Since joining the Dutch company ASML in 1984, Martin van den Brink has served here for 40 years.

In 2024, it is the 40th anniversary of ASML's establishment, and also this year, van den Brink is about to retire from his position as Co-CEO and Chief Technology Officer.

From obscurity to becoming the industry leader

Van den Brink said that his arrival at the company was purely coincidental. At that time, Philips was one of the few technology companies in the Netherlands, so he joined here. In 1984, he became interested in a photo of a lithography machine inside the company.

Van den Brink said in an interview with MIT Technology Review: "I found that this photo contains knowledge in the fields of mechanics, optics, and software, and it looks like a complex machine. My colleagues told me that I could continue to develop it, but the new company would not be part of Philips."At that time, Philips was in the process of establishing a joint venture with ASM International, and after the joint venture, it would no longer be Philips. But I agreed, because I didn't care about the outcome, and that's how it all started.

Advertisement

Undoubtedly, ASML is the undisputed leader in today's chip ecosystem. However, you might find it hard to believe that ASML's dominant position can only be traced back to 2017. That year, ASML released its EUV (Extreme Ultra Violet) machine, which took 17 years to develop, and it revolutionized the traditional chip manufacturing process as soon as it was introduced.

This is the result of ASML's steadfast investment despite being undervalued. Since the 1960s, lithography technology has enabled computer chips to encapsulate an increasing number of components.

This process includes: guiding the beam of light through a series of mirrors and lenses, illuminating the light on a mask containing patterns, and creating miniature circuits.

When the complete circuit is constructed, it also signifies the birth of the computing modules required for smartphones and artificial intelligence. For lithography machines, they must use small-sized design tools, but the number of these tools is very limited, and the type of light is the key among them.In the 1960s, photolithography machines used visible light. However, the smallest features that this light could draw on a chip were still quite large, akin to drawing a portrait with a marker. Consequently, photolithography machine manufacturers gradually began to use light with shorter and shorter wavelengths.

In the early 1980s, people were already able to manufacture chips using ultraviolet light. At that time, Nikon and Canon were the main industry leaders. It was not until 1984 that ASML, as a Dutch subsidiary of Philips, was just established and was still an unknown entity, hardly comparable to other mainstream photolithography machine manufacturers at the time.

Van den Brink said, "It was not until the 1990s that our products began to sell well. During the period before that, we were on the brink of bankruptcy several times. So, for us at that time, there was only one mission: to survive and demonstrate to customers that we could make a difference."

By 1995, ASML had already established a foothold in the industry, not only able to compete with rivals Nikon and Canon, but also completed a public listing. However, all photolithography machine manufacturers, including ASML, were striving to manufacture smaller components on chips.

At the end of the 1990s, in order to stand out from a crowd of competitors, ASML held a meeting and proposed an idea called extreme ultraviolet lithography (which was still a rumor at the time). However, the participants were also concerned that it might never work.At that time, the demand for miniaturizing chips far exceeded the limits of technological capabilities. It seemed that every company was chasing after the EUV lithography machine, which uses a smaller wavelength of light (13.5 nanometers at the time) to pattern the chips.

To achieve this, ASML had to figure out how to create, capture, and focus this light. These processes had already troubled researchers in the field for decades, not to mention establishing a supply chain for specialized materials, including making the smoothest mirrors ever and ensuring that the price would not scare off customers.

Meanwhile, Canon and Nikon were also developing EUV, but the US government did not allow them to participate in the joint research with ASML and US national laboratories. As a result, Canon and Nikon both withdrew from this field. In the meantime, ASML acquired SVG, the fourth company in the industry developing EUV, in 2001.

By 2006, ASML had only delivered two EUV prototypes to research institutions, and it was not until 2010 that the first one was handed over to a customer. Five years later, ASML warned in its annual report that the sales volume of EUV was still very low.

Given the slow manufacturing speed of this technology on the production line, ASML's customers were not eager to adopt it. If this situation continued, it could have a substantial impact on ASML's business, considering the huge investment made in the early stages.However, in 2017, after 17 years and an investment of 6.5 billion dollars in research and development funds, ASML's bet began to pay off. That year, ASML delivered 10 EUV machines, each costing over 100 million dollars, and at that time it announced that there were dozens of machines still in the backlog.

These EUV machines were sold to semiconductor manufacturing giants such as Intel, Samsung, TSMC, and a few other companies. With a brighter light source, it means that less time is needed to transfer patterns, so lithography machines can work faster.

For chip manufacturers, EUV is a tool that can bring great economic benefits, which is why ASML has been in a monopoly position since 2017.

Chris Miller, a history professor at Tufts University in the United States and author of the book "Chip Wars: The Battle for the World's Most Critical Technology," said in an interview with MIT Technology Review: "Investing in unpromising technology is a kind of stubborn obsession.

At that time, no one was willing to bet on EUV because it involved the limits of physics, engineering, and chemistry, so its development process was both lengthy and expensive." A key factor in ASML's growth is its control over the supply chain. ASML acquired some companies that often cooperated, such as the light source manufacturer Cymer. This targeted supply chain control strategy also extended to ASML's customers. In 2012, ASML provided shares to its three major customers, Intel, Samsung, and TSMC.And the reason why these three major customers have been able to maintain their market-leading position is partly due to their reliance on ASML's high-end manufacturing capabilities. "Our success depends on their success," van den Brink told MIT Technology Review.

With 7 airplanes and 50 trucks, the first high-numerical-aperture machine was delivered to Intel.

Now, ASML has launched a high-numerical-aperture (high-NA) extreme ultraviolet lithography machine and is planning the development of a super-numerical-aperture extreme ultraviolet lithography machine. That is to say, ASML has begun to strive to control another variable - the numerical aperture.

This is a measure of how much light a system can focus. If ASML can increase the numerical aperture, then ASML's machines can print smaller components. However, doing so also means countless changes.ASML had to procure a larger, ultra-smooth mirror set from its supplier Carl Zeiss. At the same time, Zeiss had to manufacture an entirely new machine, with the sole purpose of meeting the smoothness required for the mirrors needed by ASML.

Concurrently, ASML also had to strive to reduce the high costs that this change might bring to other supply chain companies. In December 2023, ASML began delivering the first high numerical aperture machine of its next-generation lithography equipment to Intel's factory in the United States.

Intel is also the first customer to place an order for this product, and it had already placed the order at the beginning of 2022. The machine that Intel received is a research and development version, and it is also the only one in this field to date.

ASML used 7 airplanes and 50 trucks to transport it to Intel's factory, and it took 6 months to install this machine, which is larger than a double-decker bus. For high numerical aperture machines, their mission is to produce the most precise advanced chip layers.

However, for many other chip designs, the previous generation EUV machine, or the older DUV (Deep Ultra Violet, deep ultraviolet) machine, will still be used for printing.At present, ASML's existing EUV customers have already placed orders for high-numerical-aperture machines with ASML. It is reported that the cost of just one such machine is 380 million US dollars. Miller said, "There are obvious benefits and obvious risks for Intel to be the first user."

He stated that figuring out which chips these machines can be used for, and how to make money from them, will be a challenge for Intel. And if the launch of these machines is really successful, it could become the pinnacle of Van den Brink's career.

At the same time, around the beginning of the COVID-19 pandemic, people's interest in the field of artificial intelligence and the progress in this field have driven a surge in demand for computing power. Companies like OpenAI need increasingly powerful computer chips.

By the end of 2022, the frenzy and investment in artificial intelligence began to gradually boil. Currently, Van den Brink has already started the next step of work. He and other ASML executives interviewed by MIT Technology Review said that the company's next big goal is super-numerical-aperture technology.

The aperture of ASML's high-numerical-aperture machine is 0.55 nanometers, while the aperture of the super-numerical-aperture tool will exceed 0.7 nanometers. This means that if the super-numerical-aperture route can be successful, ASML will be able to build machines that allow chip manufacturers to further reduce the size of transistors.Of course, the premise is that researchers can design chip components that can work normally at such a small size. Just like EUV in the early 21st century, it is still uncertain whether a super numerical aperture is feasible, after all, its cost will be very high.

However, Van den Brink is cautious and optimistic. He said that if everything goes well, ASML is very likely to eventually provide three products: low numerical aperture, high numerical aperture, and super numerical aperture. "The risk of super numerical aperture is greater. Therefore, we will be more cautious about cost issues in the future. But if we can achieve this, we will have an invincible 'triple carriage', which will monopolize all advanced manufacturing in the foreseeable future." Van den Brink said.

"There is no reason to believe that Moore's Law will come to an end."

The real issue may not be who manufactures these machines, but whether Moore's Law will continue to hold true. For many years, with the lithography machines produced by ASML, chip manufacturers have been able to produce the world's most advanced chips.

If Gordon Moore is praised for formulating the Moore's Law that drives the industry's progress, then Van den Brink and ASML are praised for enabling technological progress. Therefore, ASML must strive to maintain a lead in the requirements of Moore's Law.Moore's Law states that the number of transistors on an integrated circuit will double approximately every two years. Essentially, this means that chip manufacturers are always trying to shrink the transistors on microchips so as to pack more transistors onto the same size chip.

However, as the size of transistors has been reduced to just a few nanometers, it is increasingly difficult to maintain this doubling pace. In recent years, ASML's lithography machines have been delaying the end of Moore's Law. Nowadays, all high-end chip manufacturers around the world must rely on lithography machines to mass-produce chips.

Van den Brink said that it is Moore's Law that has driven the industry to move forward year after year. Van den Brink used a rice and chessboard analogy to illustrate how maintaining Moore's Law has been a great achievement since he joined ASML in 1984.

In this analogy, the number of rice grains represents the number of transistors, and now the number of rice grains can double on each consecutive square of the chessboard.

Since 1959, the number of transistors that can be packed into a chip has grown exponentially, which means that a grain of rice at that time is equivalent to three oil tankers full of rice now, and each tanker is 240 meters long.Even so, Moore's Law continues to drive ASML and the entire technology industry forward. Van den Brink explained that every era of computational power explosion, as well as the recent generative artificial intelligence era, has brought an increasing demand.

In other words, while three ships full of rice may sound a lot, tomorrow humanity may need six, twelve, or even twenty-four ships to carry this rice.

Van den Brink said that ASML's technology can ensure to meet the above-mentioned demand, because ASML has made a substantial investment in creating finer and more advanced tools.

NVIDIA CEO Huang Renxun announced that Moore's Law is dead. But when Van den Brink was asked what would ultimately lead to the failure of Moore's Law, he completely denied the premise.

"There is no reason to believe it will end, you won't hear from me where it will end," he said, "It will end when we run out of ideas. It will end when the value we create with all this cannot be balanced with the cost required. Not because of a lack of creativity." He believes that even if he is no longer there, his successor still has the opportunity to "fill the chessboard with rice."