Matlantis
Matlantis is a Tokyo-based cloud SaaS platform that uses a proprietary universal machine-learning interatomic potential (PFP) to deliver fast, DFT-level atomistic simulations for materials R&D teams in automotive, chemicals, semiconductor, battery, and energy companies worldwide.
- Company typePrivate
- Founded2021
- HeadquartersTokyo, Japan
- Headcount11–50
- GTM typeB2B
- OfferingSoftware
What Matlantis does
Matlantis Corporation, headquartered in Tokyo and founded in June 2021 as Preferred Computational Chemistry, is a private cloud-based SaaS company that delivers atomistic materials simulation through its proprietary PFP (Preferred Potential) universal machine-learning interatomic potential. The platform lets chemists, materials scientists, and R&D engineers run DFT-level accuracy simulations of molecular dynamics, reaction pathways, crystal structures, and surface processes for systems of up to 44,000 atoms across all 96 naturally occurring elements — work that previously required either months of high-performance computing time or specialized per-system modeling. The product is delivered entirely through a browser-based cloud interface with regional data centers in Japan, the US, and Germany, removing the need for customers to provision their own HPC infrastructure.
The company's product surface has expanded from a single atomistic simulator into a portfolio of AI-driven capabilities: PFP as the core model, LightPFP for systems beyond 44,000 atoms, Matlantis CSP for crystal structure discovery, RestScan/ReactionString and GRRM20 integrations for reaction pathway analysis, and as of May 2026 an agentic layer that embeds Anthropic's Claude Code inside the Matlantis terminal. Matlantis monetizes through subscription-based cloud access and enterprise agreements, supplemented by a 110,000 JPY paid hands-on seminar training program, and serves more than 150 enterprise customers and academic institutions across automotive, chemicals, semiconductors, and energy — including Honda R&D, Toyota Battery, Hyundai, Volkswagen, GE Vernova, Tokyo Electron, Kuraray, AGC, and ENEOS. The company is jointly owned by Preferred Networks, ENEOS Corporation, and Mitsubishi Corporation (which joined as a strategic investor in June 2024), and operates additional offices in Cambridge, Massachusetts.
Matlantis firmographics
Firmographics- Name
- Matlantis
- Legal name
- Matlantis Corporation
- Website
- https://matlantis.com
- Company type
- Private
- Founded year
- 2021
- Operating status
- Operating
- Headcount range
- 11–50 employees
- Short description
- Matlantis is a Tokyo-based cloud SaaS platform that uses a proprietary universal machine-learning interatomic potential (PFP) to deliver fast, DFT-level atomistic simulations for materials R&D teams in automotive, chemicals, semiconductor, battery, and energy companies worldwide.
- Ownership category
- akta.pro rank
Where Matlantis is headquartered
LocationHeadquarters
- HQ city
- Tokyo
- HQ country
- Japan
- HQ region
- Asia
Offices2 records
Markets served
Matlantis business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Technology or R&D, Infrastructure, Personnel, Marketing or Sales, Operations
Revenue model
- Cloud-based SaaS Subscription: Subscription-based cloud service for atomistic simulation, accessible via browser without local infrastructure requirements. Pricing varies by usage tier and organization size.
- Enterprise Licensing: Enterprise agreements for larger organizations with dedicated support, custom onboarding, and potentially higher usage limits.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Unit Pricing | One time | Hands-on Seminar Training |
| Subscription | Annual | Standard Cloud Subscription |
Go-to-market motion2 records
Distribution channels3 records
Marketing channels8 records
Matlantis product offering
Product offeringCore offering
Matlantis provides a cloud-based AI-powered universal atomistic simulator that enables fast and accurate atomic-level simulations for materials discovery. The platform combines deep learning models (Preferred Potential/PFP) trained on DFT data with computational chemistry to support simulations across all 96 naturally occurring elements and systems up to 44,000 atoms. Delivered as a browser-accessible SaaS, it serves chemicals, semiconductors, batteries, and materials science R&D teams by reducing verification cycles from years to weeks.
Differentiator
Problem solved
Functional benefit
Products and services
- Matlantis Cloud Platform Universal atomic-level AI simulator that integrates deep learning models (PFP) with traditional atomistic simulation. Enables fast, accurate atomistic simulations for materials discovery across catalysts, batteries, semiconductors, alloys, lubricants, ceramics, and chemicals. Delivered as cloud-based SaaS accessible via browser without environment setup, targeting enterprise R&D teams and academic researchers.
- PFP (Preferred Potential) Universal Machine-Learning Interatomic Potential (uMLIP) serving as the core AI model powering Matlantis simulations. Trained on DFT calculation results using Neural Network Potential technology. PFP v8 (July 2025) achieves approximately twice the accuracy of previous versions through r²SCAN training data, supports all 96 naturally occurring elements, and handles systems up to 44,000 atoms.
- LightPFP Lightweight potential model for large-scale simulations enabling computation of systems with significantly more atoms than standard PFP. Successfully simulated systems of approximately 72,000 atoms for SiO2 slabs. Uses active learning techniques for efficient model training on specialized systems and integrates with NVIDIA ALCHEMI Toolkit-Ops for optimized performance.
- RestScan and ReactionString PFP-based applied technology for molecular structure analysis and reaction characterization, enabling researchers to explore chemical reaction pathways and mechanisms at the atomic level.
- GRRM20 with Matlantis PFP-based applied technology for reaction pathway analysis, enabling automated exploration of chemical reaction mechanisms through GRRM20 tools combined with Matlantis's atomistic simulation capabilities.
- Matlantis CSP (Crystal Structure Prediction) Capability for rapidly discovering previously unknown stable crystal structures from atomic configuration and composition search spaces. Uses PFP for high-throughput structure evaluation (seconds to minutes per structure) and proprietary algorithms for efficient composition space exploration. Offers Global Search and Substitutional Structure Search modes with 3-6x improvement over random search.
- PFP Descriptors PFP-based feature extraction technology for characterizing material properties at the atomic level, enabling downstream machine learning workflows on top of Matlantis simulation outputs.
- Matlantis Skills Library Public GitHub repository (matlantis-pfcc/matlantis-contrib) packaging Matlantis-specific knowledge including functions, APIs, and representative workflows that AI agents can load and reference. Initial workflows include structure relaxation, molecular dynamics, reaction pathway exploration, crystal structure prediction, visualization, and retrieving structures from external databases.
- Hands-on Seminars Offline paid educational program for training users on Matlantis. Covers Matlantis operation, molecular dynamics tutorials, reaction pathway analysis (NEB method), practical exercises including paper reproduction, and advanced feature introductions. Priced at 110,000 JPY per participant.
Quantifiable outcome
- Reduced materials verification process from 2-3 years to 6 weeks (Kuraray)
- +5 more outcomes
Companies that use Matlantis
Customer profileNamed customers34 records
Segments4 records
Ideal customer profiles4 records
Matlantis technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration6 records
AI capability8 records
Feature8 records
Matlantis partnerships and signals
Strategic signalPartnerships
Seven partnerships are on record, tiered core, strategic, minor and flagship.
- AnthropiccoreIntegration of Anthropic's Claude Code into Matlantis terminal environment, enabling researchers to create, edit, and run simulations through natural language instructions. Includes public Skills library on GitHub.
- NVIDIA (Ising Quantum AI)strategicPartnership for Ising quantum AI model adoption, focusing on calibration and error correction in quantum systems, building AI factories and high-efficiency scientific computing setups.
- NVIDIAcoreIntegration of NVIDIA ALCHEMI Toolkit into Matlantis platform for GPU-accelerated atomistic simulations. Previous collaboration included NVIDIA Warp-optimized kernels achieving up to 10x speed improvements. Future roadmap includes PFP integration with the toolkit for enhanced industrial-scale simulations.
- Preferred Networks (PFN)coreJoint development of Matlantis core technology. PFN provides AI technology and computing infrastructure. Matlantis is a materials discovery arm of PFN.
- ENEOS CorporationcoreJoint development partner providing domain knowledge and know-how in the chemical field. Original co-developer of Matlantis alongside PFN.
- Massachusetts Institute of Technology (MIT) - Prof. Ju LiminorProf. Ju Li from MIT serves as Technical Advisor to Matlantis, appointed August 2024.
- Honda R&DflagshipEarly adopter of Matlantis CSP (Crystal Structure Prediction), collaborating on multi-component system and metastable structure exploration for materials development.
Scale indicators10 records
Recent moves10 records
Expansion highlights6 records
Matlantis competitors and assessment
Company assessmentEmerging players
- CuspAI: CuspAI is an emerging generative-AI platform for materials discovery, focused on designing new functional materials using foundation models. Comparable to Matlantis as an AI-for-materials-discovery peer targeting similar R&D use cases, with overlapping investor ecosystem and similar stage.
- Kebotix: Kebotix combines AI and robotics for accelerated materials discovery, selling to enterprise chemicals and materials companies. Comparable to Matlantis in target verticals (chemicals, materials) and AI-first value proposition, with similar scope but different modality emphasis (lab automation vs. cloud simulation).
- MaterialsZone: MaterialsZone offers an AI-driven materials informatics platform for R&D teams in chemicals, batteries, and advanced materials. Directly comparable to Matlantis in serving enterprise materials R&D with AI-based decision support, though more focused on data aggregation/informatics and less on first-principles atomistic simulation.
- Orbital Materials: Orbital Materials is building AI foundation models for materials discovery, with a focus on industrial applications like batteries and catalysts. Comparable as an emerging AI-for-materials science company targeting similar enterprise verticals, though with narrower specialization and earlier stage.
Broad incumbents
- Ansys: Ansys is a multi-physics engineering simulation incumbent that also offers materials-focused simulation tools. Comparable to Matlantis as an enterprise simulation platform for materials and process engineering, though Matlantis is more specialized and AI-native where Ansys is broader and physics-simulation-centric.
- Dassault Systèmes (BIOVIA): Dassault's BIOVIA brand delivers broad computational chemistry and materials modeling software (Materials Studio, BIOVIA Discovery Studio) used by enterprise R&D. Overlaps with Matlantis on atomistic simulation but sits inside a much broader PLM/3DEXPERIENCE portfolio rather than specializing in ML-driven atomistic simulation.
- Google DeepMind: Google DeepMind has produced foundation-model work in atomistic simulation (e.g. GNoME, GraphCast-adjacent efforts) that competes in the same ML-for-materials science category as Matlantis's PFP. While not a direct commercial peer, its open releases shape the competitive set Matlantis must differentiate against.
Direct peers
- Citrine Informatics: Citrine Informatics provides an AI platform for materials and chemicals R&D, helping enterprises accelerate materials development. Closely comparable to Matlantis in target customer (enterprise R&D in chemicals/materials) and value proposition (AI-driven materials discovery and optimization).
- Schrödinger: Schrödinger operates a computational chemistry and materials science platform combining physics-based simulation (DFT, MD) with ML for materials and drug discovery. Most directly comparable to Matlantis in offering cloud-based atomic-scale simulation tools to enterprise R&D teams in chemicals and pharmaceuticals.
Others
- Rescale: Rescale is a cloud HPC platform used by enterprise engineering and R&D teams to run simulations, including materials and chemistry workloads. Adjacent/ecosystem peer to Matlantis — not a direct MLIP competitor, but a partner-like infrastructure layer that could also become a competing distribution surface for atomistic simulation software.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat6 records
Key risks5 records
Key highlights7 records
Customer concentration
Matlantis social profiles
Digital presenceMatlantis financial estimates
Financial estimateRevenue estimate
Valuation estimate
Matlantis leadership team
Management profileNumber of profiles
Profiles5 records
Matlantis funding detail
Funding detailFunding overview
Funding rounds
Investors
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Matlantis M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Matlantis
What does Matlantis do?
Matlantis provides a cloud-based AI-powered universal atomistic simulator that enables fast and accurate atomic-level simulations for materials discovery. The platform combines deep learning models (Preferred Potential/PFP) trained on DFT data with computational chemistry to support simulations across all 96 naturally occurring elements and systems up to 44,000 atoms. Delivered as a browser-accessible SaaS, it serves chemicals, semiconductors, batteries, and materials science R&D teams by reducing verification cycles from years to weeks.
Is Matlantis a public or private company?
Matlantis is a private company. It is classified as corporate owned and is currently operating.
When was Matlantis founded?
Matlantis was founded in 2021. It employs 11 to 50 people.
Where is Matlantis based?
Matlantis is headquartered in Tokyo, Japan, in the Asia region.
How does Matlantis make money?
Two revenue lines are on record. Cloud-based SaaS Subscription is the primary driver. The others are enterprise Licensing.
Who are Matlantis's main competitors?
Emerging players on record are CuspAI, Kebotix, MaterialsZone and Orbital Materials. Broad incumbents are Ansys, Dassault Systèmes (BIOVIA) and Google DeepMind. Direct peers are Citrine Informatics and Schrödinger. Rescale is listed as an others.
Does Matlantis have an API?
Yes. Matlantis provides a simulation API accessible through Python in Jupyter Notebook environments, with a pfp-api-server for running atomistic calculations. Users can invoke simulations programmatically via the notebook interface. Additionally, Matlantis has integrated Claude Code (Anthropic's AI agent) to operate within the Matlantis terminal environment, allowing researchers to create, edit, and run simulations through natural language instructions. Developer documentation is at docs.matlantis.com/atomistic-simulation-tutorial/en.