Trying to make sense of all the complex components of a single computer chip can be dizzying: multi-layered microscopic components connected to each other by a copper-wire highway, some of which are only a few strands of DNA wide. Between these wires there is an insulating material called a dielectric that ensures that the wires do not touch and short-circuit.
Zooming in further, we can see that a special dielectric is placed between the chip and the structure underneath it, a material called a dielectric film that is as thin as a white blood cell.
Competitors have struggled to outperform them, and today Ajinomoto has more than 90% of the market share for the product, which is used in everything from laptops to data centers.
But now, a Berkeley, Calif., startup is making a tough effort to topple Ajinomoto and bring that small portion of the chip manufacturing supply chain back to the United States.
Thintronics has pledged a product built specifically for the computing needs of the AI era, which the company claims is a new set of materials with higher insulation properties and, if adopted, could mean faster computing speeds and lower energy costs for data centers.
The company is at the forefront of the coming wave of new U.S. companies, fueled by the $280 billion CHIPS and Science Act, which is seeking to get a slice of the semiconductor industry, which is currently dominated by a handful of international companies. But to succeed, Thintronics and its peers must overcome a series of challenges – solving technical problems, disrupting long-standing industry relationships, and convincing global semiconductor giants to embrace new suppliers.
Stefan Pasteine, Founder and CEO of Thintronics, said, "It's very difficult to invent a new material platform and bring it to the world. It is "not for the faint of heart."
ABF's story
If you know the name Ajinomoto, you might be surprised to learn that it plays a key role in chips: the company is better known as the world's leading supplier of MSG flavoring powders. In the 90s of the 20th century, Ajinomoto discovered that MSG was a by-product of MSG that could make good insulators, and since then, Ajinomoto has enjoyed a near-monopoly in this niche material field.
According to Ajinomoto, the story of ABF began in the 1970s, was first adopted in personal computers in the late 1990s, and has evolved to the present day with the improvement of CPU performance.
With the shift from MS-DOS to the Windows operating system, the rise of large-scale integration of CPUs for personal computers, and the increase in the number of terminals from about 1990 in the early days to thousands of terminals, the demand for advanced CPU substrates grew rapidly in the 40s. or more. This led to a shift from a "leadframe" configuration to a CPU mounted on a multi-layer circuit substrate containing complex wiring patterns, which in urgent need of new insulation.
Since the 1970s, the Ajinomoto Group has applied amino acid chemistry to basic research on epoxy resins and their composite materials. This eventually led to the development of advanced insulators for CPU substrates. As a latecomer in this field, the Ajinomoto Group's focus on thin films has allowed the company's products to stand out from traditional ink-type insulators and form a material that solves the major problems posed by the use of traditional insulators in high-performance CPUs. When ABF became available to manufacturers, it met the rapidly growing global demand.
Ajinomoto said that the basic goal of achieving R&D developed by ABF is to find a resin composition that will determine the properties of the insulating material, provide the necessary functions of the electrical material, and promote film formation. The Ajinomoto Group's expertise in fine chemistry is used to develop formulations that combine organic epoxy resins, hardeners, and inorganic particulate fillers. The main challenges include developing a method for homogeneous mixing of organic and inorganic substances that inherently resists uniform dispersion and provides excellent insulating properties and excellent processing characteristics.
To address these challenges, the R&D team created a thermoset film with high durability, low thermal expansion, ease of processing, and other important features. The film, called ABF, was first adopted by a major semiconductor manufacturer in 1999. It has since become the product of choice for almost all high-performance CPUs and is supported by evolving R&D to meet the needs brought about by rapid development in circuit integration. Since then, ABF has been leading the market.
Advances in circuit integration have made it possible to make CPUs composed of nanoscale electronic circuits. These circuits must be connected to millimeter-level electronic components in electronic devices and systems. This can be done by using a CPU "bed" made up of multiple layers of microcircuits, known as a "stacked substrate". ABF helps to form these micron-scale circuits because its surface can accept laser processing and direct copper plating. Today, ABF is an important material for forming circuits that direct electrons from nanoscale CPU terminals to millimeter-level terminals on printed substrates.
Ajinomoto said that with the rapid improvement of CPU performance, the quality of ABF is also improving. This requires continuous development of insulating resins with different properties, improved product properties, processing technologies that meet the requirements of emerging customers, and repeated testing and validation.
Achieving the thermal stability required for a hot CPU environment, optimizing the plating process that is essential for circuit formation, and facilitating laser processing are just a few of the many challenges that require specialized knowledge and expertise. As CPUs evolve and diversify, these skills are essential to produce the best ABF-based circuit boards that meet customer requirements.
Insulator bottleneck
But MIT Technology Review said in an article that Ajinomoto does not produce any other parts in the chip. In fact, the insulation in the chip relies on a fragmented supply chain: one layer uses Ajinomoto materials, another uses another company's materials, and so on, with no layer optimized for working together. The final system works fine when data is transmitted over short paths, but over longer distances (e.g. between chips), weak insulators become bottlenecks, wasting energy and slowing down computations.
Recently, this issue has been a growing concern, especially as AI training has become more expensive on a scale and consumes a staggering amount of energy. Moreover, some problems are beginning to come to the fore.
None of this made much sense to Pastine, a chemist who sold his former company, which specialized in recycling hard plastics, to an industrial chemicals company in 2019. Around that time, he began to believe that innovation in the chemical industry could be slow, and he believed that the same model was holding chipmakers back from finding better insulation. In the chip industry, he said, insulators are "kind of seen as redheaded stepchildren" — they haven't seen progress on transistors and other chip components.
That same year, he launched Thintronics, hoping that cracking code on better insulators would provide faster computing speeds for data centers at a lower cost. The idea is not groundbreaking – new insulators are constantly being researched and deployed – but Pastin is confident he can find the right chemistry to make the breakthrough happen.
Thintronics says it will manufacture different insulators for all layers of the chip for a system designed to replace an existing production line. Mr Pastine said the materials are currently being tested by a number of industry players. But he declined to provide his name, citing a non-disclosure agreement, and likewise declined to disclose details of the formula.
Without more details, it's hard to say exactly how well Thintronics materials compare to the competition. The company recently tested the DK value of its materials, which is a measure of the effectiveness of a material's insulator. Venky Sundaram, a researcher who has founded several semiconductor startups but is not involved in Thintronics, reviewed the findings. He says their most impressive Dk values are better than any other material available today, he says.
Bumpy roads
Thintronics' vision has already gained some support. The company received a $20 million Series A funding round led by venture capital firms Translink and Maverick in March, as well as funding from the National Science Foundation.
The company is also seeking funding for the CHIPS Act. The bill, signed into law by President Joe Biden in 2022, aims to promote companies like Thintronics in order to bring semiconductor manufacturing back to U.S. companies and reduce reliance on foreign suppliers. A year after the bill became law, the government said more than 450 companies had submitted letters of intent in the hope of receiving CHIPS funding to carry out work across the industry.
Most of the legislation's funding will go to large manufacturing facilities, such as Intel's factories in New Mexico and TSMC's operations in Arizona. But U.S. Commerce Secretary Gina Raimondo said she would like to see smaller companies receive funding as well, particularly in the materials sector. In February, applications for $300 million in funding dedicated to materials innovation opened. While Thintronics declined to disclose how much funding is being sought or from which projects, the company does see the CHIPS Act as a major driver.
But building an on-island chip supply chain, which currently relies on dozens of companies around the world, would mean reversing decades of specialization in different countries. Industry experts say it will be difficult to challenge today's dominant insulator suppliers, who often have to adapt to new competition.
"For more than two decades, Ajinomoto has been a material that has a market share of more than 90 percent," Sundaram says. "This is unheard of in most businesses, and you can imagine that they wouldn't have achieved it by not changing. ”
A major challenge is that the dominant manufacturers have decades-long relationships with chip designers such as Nvidia or Advanced Micro Devices, as well as manufacturers such as TSMC. Having these players swap out materials is a big deal.
"The semiconductor industry is very conservative," says Larry Zhu, a semiconductor researcher who has worked in the dielectric industry for more than 25 years. "They like to use suppliers that they already know very well, and they know the quality of those suppliers. ”
Another obstacle Thintronics faced was the technical aspect: the insulation, like other chip components, followed very precise manufacturing standards that were difficult to understand. The layer in which Ajinomoto dominates is thinner than a human hair. The material must also be able to accommodate small holes, which house wires that pass perpendicular through the film. Sundaram says that every new iteration is a large-scale R&D effort, and that the incumbent has the upper hand in this area with years of experience.
If this is all done successfully in the lab, there is another obstacle: the material must retain these properties in a high-volume production facility, and this is where Sundaram's past efforts have failed.
"Over the years, I have consulted with several material suppliers who have tried to break into the [Ajinomoto] business but have been unsuccessful," he says. "They all end up having problems that aren't as easy to use on high-volume production lines. ”
Despite all these challenges, one thing could work in Thintronics' favor: American tech giants such as Microsoft and Meta are making their first progress in designing their own chips. The plan is to use these chips for in-house AI training as well as cloud computing capabilities for renting to customers, both of which will reduce the industry's reliance on Nvidia.
Although Microsoft, Google, and Meta declined to comment on whether they are pursuing advancements in materials such as insulators, Sundaram said these companies may prefer to work with new U.S. startups rather than default to the old way of making chips: "They have a more open approach to the supply chain than the existing big companies." ”
[Source: Semiconductor Industry Observation]