Most people have never heard of Ajinomoto Build-up Film, yet factories in Taiwan, the US and South Korea cannot ship their most advanced processors without it. Behind almost every cutting-edge Nvidia AI chip, a century-old Japanese food company sits in an unexpected position of power.
From noodle seasoning to chip supply choke point
For most Japanese shoppers, Ajinomoto is the brand on instant noodle packets and bottles of seasoning. The group posts billions in annual revenue from food, sweeteners and additives such as monosodium glutamate (MSG). Hidden inside this consumer empire sits a high-tech materials division that has become critical to global chipmaking.
Ajinomoto now produces about 95% of the world’s supply of a specialist insulating film used in advanced semiconductor packaging. That material, known as Ajinomoto Build-up Film, or ABF, is baked into the substrates that hold modern CPUs and GPUs together.
ABF is a thin insulating film that lets chipmakers pack far more connections into a tiny area without short circuits, heat failures or warping.
Every time Nvidia, AMD or other chip designers push to higher performance, they demand even denser substrates. That keeps Ajinomoto’s material at the centre of the value chain, even if its name never appears on the box.
A taste experiment that led to umami – and then to microchips
The chemistry roots go back to a kitchen in 1908
The company’s technological story starts far from clean rooms. In 1908, Tokyo scientist Kikunae Ikeda tried to understand why his wife’s seaweed broth had a deep, savoury taste that did not match sweet, sour, bitter or salty. By isolating glutamate from the kelp, he identified what would later be called “umami”, often described as the fifth taste.
Ikeda patented a process for producing glutamate seasoning and, in 1909, founded Ajinomoto, meaning “the essence of taste”. From the beginning, the company focused on fermentation and chemistry as much as food. Over decades, research into amino acids, fermentation by-products and industrial chemistry built a base of expertise that would later translate into electronic materials.
Turning chemical waste into a strategic asset
In the 1970s, Ajinomoto faced a problem. Manufacturing amino acids at scale created large volumes of chemical by-products. Treating or dumping them was expensive. Researchers began to test these residues more carefully, looking for useful properties instead of simply disposing of them.
They found some compounds with strong insulating behaviour, good thermal stability and predictable mechanical responses under heat and pressure. At the time these traits had no obvious market. But Ajinomoto kept the work alive, refining the materials and recording how they behaved.
➡️ “My landlord didn’t believe it” – until I fixed mould using just salt and an old sock
➡️ Major breakthrough: a Japanese drug could regrow teeth within 5 years
➡️ I realized some cleaning tasks were stealing time without improving anything
➡️ The simple psychological trick to politely say “no” to any request without feeling guilty
➡️ The hole in your pan handle isn’t just for hanging: here’s the little-known use
➡️ 3I/ATLAS: scientists detect a strange radio signal coming from the interstellar comet
By treating industrial waste as a potential resource, Ajinomoto stumbled on the foundations of a future semiconductor material.
Two decades later, that decision paid off when the chip industry reached a scaling wall.
Intel’s miniaturisation crisis opens a door
When traditional insulating inks stopped working
By the mid‑1990s, Intel and its rivals were pushing lithography to new limits. On the packaging side, electrical traces on chip substrates were moving closer together. The insulating inks used to separate those traces were starting to fail.
Manufacturers were wrestling with bubbles, uneven drying, chemical contamination and random defects that ruined yield. As wiring densities climbed, even tiny imperfections translated into unstable chips or complete failures.
Somewhere in Intel’s supply chain, a materials engineer suggested a different route: instead of printing liquid ink, what about a continuous insulating film that could be laminated, drilled and etched with high precision? That question landed on Ajinomoto’s desk.
Four months to build a film no rival could match
Ajinomoto accepted the challenge and mobilised chemists and process engineers. Within roughly four months, they produced a new material: a uniform, ultra-thin film that could be laminated onto substrates, then laser-drilled and patterned without trapping air or deforming under heat.
This became Ajinomoto Build-up Film. It changed how high-end substrates were made.
- Withstands temperatures above 200°C without losing insulating strength
- Can be etched and drilled at micrometre scale while staying dimensionally stable
- Supports direct copper integration for high-speed signal paths
- Maintains performance as line widths and spacing shrink each generation
Each new chip design forces tweaks to the ABF recipe: thickness, thermal expansion, laser behaviour and compatibility with new production steps. That constant tuning has created a moving target for would-be competitors.
How Ajinomoto ended up inside Nvidia’s AI monsters
Advanced packaging makes ABF unavoidable
Modern AI accelerators like Nvidia’s latest GPUs are no longer just flat pieces of silicon. They are multi-layer structures, often combining several compute dies with stacks of high-bandwidth memory (HBM) on a single base.
TSMC, Nvidia’s main manufacturing partner, uses a technology called CoWoS, short for “Chip-on-Wafer-on-Substrate”. In this scheme, the GPU die and HBM stacks sit on an intermediate base known simply as the substrate.
That substrate looks like a tiny, extremely advanced circuit board. It must support thousands, sometimes tens of thousands, of electrical connections. Signals travel at very high frequencies. Power delivery is intense, and the structure has to handle repeated heating and cooling without bending or cracking.
CoWoS substrates for Nvidia’s AI chips rely on ABF layers to insulate and route dense forests of microscopic copper lines.
Without ABF or an equivalent material, the substrate could not handle such density and heat reliably. Even if chipmakers had perfect silicon from the most advanced fabs, they would not be able to package it into a usable product.
From CPUs to agentic AI chips
ABF first entered mass production in the late 1990s with Intel’s premium processors. Since then, virtually every major chip maker using advanced packaging has adopted it: AMD, Broadcom, Qualcomm and many others.
AI-specific chips raise requirements further. Architectures for “agentic” AI – which perform long chains of reasoning across multiple steps – rely on rapid data movement between compute cores and large pools of memory. That translates into fatter substrates, more layers and even tighter spacing between wires.
Nvidia’s latest platforms, designed for huge AI models and complex workflows, push substrates to their limits. With each generation, more ABF is used per package. That keeps Ajinomoto directly tied to the growth of the AI sector.
A near-monopoly that has already shaken the supply chain
When one film can slow the entire industry
Ajinomoto currently dominates ABF with an estimated 95% share of global production. There are research efforts in other countries to create competing materials, but none has reached comparable scale or reliability.
The risks of this concentration became stark in 2021–2022. The broader chip shortage is often blamed on a lack of wafers or manufacturing tools. Yet several companies saw lead times balloon because they simply could not secure enough ABF-based substrates.
At one point, networking and chip specialist Broadcom reportedly saw delivery estimates stretch to well over a year. The bottleneck was not lithography equipment or silicon wafers; it was the humble insulating film supplied mainly from Japan.
A single Japanese supplier, better known for soup seasonings, can delay cloud data centres, cars and AI servers worldwide if its film runs short.
This kind of choke point worries governments and big tech buyers. It adds another layer of fragility to an industry already exposed to geopolitical tensions and natural disasters.
Rising demand from AI and what comes next
Ajinomoto’s expansion plans
Ajinomoto expects global demand for ABF to rise by high single digits each year, driven by AI accelerators, high-performance computing and advanced consumer chips. The firm has announced plans to boost production capacity by about 50% by 2030.
Several trends are driving that growth:
- AI GPUs and specialised accelerators use larger, more complex substrates
- High-bandwidth memory requires more layers and tighter routing
- Data centres upgrade servers faster to keep up with AI workloads
- Automotive and industrial chips adopt more advanced packaging
Each of these shifts increases the total surface area of ABF per chip. That gives Ajinomoto a direct financial upside from the AI boom, even as it continues to sell instant noodles and seasonings on supermarket shelves.
Key terms and practical implications
Three concepts that shape this hidden dependency
| Term | What it means in practice |
|---|---|
| Substrate | Base layer that connects a chip to the circuit board; packed with tiny copper lines separated by insulating films like ABF. |
| Advanced packaging | Techniques for stacking or tightly coupling multiple chips and memory to boost performance and cut energy use. |
| ABF (Ajinomoto Build-up Film) | Special resin film used layer by layer to insulate and support dense wiring on substrates for high-end CPUs and GPUs. |
For a cloud provider planning next-generation AI data centres, the state of ABF supply is not a niche technical detail. It affects how many servers can be delivered, when upgrades arrive, and how quickly new AI tools reach customers.
For policymakers in the US and Europe, the situation raises questions about resilience. Billions are being poured into local chip fabs and packaging plants, yet one of the core materials remains almost entirely controlled by a Japanese group with food as its main public identity.
Scenarios, risks and possible shifts
Several scenarios are now on the table. If Ajinomoto’s expansion keeps pace with AI demand, the bottleneck might stay manageable, albeit with tight margins during peaks. If demand spikes faster than capacity, lead times for AI chips could lengthen again, slowing cloud rollouts and hardware launches.
Competing materials could gradually reduce Ajinomoto’s dominance, but any challenger must match decades of reliability across extreme temperatures, frequencies and mechanical stresses. A single defect pattern could scrap entire batches of very expensive chips, so manufacturers move slowly when changing materials.
On the upside, the ABF story shows how skills from seemingly unrelated industries can gain new strategic weight. A company that started by analysing the taste of seaweed soup now shapes the pace of AI progress, through a film most users will never see and a name that rarely appears on the label.