Should you build or buy Materials Informatics?
Materials informatics software applies machine learning to a company's own experimental materials data, helping formulation and R&D teams predict properties, screen combinations, and cut lab trials before physical testing. Buyers are chemical, materials, and manufacturing companies with proprietary formulation data they want to turn into faster product development.
Bridge it. License a platform for data capture, schemas, workflow and reference data, and keep your own models, feature engineering and experimental design in-house where the proprietary advantage lives. Prove the platform on a real pilot: import lab data, preserve scientific context, and test a model on the decision it is meant to inform before committing.
Build it, buy it, or bridge?
When building makes sense
Materials informatics is one of the few categories where the data itself, not the workflow around it, is the real prize. A company's own experimental results, from years of formulation trials, are what any predictive model needs to be useful, and that data cannot be bought from a vendor. AGC built its ARDIS data platform and AMIBA analysis tool in-house and moved them into full-scale R&D use in June 2022, and Sekisui Chemical built a similar in-house database using Hitachi's proprietary technology to organize its own experimental data. Both are large chemical companies that took years and dedicated internal expertise to get there. Build your own models and feature engineering on top of a captured, well-organized dataset once you have one. The proprietary advantage lives in what you predict from your own data, not in the plumbing that stores it.
When buying makes sense
Buying the platform layer is the practical starting point for most teams. Ansys Granta MI Enterprise gives access to more than 250,000 materials in its reference database, plus restricted-substance checks against more than 70 million safety data sheets. Rolls-Royce built its Granta MI deployment into a globally accessible 'gold source' used by more than 2,000 engineers, with certified savings of approximately $10 million a year from reduced duplicate testing and waste, and Baker Hughes adopted the same platform to migrate all of its material data into one digital source. Citrine's platform has run projects like a Siemens and SABIC collaboration that let a polymer producer offer customers faster masterbatch-to-polymer ratio recommendations by model prediction instead of physical trials. Buy the data capture, schema, and reference layer, and prove it on a real pilot before committing further.
The desk read
Build-versus-buy analysis for Materials Informatics is being written. In the meantime, the framework that drives every B4 call is on the B4 Index page.
Frequently asked
Has any real company saved money by building its own materials informatics platform?
Yes. AGC built its own materials informatics database (ARDIS) and analysis tool (AMIBA) in-house, moving them into full-scale R&D use in June 2022. Sekisui Chemical built a similar in-house database using Hitachi's technology to organize its own experimental data. Both are large chemical companies with dedicated data teams; a licensed platform like Ansys Granta MI is the more common path even for large industrials, and Rolls-Royce's own Granta MI deployment has produced certified savings of approximately $10 million a year.
What's the difference between the platform and the proprietary part in this category?
The platform is the data capture, schema, reference library, and workflow tooling, which a vendor like Ansys or Citrine already provides. The proprietary part is your own experimental data and the models you train on it, which no vendor can supply because it comes from your own lab history.
Can a vendor's platform actually reduce lab experiments?
Yes, in documented cases. A Citrine Platform project with Siemens and SABIC let a polymer producer's customers identify workable masterbatch and polymer ratios through model prediction, without running a physical tool trial for every combination they wanted to evaluate.
Is materials informatics only for large chemical companies?
The clearest wins so far come from large industrials like Rolls-Royce and Baker Hughes with dedicated materials data teams. Smaller R&D operations can still use a vendor platform's reference data and workflow tools, but the payoff scales with how much proprietary experimental data there is to build models on.