From a kitchen table to an industrial revolution with generative AI
Automation has always been one of the big bets in modern industry, but few have managed to turn that idea into something as tangible and impactful as Takayuki Hirayama. In 2006, he founded ARUM Inc. in Kanazawa, Japan, starting from a simple kitchen table. Nearly twenty years later, the company sits at the center of a quiet revolution in precision manufacturing, combining intelligent software, fully automated machines, and now generative artificial intelligence to scale its business beyond Japanese borders.
The road to get here was anything but a straight line. Economic crises, a shortage of skilled labor, and an increasingly demanding market were the ingredients that forced a strategic pivot. And it was precisely from that pressure that ARUMCODE and TTMC were born — two solutions that together cover the entire metal processing cycle and are now used by more than 200 companies across Japan. Dozens of units of the fully automated TTMC milling machine have already been shipped to clients throughout the country. But Hirayama is not stopping there. With export plans targeting the U.S., South Korea, and India by August 2027 and integration with GPT-5 via Azure OpenAI on Microsoft Foundry, ARUM is positioning itself as one of the most compelling companies in the global industrial sector. 🚀
An origin story shaped by crisis and talent scarcity
ARUM started out as an ODM — original design manufacturer — meaning it designed and fabricated precision parts on demand for companies in the automotive and semiconductor sectors. The work was technically complex, required a high level of expertise, and yet profit margins remained pretty thin. It was the kind of business that worked fine as long as market conditions stayed stable, but suffered badly when things tightened up.
And things tightened up dramatically in 2008 with the global financial crisis. Many of ARUM’s partners, smaller companies also involved in metal processing, simply went under and disappeared. Hirayama realized then that a growing gap was forming in the market. If there was demand for precision parts and fewer and fewer companies capable of making them, there was also a massive opportunity for anyone who could automate that process and offer it as a service or product.
At the same time, Japan was dealing with a structural problem that only got worse over time: a declining number of skilled workers capable of operating CNC machines and performing metal machining at the precision levels the industry demanded. This type of equipment, essential for manufacturing automotive, electronic, and semiconductor components, depended on operators trained over years, and the younger generation simply was not entering the field. Hirayama saw this bottleneck not just as a problem, but as motivation to pursue full automation using artificial intelligence.
The birth of ARUMCODE and TTMC
ARUM’s original concept was to automate metal processing, which consists of 12 distinct processes. ARUMCODE was created to cover the first three, which involve machine programming. It is a CAM programming automation software that automatically generates the toolpaths needed for metal part machining. What used to take hours of work from a human specialist could now be done in minutes, with far less room for error and without relying on a large, expensive team.
From there, the next step was a natural one: automate the remaining nine processes. That is how TTMC was born — ARUM’s fully automated milling machine. While ARUMCODE handles the intelligence behind the programming, TTMC is the hardware solution — a line of machining equipment designed to operate in full integration with the company’s software. Together, they cover the complete metal processing cycle, from digital planning to the finished part, creating a closed and highly efficient ecosystem.
The development was neither fast nor simple. Hirayama says that in the beginning, ARUMCODE was a piece of software that simply could not produce anything usable. The team ran tens of thousands of tests, adjusting, refining, and letting the system learn from each iteration. It was not until around 2020 that the software reached a level of maturity sufficient to be considered functional in a production environment. At that point, Hirayama had the confidence that the product would sell and that ARUM could bet on it as the flagship of the business.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
Profits multiplied and a new brand identity
The strategic pivot delivered impressive financial results. According to Hirayama, comparing the company before and after the development of ARUMCODE and TTMC, profits increased eight to ten times. That is a remarkable transformation for a company that started out with tight margins manufacturing parts on demand.
Beyond the financial impact, there was an important shift in brand perception. When ARUM was just an ODM, it was extremely rare for the company’s name to have any visibility at all. Parts left the factory with no identification of the original manufacturer. With ARUMCODE and TTMC, the products started carrying the ARUM name, and the brand began to be recognized not only in Japan but worldwide.
The company also went through a transformation in its business model. Today, ARUM considers itself a fabless company — focused on design and development, leaving precision parts manufacturing to partner companies. It is a radical shift from the company’s origins and reflects a mindset more aligned with tech companies than with traditional industrial manufacturers. For fiscal year 2026, the goal is to reach revenue of 7 billion yen, roughly $44.4 million, with a net profit of 2 billion yen, around $12.7 million — all with a team of just 40 people. The company also plans to go public in 2030.
Without ARUMCODE or TTMC, Hirayama estimates revenue would have stayed at just 1 billion yen, about $6.4 million, with a profit of maybe 50 million yen, a little over $318,000. The company would have survived as a subcontractor, with steady work coming in, but without big ambitions and without any prospect of competing on the international stage.
Generative artificial intelligence as the next leap
With more than 200 Japanese companies already using ARUM’s solutions, the company’s installed base represents a valuable and strategic asset. Each of those factories continuously generates data: information about cycle times, tool wear, cutting parameters, programming errors, and adjustments made by operators. That data, when properly handled, is pure gold for any artificial intelligence model. And that is exactly what ARUM is doing now — using this rich and diverse history to train and feed AI systems that will make ARUMCODE even more powerful and autonomous.
The integration with GPT-5, OpenAI’s large language model available through Azure OpenAI on Microsoft Foundry, is one of the company’s boldest moves in this direction. The idea is to let factory operators interact with machines using natural language. In practice, this means a person can start operating the equipment on their very first day at the factory, without extensive technical training. Just describe what you need in plain language and the system takes care of the rest.
Currently, the system is programmed in Japanese, but translating it into other languages using AI is straightforward and entirely feasible, which greatly supports the internationalization plans. This further democratizes access to advanced automation, allowing smaller companies with less specialized teams to operate at the same efficiency level as major players in the manufacturing sector.
On top of that, combining generative AI with real shop floor data creates a virtuous cycle of continuous learning. All design data and processing formulas are collected automatically and can be used to fuel machine learning. Every new part machined, every adjustment made, every problem solved improves the model and its future suggestions. Over time, the system becomes increasingly precise, anticipating problems before they happen and recommending optimizations that a human programmer would struggle to identify just by looking at raw data. It is the kind of applied intelligence that turns a factory into a living organism, capable of learning and adapting in real time. 🤖
A cloud-connected manufacturing network
One of ARUM’s most ambitious near-term plans is the creation of a procurement network, or supply network, connecting more than 100 TTMC units spread across various locations in Japan. All of it managed through a consolidated platform on Microsoft Azure.
The logic behind this network is as practical as it is clever. Hirayama uses an example to illustrate: if the Tohoku region in northeastern Japan is hit by a major earthquake and the local factory goes offline, a TTMC unit in the south of the country — say in Kagoshima — can take over production and manufacture the needed parts without any interruption. It is industrial resilience in its most tangible form.
Beyond geographic redundancy, this network allows ARUM to serve multiple sectors simultaneously, including automotive, semiconductors, and defense. And since the company operates the cloud infrastructure connecting all these machines, it positions itself as what Hirayama calls a manufacturing infrastructure vendor — essentially an industrial backbone for these industries.
Global expansion focused on three strategic markets
Success in Japan has opened the door to an ambitious expansion plan. ARUM is targeting the U.S., South Korea, and India, with exports planned for August 2027, assuming all necessary approvals are secured. These are three regions with very different characteristics, but they share a common denominator: a manufacturing industry in the middle of a digital transformation, hungry for solutions that boost efficiency without blowing up costs.
In the United States, the reshoring movement — the trend of bringing production lines back from overseas — is creating enormous demand for industrial automation. American wages require factories to be far more productive per worker than their Asian counterparts, and solutions like ARUMCODE and TTMC fit perfectly into that equation.
In South Korea, the landscape is different but equally favorable. The country has a highly sophisticated electronics and semiconductor industry, with component suppliers that need extreme precision in part machining. The level of control and repeatability offered by ARUM’s solutions is critical for meeting the quality standards demanded by the region’s major chipmakers.
In India, the appeal leans more toward cost-effectiveness: a country with a growing industrial base, intense pressure for global competitiveness, and a chronic shortage of specialized technical professionals. That is exactly the problem ARUM was built to solve.
To make this expansion viable, the company is investing in product localization and multilingual support. The integration with Azure OpenAI also helps in this process, as it allows the system interface to be adapted for different languages much more fluidly than a traditional software approach would permit. An operator in Mumbai or Detroit will be able to interact with ARUMCODE in the same way as an operator in Kanazawa, with the same ease and efficiency.
The pride of someone who invented it and delivered it to the world
When asked what makes him most proud looking back at the entire journey, Hirayama is straightforward: ARUM was born from his idea. He is the original inventor, and now he gets to offer that technology to the world and receive appreciation from customers. For him, that is the greatest source of joy and pride.
It is a simple statement, but it carries the weight of nearly two decades of persistent work, countless tests, adjustments, and fresh starts. It is not every day that someone manages to turn a personal idea into a product that changes the dynamics of an entire industry.
What this story means for the future of industrial AI
ARUM’s trajectory is a concrete example of how artificial intelligence is moving beyond conferences and research labs to become a practical tool integrated into the daily operations of factories. What Hirayama built over nearly two decades is not just a product — it is a methodology for solving real problems using technology in a progressive and consistent way. First, he automated the programming process with ARUMCODE. Then, he closed the loop with TTMC hardware. Now, he is adding a layer of generative artificial intelligence that promises to make the system increasingly autonomous and accessible.
This model of gradual evolution is interesting because it shows that AI adoption in manufacturing does not have to be a massive disruption. The most successful solutions tend to be the ones that integrate with what already exists, amplifying the capabilities of systems and people without creating a learning curve that is impossible to scale. ARUMCODE was already a powerful tool before AI entered the picture; with it, the software becomes an intelligent platform that learns, adapts, and evolves alongside every factory that uses it.
For anyone following the advance of artificial intelligence in industry, ARUM is a case worth paying attention to. Not because of the hype, but because of the substance: a company founded with a clear purpose, one that built a solid foundation over years, accumulated real operational data, and is now using those assets to make a qualitative leap with AI. It is the kind of story that tends to age really well, and one that can serve as a reference for many other companies still trying to figure out how to turn the promise of smart automation into concrete results on the factory floor. 💡
