Hitachi is building a global AI Factory using Nvidia GPUs to deliver physical AI – industrial AI, plus domain-specific SLMs/LLMs – for rail, energy, factories, and infrastructure, integrating IoT, digital twins, edge AI, and (some) private 5G. It has also agreed a deal to buy German industrial AI consultancy Synvert, to be integrated with its US integrator business GlobalLogic.
In sum – what to know:
Private AI platform – Hitachi is building a distributed global AI cloud with Nvidia GPUs to deliver physical AI for rail, energy, factories, and infrastructure.
New Industry 4.0 – the initiative seeks to mesh new domain-specific AI mechanics with the traditional Industry 4.0 disciplines of IoT, robotics, automation.
Strategic expansion – the firm is strengthening its AI capabilities with the purchase of Synvert, which will be integrated with US-based GlobalLogic.
Inevitably, like its peers, Japan-based industrial conglomerate Hitachi is leaning into the new wave of ‘physical AI’ with a distributed “AI factory”, localised in major regional markets and populated (invariably) with Nvidia AI chips, to bring private cloud infrastructure and domain-specific AI, combining enterprise-geared small and large language models (SLMs and LLMs), to railways, energy grids, factories, and other industrial infrastructure.
The setup (‘Hitachi AI Factory’) is distributed across the US, Japan, and the broad EMEA region, said a press note – to “ensure Hitachi’s engineers can collaborate seamlessly and access powerful computing resources with low latency, no matter where they are”. It went on: “This interconnected network will support the creation of a wide range of physical AI applications, driving new levels of efficiency, productivity, and safety across industries.”
Physical AI, as terminology, has arisen relatively recently in order to distinguish the physical application of AI, as commonly defined in robots and machinery in Industry 4.0 scenarios, from data-based LLM-style AI, as confined to text, images, or numbers on a server. It describes the difference between a virtual chatbot and a physical robot – that adjusts a factory line in real time, or triggers a maintenance crew on a railway line.
In reality, physical AI has been around for years; the new definition is only because the GPU mechanics behind big LLM systems can be modified and trained on proprietary corporate and niche industrial data to support and accelerate traditional ML- and AI-geared industrial analytics systems. It draws on most of the same Industry 4.0 elements: sensors and cameras to perceive, digital twins and AI models to reason, and robotics and actuators to act.
Hitachi’s new tie-up with Nvidia – its AI factory uses Nvidia’s ‘AI factory’ reference architecture, which specifies use of its “full-stack AI platform” (its AI Enterprise, Omniverse, and Blackwell GPU products) – positions it as a go-to automation supplier for AI-accelerated industrial solutions, layering in higher-grade AI on top of the classic Industry 4.0 stack, which mixes IT, OT, and IoT, plus pattern matching ML techniques.
Hitachi called it a “centralized infrastructure… to accelerate the development and deployment of physical AI solutions across [its] core business sectors”, as listed at the top of the piece. Nvidia’s stack is geared to train general-purpose LLMs and domain-specific SLMs, fine-tuned industrial vocabularies, sensor data, and OT control systems. The focus on physical AI implies a heavier tilt toward domain-specific and multimodal models.
There is clear crossover with industrial IoT, and implicit crossover with industrial 5G. Hitachi’s old Lumada IoT unit integrates sensor data; its new Lumada 3.0 vision extends this, somehow, into an AI-savvy operating model for enterprises to solve “business and societal problems through co-created digital transformation”, it said. The company has also deployed a number of private 5G networks, from others, to support the same transformation goals.
Notably, it is using its own system integration division, GlobalLogic, to deploy AI, IoT, and 5G solutions at its own companies and factories – as part of its self-fulfilling ‘One Hitachi’ strategy to buy and sell digital change solutions among its own businesses, as well as to serve other industrial enterprises in other industrial segments. Earlier this month, GlobalLogic said it had knitted together componentry from Hitachi’s broader tech portfolio into a private 5G network from Ericsson at a US manufacturing plant belonging to Hitachi Rail, its division for rolling stock.
It said its new AI factory will use its HMAX family of industrial AI solutions, used by Hitachi Rail for railway operations and maintenance, as well as Hitachi Vantara’s AI solution portfolio Hitachi iQ, and Hitachi’s liquid-cooled AI data centres. Jun Abe, general manager of the company’s digital systems and services division, said: “The collaboration [with Nvidia] is a key engine for solving complex real-world problems… By establishing a global AI factory, we can now operate as a true ‘One Hitachi’ across regions and organizations.”
Related, Hitachi has just agreed a deal to buy Germany-based industrial AI consultancy Synvert (stylised as ‘synvert’) – as a wholly owned subsidiary of GlobalLogic, to accelerate the deployment of the HMAX suite of agentic and physical AI solutions. Synvert claims around 200 clients, notably in Germany, Switzerland, Spain, Portugal, and the Middle East/ It has “advanced partnerships with leading cloud and data platform vendors”, apparently.
Its supply-side partners include Databricks, Snowflake, plus all major public cloud vendors (AWS, Microsoft Azure, Google Cloud). The deal is expected to close by March 2026.
Abe said: “Hitachi is accelerating social innovation by addressing urgent societal challenges – such as labour shortages and knowledge transfer among frontline workers in the transportation, energy, gas, and railway sectors – through the advancement of its own AI-driven transformation and the delivery of its outcomes to customers. By integrating Synvert’s outstanding data analytics and consulting capabilities with GlobalLogic’s digital engineering expertise, we will enhance competitiveness through agentic AI and accelerate HMAX deployment.”
Srini Shankar, president and chief executive at GlobalLogic, said: “Integrating Synvert into GlobalLogic will play a key role in accelerating our global strategy. It will strengthen our data and consulting capabilities and expand our presence in Europe and the Middle East across industries like energy, retail, financial services, and insurance. Once the acquisition is completed, we expect to rapidly deliver innovation through end-to-end development of data-driven AI services, and accelerate efforts in agentic and physical AI.”
André Holhozinskyj, chief executive at Synvert, said: “Joining the Hitachi Group, which has strengths in OT and products, is an ideal step in Synvert’s growth story. GlobalLogic’s capabilities and regional strategy align well with synvert, and we believe this partnership will drive further growth. Synvert’s identity resonates with Hitachi’s purpose-driven culture, and we look forward to what we can achieve together.”