Japan is entering a new chapter in artificial intelligence. For decades, the country built global leadership in electronics, automobiles, industrial automation, and advanced manufacturing. Now those capabilities are converging with generative AI, foundation models, AI semiconductors, and autonomous systems.
The numbers illustrate the scale of this transformation. Japan’s AI market is projected to reach $123.9 billion by 2032, growing at a CAGR of 34.4%. [1]
Japan’s current AI ecosystem looks quite different from Silicon Valley’s. The country certainly has companies developing large language models and generative-AI platforms, but some of its most interesting opportunities lie in industrial AI, robotics, autonomous machines, semiconductor technology, and Japanese-language AI.
Below, I’ve featured top Japanese AI companies (including both tech giants and AI-focused corporations) that could play a major role in shaping the country’s next decade of technological growth.
Did you know?The Japanese government plans to provide more than $62 billion in public financial support through FY 2030 for AI and semiconductor industries. [2]
Table of Contents
12. AI inside
Founded: 2015Key AI Products: DX Suite, PolySphere LLM
Competitive Edge: Japanese-language specialization
AI inside originally became famous for DX Suite, an AI-powered document recognition and data-entry platform. It is now expanding into Japanese-language LLMs, multimodal RAG, AI agents, edge AI hardware, and distributed AI inference infrastructure.
In 2025, the company made AI agents a standard component of DX Suite. The agent can automatically identify documents, determine appropriate OCR settings, process files, output structured data, and store the results.
AI inside has also developed its own LLM family, called PolySphere, specifically for Japanese documents and enterprise data. It is integrated into DX Suite, where the company reported an average OCR accuracy of 99.6% on real-world data.
AI inside technologies have now been deployed to more than 70,000 users, including government organizations, local authorities, and private companies. Its document-processing technology has performed more than 9.8 billion readings, giving it an unusually large operational dataset for Japanese document AI. [3]
11. Hitachi
Founded: 1910Key AI Products: Lumada, HMAX
Competitive Edge: Has enormous quantities of operational data
Hitachi is more of an AI-enabled social-infrastructure company than a conventional AI company. It is trying to make AI work reliably in railways, electricity grids, factories, public infrastructure, finance, and other mission-critical environments.
Hitachi refers to the physical equipment and systems that generate useful data as “Digitalized Assets.” It combines these assets with industry expertise and digital services to create additional value.
At the center of this approach is Lumada, Hitachi’s core digital platform. Lumada combines generative AI with decades of industry knowledge to support automated decision-making, predictive maintenance, and frontline operations.
In FY2025, Lumada generated nearly $24.4 billion in revenue, accounting for 38% of Hitachi’s total revenue. That was up from $18.7 billion, or 31% of revenue, a year earlier.
HMAX is another key part of Hitachi’s AI strategy for physical industries. It is designed to go beyond reporting what happened by helping AI understand conditions, predict problems, recommend actions, and eventually coordinate physical operations.
The technology is already being used in mobility and energy, with expansion into other industrial sectors.
10. Sony AI
Key AI Products: Woosh, Gran Turismo Sophy
Competitive Edge: Proprietary semiconductor & sensors
Sony AI is one of Japan’s most distinctive AI companies because it sits at the intersection of artificial intelligence, entertainment, robotics, imaging, sensing, and creativity.
Its objective is to develop fundamental AI research and then connect that research with Sony’s enormous ecosystem of PlayStation, Sony Pictures, Sony Music, image sensors, robotics, and entertainment technology.
For example, Sony AI has developed an autonomous racing agent called Gran Turismo Sophy. It is trained using deep reinforcement learning and the Gran Turismo simulation environment. Sophy learned not only vehicle control but also racing strategy and sportsmanship, including overtaking, defensive driving, and collision avoidance.
One of Sony AI’s most notable achievements is Project Ace, an autonomous table-tennis robot. It is the first known real-world autonomous system capable of competing at the elite/professional level in human table tennis. In fact, Ace won matches against seven ranked professional players in 2026 under official competition rules. [4]
The company also introduced Woosh, a foundation model specifically designed for sound-effect generation. It handles both text-to-audio and video-to-audio generation.
9. ABEJA
Founded: 2012Key AI Products: Sarashina, Sarashina mini/API
Competitive Edge: 300+ AI implementation projects
ABEJA Platform provides the technology and operational infrastructure needed to introduce AI into corporate processes. Over the years, ABEJA has supported more than 300 companies with AI and digital-transformation projects.
Its traditional AI business has increasingly been transformed by generative AI. In FY 2025, LLM-related projects accounted for more than 50% of total revenue, compared with just more than 20% in FY2024.
The company is now expanding beyond software and digital workflows into Physical AI and robotics. In 2026, ABEJA and Murata Manufacturing demonstrated a Vision-Language-Action (VLA) model that controlled a dual-arm robot to perform coordinated component-handling operations.
8. ExaWizards
Founded: 2016Key AI Product: exaBase
Competitive Edge: Exceptional government penetration
ExaWizards focuses less on building a single frontier foundation model and more on turning AI into usable products, AI agents, and enterprise transformation services.
Its flagship AI platform, exaBase, allows ExaWizards to combine different AI technologies with business applications instead of building each project from scratch. The portfolio includes exaBase AI for Government, exaBase Sales Agent, exaBase DX Assessment & Learning, and exaBase Studio.
This “platform + products + implementation” approach is important because many Japanese companies need help integrating AI into their existing workflows, data, and legacy systems rather than simply buying access to an LLM.
ExaWizards has built a strong presence across Japanese businesses and government. Its exaBase AI has more than 150,000 users across 1,700 companies, while its government products are used by 25 of Japan’s 47 prefectures and more than 200 government agencies and municipalities.
In FY2026, ExaWizards reported revenue of $74.8 billion, up 22.3%, while operating profit reached $10 million.
7. Fujitsu
Key AI Products: Fujitsu Kozuchi, Takane
Competitive Edge: HPC and quantum systems integration
Fujitsu is building an enterprise AI stack around its Fujitsu Kozuchi AI platform, Takane enterprise LLM, domain-specific AI, AI agents, graph AI, and security technologies.
Takane is particularly important because it is fine-tuned for Japanese enterprise applications. The goal is not just to handle the Japanese language, but to understand the terminology, business practices, and specialized knowledge used by Japanese companies.
Takane can also be combined with domain-specific AI for industries such as manufacturing and healthcare, where general-purpose LLMs can struggle with highly specialized information. [5]
The company is also moving quickly from generative AI toward agentic AI. Its Multi AI Agent Framework allows multiple AI agents to work together on complex tasks instead of relying on a single chatbot to respond to prompts.
A notable example is its work with Panasonic Electric Works, where Fujitsu’s AI agents and data platform helped manage more than 200,000 part numbers across 3,000 sites. The system also supported more than 300 predictive models for demand forecasting.
Plus, Fujitsu isn’t betting only on conventional AI computing. It is developing the FUJITSU-MONAKA processor and a quantum-computing roadmap that targets 10,000 qubits in 2030. [6]
6. SoftBank / SB Institutions
Founded: 1986Key AI Products: Sarashina, Sarashina mini/API
Competitive Edge: Sovereign Japanese LLM strategy
SB Intuitions is the wholly owned AI company of SoftBank Corp. Its flagship Sarashina model reached around 460 billion parameters in FY2024. The company later developed Sarashina mini, a smaller 70-billion-parameter model designed for practical use in enterprise settings.
The longer-term goal is to combine specialist models to build models approaching one trillion parameters.
The main idea behind Sarashina is sovereign AI. SoftBank argues that relying on foreign AI models can create risks around data sovereignty, operational continuity, Japanese-language nuances, and regulatory requirements. [7]
SB Intuitions focuses on four areas of sovereignty: technology, data, operations, and laws and regulations. Its models can be trained and operated on infrastructure in Japan, allowing governments, financial institutions, universities, and companies to keep sensitive data within the country.
5. NEC
Founded: 1899Key AI Products: Cotomi, Cotomi Act
Competitive Edge: Government & defense relationships
NEC’s AI expertise goes back well before the current generative AI boom. The company has more than 50 years of AI research experience, and its history includes computer research, OCR, biometrics, supercomputing, and advanced data-processing technologies.
NEC’s AI portfolio is centered on NEC Cotomi, its proprietary generative AI technology, and Cotomi Act, its AI-agent technology for executing digital tasks. Around these technologies, NEC is building applications for enterprise search, customer service, sales support, cybersecurity, government, and industrial operations.
In its WebArena evaluation, NEC reported an 80.4% task-success rate, compared with a 78.2% human benchmark. This was the first time an AI system exceeded the human score in this benchmark. [8]
4. Nippon Telegraph and Telephone Corporation
Key AI Products: Tsuzumi 2
Competitive Edge: Massive infrastructure, Proprietary telecom data
NTT’s AI story is fundamentally different from that of a startup. The company has been conducting AI and computing research for decades through NTT Research and its laboratories, and it has one of Japan’s largest corporate R&D organizations.
NTT has 2,400 researchers, holds 22,000 patents, and publishes around 2,700 academic papers and presentations each year.
They have developed a large language model called tsuzumi. It is designed to be much smaller and more computationally efficient than the giant models developed by US hyperscalers. Its main selling points are strong Japanese-language processing, low operating cost, low power consumption, and the ability to operate in closed/on-premises environments.
The company is also expanding data centers, AI computing, and optical networking infrastructure. The data-center business has nearly 2,000 MW of total power-receiving capacity across the group and is targeting over 2,700 MW, with a longer-term goal of approximately 3,000 MW by FY2030.
3. PKSHA Technology
Key AI Products: PKSHA Dialogue, Voice, AI FAQ Assistant
Competitive Edge: Enterprise AI SaaS + proprietary algorithms
PKSHA focuses on deploying AI in real-world corporate workflows — customer service, contact centers, internal help desks, knowledge management, voice operations, sales, and increasingly autonomous AI agents.
The company has accumulated more than a decade of proprietary AI/NLP experience, 4,400+ corporate deployments, and over 7,000 AI agents in operation.
A 2026 Fuji Chimera Research Institute survey ranked PKSHA’s products No. 1 in three categories based on FY2024 results: chatbots, voicebots, and FAQ/knowledge-management systems.
PKSHA ChatAgent has handled more than 750 million conversations, while VoiceAgent processes over 300,000 incoming calls a month. Its FAQ product is used by more than 800 companies across 1,500+ sites.
In FY2026, PKSHA generated about $217 million in revenue. Its customer base is particularly impressive, with more than 70% of Japan’s top 100 companies by market capitalization using its AI solutions or SaaS products.
2. Preferred Networks
Key AI Products: PLaMo model, MN-Core processor series
Competitive Edge: Vertical integration, Energy efficiency
Preferred Networks (PFN) is trying to control four layers of the AI stack itself: AI semiconductors, computing infrastructure, foundation models, and AI applications.
The company has developed the MN-Core family of processors specifically for AI workloads. Its MN-3 supercomputer, powered by the first-generation MN-Core, ranked first on the Green500 list of the world’s most energy-efficient supercomputers three times.
PFN subsequently developed MN-Core 2 and launched the Preferred Computing Platform (PFCP) in 2024, enabling customers to access its AI computing infrastructure as a cloud service.
Its second biggest technology is PLaMo, its Japan-developed foundation-model family. Unlike rivals that simply fine-tune an American foundation model, PFN developed PLaMo fully from scratch, using training data rich in Japanese content. [9]
The commercial significance is that PLaMo is growing. For instance, PLaMo models are being incorporated into third-party products including miibo, Tachyon Generative AI, and QommonsAI. The latter is reportedly used by more than 150 Japanese local governments.
So far, PFN has raised more than $270 million across multiple rounds, and its valuation has reached $2 billion. Its investors and strategic partners include Toyota, NTT, FANUC, Hitachi, Mizuho, Mitsubishi Corporation, and Mitsui.
1. Sakana AI
Key AI Products: Sakana Fugu, Sakana Marlin, Sakana Chat
Competitive Edge: Evolutionary/multi-agent architectures
Sakana AI is not trying to beat OpenAI by training the world’s largest model from scratch. Instead, it is pursuing collective intelligence, model orchestration, self-evolving AI, efficient models, and Japan-specific post-training.
Its research portfolio includes the Darwin Gödel Machine, which explores self-improving AI systems; ShinkaEvolve, which evolves LLM-generated programs; AB-MCTS, which coordinates multiple models; and The AI Scientist, a multi-agent system that automates parts of the scientific research process.
The company is increasingly turning that research into enterprise and government applications. For instance, in 2025, it signed a three-year-plus partnership with MUFG Bank, Japan’s largest banking group, to develop banking-specific AI.
In 2026, Sakana AI launched Sakana Chat, Sakana Marlin, and Sakana Fugu, followed by Sakana Translate. Marlin is the company’s first commercial product, designed to conduct long-horizon strategic research autonomously for up to eight hours. It can produce reports of up to 100 pages along with executive-summary slides. [10]
Read More
- 12 Top Chinese AI Companies
- 17 Best AI Tools For Businesses
- 14 Leading Japanese Semiconductor Companies
- Japan’s AI market and trend analysis, Fortune Business Insights
- Bold investment for growth through public-private collaboration, Ministry of Economy and Trade
- AIinside is moving to the next growth stage, Moomoo
- Outplaying elite table tennis players with an autonomous robot, Nature
- Takane LLM successfully piloted in the central government agency, Fujitsu
- The year of practical quantum computing, Fujitsu
- Launching generative AI services using homegrown LLM “Sarashina,” SoftBank
- Accurate performance even with vague instructions, NEC
- PFN launches Japan-made LLM PLaMo Prime, Preferred Networks
- Sakana AI launches ‘Marlin’ research service for businesses, MarketWatch
