What if the irreplaceable skills of master craftsmen could live on through AI? Hitachi is betting big on Physical AI, a market it expects to reach roughly $125 billion (about ¥20 trillion) by 2030. Often described as the wave that follows generative AI, this technology could be the key to preserving decades of manufacturing expertise as Japan faces a severe labor shortage and an aging workforce. Here is how the company's unusual combination of IT, OT, and physical products positions it to lead.

What Is Physical AI?

Physical AI refers to artificial intelligence that perceives and understands the real world through sensors and cameras and generates physical actions in response. Unlike traditional AI focused on text and image generation, Physical AI excels in real-world applications such as robotics control, predictive maintenance, and the optimization of complex systems.

According to a reference figure from Grand View Research, the Physical AI (AI-equipped robotics) market is projected to reach around $124.7 billion (about ¥20 trillion) by 2030. That is roughly three times the size of the AI agent market, and it is drawing attention as the field expected to gain dominance after generative AI.

Hitachi's Unique Vision for Physical AI

Jun Abe, Hitachi's Executive Vice President, has laid out a clear vision for the company's Physical AI strategy.

"Our goal is to view the sites of social infrastructure themselves, railways, power, and manufacturing, as a single system, and to use AI to optimize everything from skilled workers' tacit knowledge to the behavior of equipment," Abe says.

While other companies focus on humanoid robots as Physical AI technology, Hitachi frames Physical AI around supporting people and keeping social infrastructure running stably. NVIDIA CEO Jensen Huang has called Hitachi a rare kind of company, noting that it holds all the capabilities needed to create value from data across IT, operational technology (OT), and physical products.

Japan's Manufacturing Crisis: The Tacit Knowledge Problem

Japanese manufacturing faces a structural crisis. According to the Ministry of Economy, Trade and Industry's Monozukuri (Manufacturing) White Paper 2025, manufacturing employment has fallen by about 1.57 million over the past two decades, with a particularly steep decline among workers under 34, while the share of workers aged 65 and older keeps rising.

At the heart of the problem is the extreme difficulty of passing on tacit knowledge, the intuitive expertise that experienced workers hold. Consider the skills master craftsmen build over decades: sensing subtle temperature changes in metal by color, distinguishing faint abnormal sounds from machinery, finishing work to micron-level precision. These abilities, often described as instinct, knack, and experience, resist documentation and standardization. Making matters worse, more than 60 percent of establishments report a shortage of people available to train and mentor younger workers.

HMAX: Hitachi's Answer

At the core of Hitachi's Physical AI strategy is HMAX (Hyper Mobility Asset Expert), built on four key elements.

First, it collects vast amounts of data from digitized assets such as power grids, railways, and manufacturing equipment through sensors and control systems. Second, it embeds into AI models the operational and maintenance expertise, the domain knowledge, that Hitachi has built over more than 110 years. Third, it integrates a range of AI technologies, including perception AI, generative AI, agentic AI, and physical AI. Fourth, it draws on strategic partnerships with global tech companies such as NVIDIA, Google Cloud, and OpenAI.

Early deployments in European railway systems have already delivered measurable results: a 15 percent reduction in energy consumption, a 20 percent decrease in train delays, and 15 percent lower maintenance costs.

How Hitachi Captures Tacit Knowledge

Through its AI Agent Development, Operation, and Environment Service, Hitachi is working to transfer expert knowledge to AI. First, using ethnography, an anthropological observation technique, it records and analyzes in detail how skilled workers behave and make decisions. Next, through interviews with those workers, it draws out knowledge and know-how that has never been put into words. It then combines this extracted tacit knowledge with explicit knowledge such as forms and design documents to build data that AI can learn from.

Within Hitachi Group companies such as Hitachi Building Systems and Hitachi Power Solutions, hundreds of business processes already use AI, giving the company a track record on which to base practical, proven solutions for customers.

Aiming to Become the World's Top User of Physical AI

Hitachi has declared its ambition to become the world's top user of Physical AI. Notably, it has decided not to compete in developing its own large language models.

"LLMs will most likely end up as commodities chosen mainly on price. That layer holds little value for us," Abe says. Instead, Hitachi positions the social-infrastructure domain knowledge it has accumulated over more than a century as its greatest asset and strength, and its strongest defensive moat.

In Japan, a shrinking working-age population is accepted as an unavoidable reality, and hopes are high for AI-enabled skills transfer. On the factory floor, the pressing question is how to digitally preserve the artisan techniques that would otherwise vanish as veteran engineers retire.

How is the manufacturing skills-transfer challenge viewed in your country? What discussions are happening around using AI to preserve and pass on skilled workers' expertise? Please share your thoughts in the comments.

References