Ajinomoto is a Japanese food company founded in 1909, and it still commands a large share of the global MSG market. Yet in the 2020s, its name began appearing in semiconductor analysts' reports — not because of its food business, but because a specialty insulating film, developed through its amino acid research, had become a critical component in the high-performance chip packaging used for AI computing.
Around the same time, TOTO, the world's largest maker of sanitary ceramics, traced a similar arc. The precision materials and sensor expertise the company built over more than sixty years of bidet-toilet technology fed directly into AI-powered health monitoring. Products that detect early indicators of diabetes and kidney disease by analyzing urination patterns are now moving through concrete commercialization steps in Japan's healthcare and insurance markets.
In May 2026, The Economist analyzed the two companies as a single phenomenon: the beneficiaries of the AI boom are not only IT firms like Nvidia and OpenAI. Quieter, steadier AI-related profits are coming out of decades-old analog businesses — toilets and seasonings. What this means for Korea's solo founders and one-person businesses is anything but simple.
How an Amino Acid Company Ended Up in the AI Chip Supply Chain
Ajinomoto Build-up Film (ABF) is an insulating material used in semiconductor substrates. A spin-off from decades of research into amino acid compounds, it began shipping to Intel and other major chipmakers in the late 1990s. As demand for AI computing surged, shortages of the material came to be cited as one of the bottlenecks in the global semiconductor supply chain — a measure of how strategically valuable it had become. Ajinomoto makes no AI chips and writes no AI software. But inside the world's major AI chips sits this company's material.
TOTO's path is different. As a product category, the toilet spent decades on the margins of technological innovation. But within Japan's culture of precision manufacturing, TOTO steadily advanced the technology packed into the Washlet — its warm-water bidet toilet — integrating compact heaters, fine water-flow control, and seat-occupancy sensors. That technology stack expanded into collecting and analyzing urination data, becoming an infrastructure that automatically gathers biometric data, in an everyday space, in a form AI models can process. To the company, a toilet is still a toilet. To the AI healthcare ecosystem, that toilet is a data collection terminal.
What the two companies share is that neither develops AI nor sells AI services. They supply the materials, components, and data-collection infrastructure the AI ecosystem needs, backed by decades in their trades. The Economist described this position as the AI boom's indirect beneficiaries: businesses that have long supplied something AI cannot run without are collecting profits, even though they never built AI themselves.
Why Solo Entrepreneurs Shouldn't Consume This Story as Easy Optimism
Before taking these cases at face value, there are points worth examining with a cold eye.
ABF's success owes less to strategic foresight than to long-term research investment and a fortunate technological fit. Ajinomoto did not set out in the 1990s to develop insulating materials because it had predicted the long-term growth of the semiconductor market. Years of deep work in amino acid chemistry produced a film material with an unusual molecular structure, and that material happened to match what the chip industry needed. It was an outcome possible only for a large corporation that could sustain a major R&D organization for decades.
And there is a clear counterargument. Ajinomoto and TOTO hold near-monopoly technology positions built on decades of capital investment and hundreds of researchers. Critics point out that small manufacturers and one-person businesses are structurally incapable of replicating this path. In Korea especially, small operators are often locked inside conglomerate subcontracting chains or platform dependencies, with little room to accumulate independent technical capability at all. If the message that analog industries have AI opportunities too is consumed too optimistically, solo founders end up projecting onto their own situation a story that holds only for players with capital and scale.
Much of that criticism is correct. But strip the scale out of the two cases, and the mechanism underneath is simple: whoever holds something AI needs becomes a beneficiary. That something might be a specialty material, data from a particular field, or hands-on experience AI cannot easily reach. It doesn't have to be big. If it's scarce, it has value.
Someone who has brokered real estate in one neighborhood for twenty years holds unlisted transaction histories and building-condition records for that area. Someone who has run a small manufacturing floor for ten years has accumulated the defect patterns of specific processes and the fixes that work on site. AI companies cannot realistically scrape this from the internet. Just as a toilet maker holds an exclusive infrastructure for collecting urination data, people who have worked one trade for a long time hold data access and field knowledge that AI struggles to replicate. And where those two things overlap, transactions are starting to happen.
The Roles Analog Businesses Can Find in the AI Ecosystem
Carry the structure of the Japanese cases over to Korea's small-business environment, and a few concrete paths come into view.
The unstructured data accumulated by someone who has worked an industry for years is a resource AI training needs but cannot find online: site photographs, customer consultation records, local transaction customs, failure cases from specialized processes. In some industries, contracts are already being signed to supply this data to AI developers, or to package it into specialized training datasets for sale. What makes these deals possible is time spent in the trade — and the records that piled up inside it.
There is also the role of reviewing and finishing what AI produces, against working professional standards. AI drafts a legal document; a lawyer edits it. AI organizes medical information; someone with clinical experience reviews it. Only people with years of real practice in a field can carry this work, and as AI tools grow more capable, the market value of this verification role keeps climbing. In Korea, there are reports of freelancers in these roles beginning to gain the upper hand in rate negotiations.
Field execution is another role. However advanced AI agents become, someone still has to do the actual interior renovation, conduct the on-site interview, machine the specialty material. The more AI automates the planning, analysis, and communication stages, the more the capacity to handle the final execution stage acquires standalone business value. The shift is already faintly visible among small interior contractors, on-location photography specialists, and neighborhood logistics operators.
Practitioners who have repeated the same work in one field for years tend to discover a common sequence. First, find the most time-consuming bottleneck inside work you already perform well and repeatedly; cut that bottleneck with AI tools; then spend the recovered time expanding customer contact or winning verification work. Run the sequence in reverse, and you end up with more AI tools and the same revenue structure. Just as the MSG company found a semiconductor material by digging ever deeper into amino acid research, the starting point is locating the bottleneck and the scarce resource inside the work you do now. Adopting AI tools first and looking for uses afterward rarely makes this structure work.
When Ajinomoto and TOTO found, inside work they had done for decades, territory AI could not easily touch, that territory became a tradable resource in ways no one expected. It is a different path from leaping into a brand-new AI business. The question this story poses to Korea's solo entrepreneurs sits in the same place: within the work you are doing right now, where is the part AI finds hard to replicate — and how, and to whom, could that part be sold?
