and structure optimization with classical or machine-learning force fields.
- Dispersion (van der Waals) force
- A weak attraction between molecules or layers. DFT calculations need a correction for it.
- Pipeline Pilot
- BIOVIA’s workflow platform, which links steps of data processing and calculation.
- Virtual twin
- A virtual counterpart of a product or material recreated inside a computer.
- Universal Registration Document (URD)
- The annual report that listed companies in Europe file.
13. References
- Dassault Systèmes 2025 Universal Registration Document (the annual report, filed 31 March 2026; English translation). Company profile, revenue, R&D expense, headcount, BIOVIA’s positioning and company history. The French version is the legally binding one. https://investor.3ds.com/static-files/ce30eb59-e3df-4f57-af64-d7f105bb3cc3
- Dassault Systèmes News release “Filing of the 2025 Universal Registration Document” (31 March 2026). https://investor.3ds.com/news-releases/news-release-details/dassault-systemes-filing-2025-universal-registration-document
- Dassault Systèmes Official website, BIOVIA Materials Studio product page. Positioning and application areas. https://www.3ds.com/products/biovia/materials-studio
- Dassault Systèmes Datasheet “What’s New in Materials Studio 2026” (document DS-10079-0626). New features: MACE, Python, FlexTS, viscosity calculation and more. https://www.3ds.com/assets/invest/2026-06/biovia-materials-studio-whats-new-2026.pdf
- Dassault Systèmes Release note T61-2025 “Materials Studio 2026” (1 December 2025). The individual enhancements. https://www.3ds.com/support/documentation/resource-library/t61-2025-materials-studio-2026
- Schrödinger Form 10-K for the year ended 31 December 2025. Names BIOVIA as a rival in materials science. A third party’s filing. https://www.sec.gov/Archives/edgar/data/1490978/000149097826000010/sdgr-20251231.htm
- Dassault Systèmes Official website (source of the logo; used to check the company profile). https://www.3ds.com/
14. Claim-to-source audit
| Claim in the text | Type | Source |
|---|---|---|
| Legal name (SE), headquarters, listing (Euronext Paris, CAC 40) and the 2015 change from SA to SE | Sourced | Reference 1 https://investor.3ds.com/static-files/ce30eb59-e3df-4f57-af64-d7f105bb3cc3 |
| Revenue, software revenue, operating income and R&D expense (2024 and 2025, IFRS) | Sourced | Reference 1 https://investor.3ds.com/static-files/ce30eb59-e3df-4f57-af64-d7f105bb3cc3 |
| About 25,000 employees, 9,968 in R&D, over 1,000 PhDs, 188 sites, over 390,000 customers, and the software revenue split by sector | Sourced | Reference 1 https://investor.3ds.com/static-files/ce30eb59-e3df-4f57-af64-d7f105bb3cc3 |
| The URD filing date (31 March 2026) | Sourced | Reference 2 https://investor.3ds.com/news-releases/news-release-details/dassault-systemes-filing-2025-universal-registration-document |
| The 2014 Accelrys acquisition and creation of the BIOVIA brand | Sourced | Reference 1 https://investor.3ds.com/static-files/ce30eb59-e3df-4f57-af64-d7f105bb3cc3 |
| BIOVIA’s positioning (virtual twins of biology and materials), the AI-native categories, and the glass packaging research project | Sourced | Reference 1 https://investor.3ds.com/static-files/ce30eb59-e3df-4f57-af64-d7f105bb3cc3 |
| Materials Studio’s positioning, application areas and cloud edition | Sourced | Reference 3 https://www.3ds.com/products/biovia/materials-studio |
| New features of Materials Studio 2026 (Python, MACE fine-tuning, FlexTS, viscosity, dispersion correction) and the descriptions of each MACE model | Sourced | Reference 4 https://www.3ds.com/assets/invest/2026-06/biovia-materials-studio-whats-new-2026.pdf |
| The 2026 release date (1 December 2025), individual enhancements, academic code versions (GULP 6.4, ONETEP 7.3.95) and the CCDC search API | Sourced | Reference 5 https://www.3ds.com/support/documentation/resource-library/t61-2025-materials-studio-2026 |
| Schrödinger’s Form 10-K names BIOVIA as a rival in materials science | Sourced | Reference 6 https://www.sec.gov/Archives/edgar/data/1490978/000149097826000010/sdgr-20251231.htm |
| Placement on the three steps of MI (data, predict and design, experiment) | Our calculation | Our own grouping from public sources (this article) |
| BIOVIA and materials revenue, dedicated headcount, investment and the R&D breakdown | Not yet confirmed | None could be confirmed, and this article does not estimate them (not stated in the sources checked for this article) |
| Investment in materials start-ups, new acquisitions and public funding for materials | Not yet confirmed | Not stated in Reference 1 as checked for this article; we do not assert that none exist |
| The next release date, numerical targets for materials, license terms of external models, and prices | Not yet confirmed | Could not be confirmed; this article does not guess |
| Reading: the policy is to take in published models and data | Inference | Our own interpretation of the datasheet description |
| Reading: machine-learning force fields are added as one more tool to existing calculation | Inference | Our own interpretation of the feature set |
| Reading: it emphasizes users growing machine-learning force fields themselves | Inference | Our own interpretation of the new features |
| Risk: reliance on outside models for the ML core | Inference | Our own interpretation of the product composition |
| Risks: dependence on academic code versions, and the gap between strategy and results | Inference | Our own interpretation of the product composition and the URD |
Last updated 22 October 2026 / Troy Technical
The figures in this article rest on Dassault Systèmes’ 2025 Universal Registration Document, its official product pages, datasheet and release notes. Market-research estimates and press-only figures are not used. Statements of performance or effect by the company were not reproduced or verified here. Items marked “Inference” are our own reading of the primary sources and are not statements by the company.