Preferred Networks
Preferred Networks (PFN) is a Tokyo-based AI unicorn founded in 2014 that vertically integrates proprietary AI semiconductors (MN-Core), compute infrastructure, the PLaMo Japanese foundation model, and applied AI products (Matlantis, Talessca) to serve enterprise customers across manufacturing, materials, life sciences, finance, and other Japanese verticals.
- Company typePrivate
- Founded2014
- HeadquartersOtemachi, Japan
- Headcount251–500
- GTM typeB2B
- OfferingSoftware
What Preferred Networks does
Preferred Networks (PFN) is a Tokyo-headquartered, privately-held AI company founded in 2014 that vertically integrates the full AI value chain across four layers: proprietary AI semiconductors (MN-Core series, including the MN-Core L line targeted at physical AI), proprietary computing infrastructure, in-house generative AI foundation models (the PLaMo family, with PLaMo 3.0 Prime released in June 2026), and applied AI products and solutions. Its flagship products include PLaMo (a domestic Japanese LLM), Matlantis (an AI-driven atomistic simulation platform using the Preferred Potential model and integrated with NVIDIA's ALCHEMI Toolkit for materials discovery, documented to compress multi-year R&D verification cycles to weeks), Talessca (an AI interview service using generative AI and AI avatars that reportedly reduces screening workloads by ~80%), and the MN-Core L AI semiconductor used in joint research with Toyota.
PFN serves enterprise customers across nine vertical markets — manufacturing, materials and chemicals, life sciences, entertainment, retail and distribution, finance, public services and infrastructure, education, and general enterprise — plus an HR/recruitment use case via Talessca. Named customers and investors include Toyota (multi-round investor since 2015 and physical-AI research partner), Mitsubishi Heavy Industries (mission-critical AI partner), Kuraray, Okinawa Cellular, Chugai Pharmaceutical, Hitachi, Mizuho Bank, SBI Group, Mitsubishi Corporation and ENEOS, among others. The go-to-market is enterprise-led via direct sales and strategic partnerships with large Japanese corporates, supplemented by technology-ecosystem integrations (NVIDIA, Kioxia, NTT/Edgecore IOWN) and thought-leadership content from co-founder Daisuke Okanohara.
The business model combines subscription/recurring revenue from AI products, foundation models and computing infrastructure with hardware sales of MN-Core semiconductors; pricing is quote-based and not publicly disclosed. PFN is recognized as a Japanese AI unicorn and has raised approximately 19 billion yen (~$125M) in December 2024 led by SBI Group with Mitsubishi Corporation, bringing cumulative disclosed funding across eleven rounds to roughly $350M. Co-founder Daisuke Okanohara (岡野原大輔) serves as Representative Director, President and CEO.
Preferred Networks firmographics
Firmographics- Name
- Preferred Networks
- Legal name
- 株式会社Preferred Networks (Preferred Networks, Inc.)
- Website
- https://preferred.jp
- Company type
- Private
- Founded year
- 2014
- Operating status
- Operating
- Headcount range
- 251–500 employees
- Short description
- Preferred Networks (PFN) is a Tokyo-based AI unicorn founded in 2014 that vertically integrates proprietary AI semiconductors (MN-Core), compute infrastructure, the PLaMo Japanese foundation model, and applied AI products (Matlantis, Talessca) to serve enterprise customers across manufacturing, materials, life sciences, finance, and other Japanese verticals.
- Ownership category
- akta.pro rank
Preferred Networks industry classification
Industry- Product category
- Artificial Intelligence Platform
- NAICS
- Semiconductor and Related Device Manufacturing (334413), Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (5182), Other Communications Equipment Manufacturing (334290)
- SIC
- Electronic Components, Nec (3679), Services-Computer Integrated Systems Design (7373)
- akta.pro primary industry
- AI Hardware IP & Chiplet/Packaging for Accelerators (EDA/IP, advanced packaging) (HDAAAAAL)
- akta.pro secondary industry
- AI/ML Engineering & Model Development Services (BPAEAHAF)
Keywords
Where Preferred Networks is headquartered
LocationHeadquarters
- HQ city
- Otemachi
- HQ country
- Japan
- HQ region
- Asia
Offices1 record
Markets served
Preferred Networks business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Technology or R&D, Personnel, Infrastructure, Supply Chain, Operations, Marketing or Sales
Revenue model
- AI Products and Solutions: Providing AI-powered products and solutions across multiple industry verticals including manufacturing, materials, life sciences, entertainment, retail, finance, public services, education, and enterprise.
- AI Semiconductors: Development and sale of proprietary AI semiconductors (MN-Core series) for AI acceleration.
- Computing Infrastructure: Proprietary computing infrastructure to support AI development and services.
- Generative AI Foundation Models: Providing access to domestically-developed foundation models (PLaMo series) for enterprise and research applications.
Go-to-market motion2 records
Distribution channels3 records
Marketing channels7 records
Preferred Networks product offering
Product offeringCore offering
Preferred Networks vertically integrates the AI value chain across four layers that it develops in-house: proprietary AI semiconductors (MN-Core series), large-scale computing infrastructure, a domestic Japanese generative AI foundation model (PLaMo series, including PLaMo 3.0 Prime), and applied AI products and solutions. Its applied products include Matlantis (atomistic simulation for materials discovery), Talessca (AI-powered interview screening), and industry-specific AI solutions for manufacturing, materials, life sciences, finance, retail, entertainment, public services, education, and enterprise.
Product overview
Preferred Networks (PFN) operates a vertically integrated AI value chain across four layers: AI semiconductors (MN-Core L Series), computing infrastructure, generative AI foundation models (PLaMo series), and AI products/solutions. The company offers PLaMo as its flagship Japanese domestic generative AI model, alongside Talessca for AI-powered interview automation and Matlantis for atomistic simulation in materials discovery. PFN's technology stack combines proprietary hardware with software solutions serving diverse industries including manufacturing, life sciences, materials, entertainment, retail, finance, public services, and education.
Differentiator
Problem solved
Functional benefit
Brands
- Talessca (Thaleska): AI interview service using generative AI and AI avatars to standardize interview evaluations, achieving 80% reduction in interview workloads
- Matlantis
- PLaMo
- MN-Core
Products and services
- PLaMo (Preferred Language Model) Domestic Japanese generative AI foundation model developed by Preferred Networks; the PLaMo 3.0 Prime version supports multimodal capabilities for enterprise and industrial applications across text generation and other modalities. Targeted at Japanese enterprises and research organizations seeking sovereign generative AI.
- MN-Core L Series Proprietary AI semiconductor developed by Preferred Networks for accelerating AI computations, particularly physical AI applications including robotics and autonomous systems. Used in joint research with the Toyota Mirai Sosei Center for high-speed physical AI processing.
- Talessca (Thaleska) AI interview service utilizing generative AI and AI avatars to standardize interview evaluations for enterprise HR departments. Achieves an 80% reduction in interview workloads during initial screening and supports talent management with industry-specific scenarios. Customers include Okinawa Cellular, which adopted Talessca for its 2026 fiscal year new graduate recruitment.
- Matlantis AI-driven atomistic simulation platform enabling researchers to evaluate thousands of material candidates at near quantum-level accuracy in seconds rather than months. Features the Preferred Potential (PFP) model integrated with NVIDIA's ALCHEMI Toolkit for industrial-scale materials discovery in batteries, semiconductors, catalysts, and other materials. Adopted by Kuraray to reduce a 2-3 year verification process to six weeks.
- Computing Infrastructure Proprietary large-scale computing infrastructure providing the massive computational power required for AI development and serving as the foundation of PFN's vertically integrated AI value chain from semiconductors to applications.
- AI Products and Solutions Applied AI products and solutions combining PFN's computing infrastructure and cutting-edge technology with industry domain knowledge, delivered across manufacturing, materials, life sciences, entertainment, retail, finance, public services, education, and enterprise verticals.
Quantifiable outcome
- 80% reduction in interview workloads for companies during initial screening
- +2 more outcomes
Companies that use Preferred Networks
Customer profileNamed customers4 records
Segments10 records
Ideal customer profiles4 records
Preferred Networks technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- No
- API docs
- API detail
Core technology
AI maturity
App detail
AI capability7 records
Feature5 records
Preferred Networks partnerships and signals
Strategic signalPartnerships
Five partnerships are on record, tiered core and ecosystem.
- Mitsubishi Heavy IndustriescoreBusiness partnership for joint development of domestic AI technology for mission-critical areas in manufacturing and industrial applications.
- Toyota (Toyota Mirai Sosei Center)coreJoint research partnership to accelerate physical AI using PFN's AI semiconductor MN-Core L series for robotics and autonomous systems applications.
- Edgecore (Accton Company)ecosystemEcosystem partner in Edgecore Open Fabric platform built on NTT's IOWN All-Photonics Network architecture, delivering 102.4 Tbps intra-datacenter optical bandwidth.
- Kioxia CorporationecosystemCollaboration on RAG and LLM optimization. Kioxia's KIOXIA AiSAQ ANNS algorithm integrated with Milvus vector database, with Kioxia LC9 Series SSDs supporting vector database scale of 20 billion vectors.
- NVIDIAecosystemMatlantis platform integrated with NVIDIA's ALCHEMI Toolkit to remove infrastructure bottlenecks, enabling Preferred Potential (PFP) model to operate at industrial scale.
Scale indicators5 records
Recent moves6 records
Expansion highlights6 records
Preferred Networks competitors and assessment
Company assessmentDirect peers
- Sakana AI: Tokyo-based AI research company building Japanese foundation models (similar to PFN's PLaMo). Closest direct peer given geographic overlap, focus on domestic Japanese AI capabilities, and foundation-model-centric approach.
- Schrödinger, Inc. Public computational chemistry / materials simulation company. Direct comparable to Matlantis in using physics-based AI/atomistic simulation for materials and drug discovery, with similar enterprise customer profile.
Broad incumbents
- NVIDIA: Dominant AI accelerator incumbent whose GPU ecosystem defines the market PFN's MN-Core ASIC competes against. PFN integrates NVIDIA tooling (ALCHEMI Toolkit) while also building an alternative compute platform.
- Anthropic: Leading global foundation model developer (Claude). Competes with PFN's PLaMo in enterprise foundation models, with a fundamentally different scale and capability frontier but overlapping customer base.
- SoftBank Group: Japanese tech conglomerate that owns Arm and has invested aggressively in AI (including a reported $500M anchor in PFN peer Sakana AI). Broader incumbent with overlapping Japan-AI ambitions.
- Hitachi: Japanese industrial conglomerate with significant AI/digital solutions business (Lumada). PFN is also a prior Hitachi strategic investor (2017), and Hitachi represents the broader incumbent landscape PFN targets in manufacturing/public services AI.
Emerging players
- Cerebras Systems: US-based AI accelerator company developing proprietary wafer-scale chips for AI workloads. Direct emerging competitor to MN-Core in the alternative-AI-hardware category.
- SambaNova Systems: US AI chip startup building dataflow accelerators for foundation model training and inference. Comparable to MN-Core as an emerging alternative to NVIDIA GPUs for enterprise AI.
- Groq: AI inference accelerator company with proprietary LPU architecture. Comparable to PFN's MN-Core in building domain-specific silicon targeting AI workloads at lower cost/latency than general GPUs.
- Recursion Pharmaceuticals: AI-driven drug discovery company. Comparable to PFN's life-sciences vertical in applying foundation models/compute to pharma R&D, though with a much more vertically focused platform.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat6 records
Key risks6 records
Key highlights7 records
Customer concentration
Preferred Networks social profiles
Digital presencePreferred Networks compliance and trust
Trust signalCompliance1 record
Preferred Networks financial estimates
Financial estimateRevenue estimate
Valuation estimate
Preferred Networks leadership team
Management profileNumber of profiles
Profiles10 records
Preferred Networks funding detail
Funding detailFunding overview
Funding rounds10 records
Investors21 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Preferred Networks M&A and investment
M&A and investmentM&A
Investments5 records
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Preferred Networks
What does Preferred Networks do?
Preferred Networks vertically integrates the AI value chain across four layers that it develops in-house: proprietary AI semiconductors (MN-Core series), large-scale computing infrastructure, a domestic Japanese generative AI foundation model (PLaMo series, including PLaMo 3.0 Prime), and applied AI products and solutions. Its applied products include Matlantis (atomistic simulation for materials discovery), Talessca (AI-powered interview screening), and industry-specific AI solutions for manufacturing, materials, life sciences, finance, retail, entertainment, public services, education, and enterprise.
Is Preferred Networks a public or private company?
Preferred Networks is a private company. It is classified as venture growth investor backed and is currently operating.
When was Preferred Networks founded?
Preferred Networks was founded in 2014. It employs 251 to 500 people.
Where is Preferred Networks based?
Preferred Networks is headquartered in Otemachi, Japan, in the Asia region.
How does Preferred Networks make money?
Four revenue lines are on record. AI Products and Solutions are the primary driver. The others are AI Semiconductors, computing Infrastructure and generative AI Foundation Models.
Who are Preferred Networks's main competitors?
Direct peers on record are Sakana AI and Schrödinger, Inc.. Broad incumbents are NVIDIA, Anthropic, SoftBank Group and Hitachi. Emerging players are Cerebras Systems, SambaNova Systems, Groq and Recursion Pharmaceuticals.
Does Preferred Networks have an API?
No public API is recorded for Preferred Networks.
What industry is Preferred Networks in?
Preferred Networks's product category is Artificial Intelligence Platform. Its primary akta.pro industry code is HDAAAAAL, AI Hardware IP & Chiplet/Packaging for Accelerators (EDA/IP, advanced packaging), with a secondary code of BPAEAHAF, AI/ML Engineering & Model Development Services. Its NAICS code is 334413 and its SIC code is 3679.