AI material simulation: from theory tool to deployment engine
AI material simulation is the use of machine learning and physics-based models to predict how atoms and molecules will behave so that engineers can design, test, and optimize new materials in software before committing to slow and expensive physical experiments. This shift is reshaping how next-generation materials reach the market. Instead of screening candidates one by one on shared supercomputers, researchers can now explore thousands of options in parallel and refine only the most promising structures in the lab. The result is compressed R&D timelines and lower development risk, because performance, stability, and manufacturability are considered earlier. In fields from specialty coatings to CO2 capture technology, this approach is starting to bridge a long-standing gap between academic discovery and commercial deployment, where promising materials often stalled due to the cost and time of traditional testing.
MOFs reach industry: from millions of possibilities to real products
Metal Organic Frameworks (MOFs) are crystalline structures with large internal surface areas that can be tuned to trap or separate specific molecules, making them strong candidates for CO2 capture technology, refrigerant recycling, deodorization, and advanced coatings. Their challenge has never been a lack of potential but an excess of it: there are millions of possible MOF structures, and testing them experimentally is slow and costly. Companies from large chemical groups to specialized firms have produced MOFs on a commercial scale for carbon capture or focused on monolithic MOF formats, yet widespread MOF commercialization has lagged. One reason, as Atomis CEO Daisuke Asari notes, is that researchers have often chased performance without fully weighing durability, cost, and scalability. AI-driven screening is starting to reorder these priorities by letting teams evaluate profitability and manufacturability alongside molecular performance from the earliest design stages.
Atomis and Matlantis: compressing the MOF development cycle
Atomis shows how computational chemistry can shorten the path from MOF idea to industrial product. Instead of a linear workflow where experimentalists wait days or weeks for supercomputer slots, Atomis uses the Matlantis AI platform to run simulations in hours and fold results into ongoing lab work. Matlantis uses a machine learning interatomic potential based on graph neural networks, trained on millions of density functional theory (DFT) calculations, including unstable and non-equilibrium structures. This lets the system model unknown materials and complex reactions with useful accuracy. Atomis screens candidate MOFs in silico, filters them using both performance and cost data, and sends only the best options to DFT validation and physical synthesis. Researchers found that Matlantis and traditional supercomputer outputs were “not significantly different,” giving them confidence to rely on AI material simulation for rapid iteration across use cases such as ton-scale CO2 separation and refrigerant recovery.
From lab data to factory scale: building a MOF commercialization pipeline
For MOF commercialization, modeling performance is only half the job. Atomis builds a database that combines structural information with practical factors such as durability and cost, then uses Matlantis and other tools to narrow candidates before scaling synthesis. The company targets mass-market applications, including CO2 separation at the ton scale, to drive down production costs and prove mass manufacturing is feasible. This stands in contrast to players that focus on specific sorbents or niche monolithic formats. Python support within Matlantis enables scripted, continuous simulation campaigns under many conditions, which speeds data accumulation and helps experimental scientists with limited simulation experience participate. Visual molecular simulations also support customer conversations by showing how gas molecules move into adsorption sites and through pores, turning abstract computational chemistry outputs into concrete proof that a proposed MOF can work in real industrial settings.
A template for other emerging materials
The Atomis–Matlantis model hints at a broader shift in how emerging materials move from research to adoption. Many promising candidates—porous adsorbents, catalytic surfaces, next-generation coatings—are stuck in academic pipelines because exploring their design space experimentally is too slow. AI material simulation and cloud-based atomistic platforms can reverse that bottleneck by making it practical to screen huge material libraries, explore non-intuitive structures, and balance cost, durability, and performance from the outset. Once a workflow is in place, the same infrastructure that supported MOF commercialization can be redirected to other families of materials, using shared databases, automated Python workflows, and visual simulations to keep stakeholders aligned. As these tools spread, the distance between a published material discovery and a deployable product could shrink from decades to years, re-shaping how industrial sectors adopt CO2 capture technology, new refrigerant systems, and specialty surface treatments.
