Recently, there’s been a highly anticipated new book in the semiconductor industry, with a title that’s both particularly short and particularly long.
This book is “Focus,” with the subtitle “The ASML Way—Inside the Power Struggle Over the Most Complex Machine on Earth.”
The book was first published in Dutch and English, and I noticed that a Chinese version has already been signed by a domestic publisher. However, after reading it, I can’t help but pour a little cold water: due to the strict requirements of domestic publishing, a large portion of the content, possibly even one-fifth, might need to be deleted or revised.
Initially, everyone thought this book was a complete biography of ASML. While that’s not entirely wrong, the author focuses on describing the current tech cold war and the great power struggles, as well as how ASML struggles and evolves amidst these challenges.
Biden and Trump were mentioned about 30 times each, and Dutch Prime Minister Rutte about 20 times, so you can guess that China leaders also appear quite a bit.
Without further ado, here’s a brief review. This is purely personal sharing, feel free to criticize if you don’t like it.
One, Link to heading
The first third of the book is a bit frustrating because, for those who have read “The Lithography Giant,” the information increment is somewhat limited. Fortunately, the author has greatly simplified similar content from “The Lithography Giant.”
The middle part, about the post-1995 IPO content, interacts well with the “Battle of Lithography Machines” series of articles, with some uncanny similarities that give a sense of camaraderie.
For instance, the author also believes that ASML’s acquisition of American SVG at an unworthy high price was to penetrate Intel. The American identity helped ASML gain benefits in EUV research and development but also forced it to take sides on issues with China today.
Another example is ASML selling shares to Intel, Samsung, and TSMC to acquire Cymer, making a desperate bet on the EUV path. Also, ASML’s official stance has always been silent on Lin Benjian’s contributions to immersion lithography.
Two, Link to heading
I found some of the anecdotes in the book quite intriguing.
At an Intel factory, the yield mysteriously dropped between 1-2 AM and then automatically returned to normal, baffling the technicians. It turned out that it was caused by cows on a nearby farm farting.
Yes, cow farts.
The book doesn’t explain why the cows farted at that specific time, but the methane in the farts indeed drifted into the cleanroom. Intel eventually had to pay to relocate the farm.
Another example involves feng shui, which many wafer fabs are indeed superstitious about. In an Arizona wafer fab, the lithography machine kept malfunctioning, but engineers found everything normal. Even more bizarre, other equipment in the fab also frequently malfunctioned.
The fab manager heard that the fab was built on an old Indian burial ground and invited a feng shui master to take a look. The master found that some flower beds or concrete at the entrance resembled tombstones and needed to be moved. Even more magically, the master donned a bunny suit and inspected the production line, saying the spirits liked red and needed some red.
The manager stacked bright red packaging boxes on the lithography machine, and the problem miraculously disappeared. For years, no one dared to move the box, and even new lithography machines were required to have red boxes placed on them.
Three, Link to heading
At a TSMC wafer fab, there was also an immersion equipment issue; every time it restarted, it inexplicably produced defective chips for the next two hours, then everything returned to normal without any adjustments.
Technicians took months to find the cause. Every time the machine stopped, five drops of water would fall on the immersion layer, causing it to swell by a few nanometers. This was enough to misalign the projection pattern. After two hours, the water evaporated, and the immersion layer returned to its original state, leaving no trace of error.
Such problems only surface in the harsh realities of factory workshops. Even after years of meticulous design, half of the issues these machines face are unimaginable.
During the pandemic, there was also an interesting anecdote. Martin van den Brink (hereafter referred to as Big Martin) disliked teleconferences and wanted to meet face-to-face with TSMC in Taiwan. We all know that at that time, various countries or regions had quarantine policies, and Big Martin’s valuable time couldn’t be wasted in quarantine hotels. So what to do?
Both parties booked a hotel in Hsinchu, with half of the ASML employees staying on one side and half of the TSMC employees on the other, completely physically separated. They were only allowed to leave their rooms for meetings, walking through dedicated corridors under the supervision of “Big Whites” to the conference room. The conference room was divided by an acrylic glass panel, allowing them to see each other but making it difficult for the virus to pass through.
Four, Link to heading
Aside from technical issues, ASML’s biggest problem has always been a lack of funds. The part about fooling Intel is particularly interesting.
We mentioned in “Why the Big Three Invested in ASML” that Intel’s investment was for the 18-inch wafer lithography machine, and they were very skeptical about EUV. ASML spent a huge sum to acquire Cymer, and shareholders were initially reluctant to support it. Big Martin indeed fooled Intel, pretending to develop the 18-inch machine, then saying if they wanted it, they had to place an order first.
Intel wasn’t about to be swayed by PowerPoint presentations; they wanted to see the machines first. In early 2013, Martin orchestrated a meeting with the tech heads of TSMC, Samsung, and Intel, where the former two directly told Intel they weren’t interested in 18-inch wafers.
Intel’s executives were caught off guard by this ambush. At that time, Intel was still at the peak of its power, and it seemed unthinkable for these Asian upstarts to dictate standards. Yet, soon after, both of these upstarts surpassed the giant.
The book highlights two main characters: ASML’s recently retired CEO, Peter Wennink, and Martin. Wennink, who started as an external accountant, is another example of a CFO becoming a CEO, and the story ends well.
Wennink’s crucial mission was to secure funding for Martin, who needed money for extremely expensive, high-risk, and long-term R&D projects. Martin, of course, was there to spend billions of euros on his big toys. This collaboration might not sound extraordinary, but those familiar with current corporate governance and public metrics will understand how unusual it is.
It’s said that Martin successfully sold an EUV machine to Samsung in a hotel after the 2006 SPIE conference, with delivery promised by 2010. Samsung demanded 100 watts of power, but Martin had only achieved 5 watts at the time. Talk about guts.
Then came the 2008 financial crisis, and the chip industry collapsed. ASML was forced to lay off about 1/8 of its workforce and cut all employees’ 13th-month pay or year-end bonuses by half, while its stock price was halved.
It was during this time that you could see how much Wennink (then CFO) supported Martin’s risky EUV gamble. Even with the meager 5 watts of power, TSMC had already announced a shift to multi-electron beam Mapper technology for the next generation (see “ASML and Electron Beam Lithography”).
According to the book, Apple’s decision to distance itself from Samsung and invest heavily in TSMC was the savior, prompting TSMC to commit to EUV.
As for the 100 watts Samsung wanted, ASML took seven years to achieve it.
Five, Link to heading
Another interesting story is how Zeiss helped ASML in its legal battle against Nikon.
We all know that with dual-stage and immersion lithography, Nikon found it hard to catch up technologically, so they resorted to patent wars. After all, having been in the lithography game for so long, accusing someone of infringement was easy. Especially since ASML, in its early days, didn’t pay much attention to patents, leaving a sword hanging over its head.
A familiar scenario unfolded: ASML won all its cases in Dutch courts but faced significant challenges in the U.S. The U.S. courts’ trump card was to ban sales if infringement was found. ASML repeatedly sought a settlement with Nikon, but Nikon refused.
That’s when Zeiss rode in on a white horse. As a century-old giant, Zeiss had a plethora of camera patents. Zeiss, together with ASML, sued Nikon for camera patent infringement. If you ban my lithography machines, I’ll ban your cameras.
Even more astonishing, Zeiss launched a camera called the ZX1, a product with no competitive edge in the smartphone era. The media was puzzled by this move, especially since Zeiss’s last camera was a film camera. The ZX1 was never available for purchase; only a few prototypes were made for influencer marketing.
Zeiss even ran numerous ads for the ZX1 because they needed to prove they had a camera that suffered losses due to infringement for the lawsuit against Nikon. These ads were clearly aimed at Nikon. Nikon got a bit nervous and eventually agreed to settle with ASML. Then, the ZX1 vanished.
Six, Link to heading
With the skyrocketing demand for EUV, ASML’s production capacity couldn’t keep up, and even DUV was in short supply due to other layer needs and sudden demand from China. Martin naively thought that simply doubling the procurement would solve the problem.
In reality, the expansion exposed a series of management issues within ASML.
Many people assumed ASML must be a highly standardized company to produce the world’s most precise equipment. However, like other large companies, ASML’s attempt to standardize KPIs and system processes faced a collapse during the pandemic.
The surge in orders and tens of thousands of parts forced ASML to launch an advanced automated warehouse, 5L. This warehouse, one kilometer long and the size of ten football fields, was equipped with various automated robots and shelves.
The system crashed as soon as it went live in 2021, unable to ship anything. After a week without resolution from IT, employees were sent home. Like other big companies, no one dared to report the issue to the boss. Eventually, Wennink discovered the empty warehouse during a tour with guests and was shocked.
ASML then rallied thousands of employees to manually ship various parts without using the system. The severe issues caused by 5L persisted for over three months. The severe chip shortage at the time hit various sectors, especially the automotive industry, hard, with chip manufacturers even willing to accept untested lithography machines rather than wait.
But uniquely, no one was fired over this; punishment was never part of ASML’s culture.
ASML acknowledged that in a rapidly growing company, organizational and process optimization is impossible, and dynamically solving problems as they arise is more efficient. Arguing over why issues weren’t prevented in advance is less productive than resolving them and moving on. Martin represents a free technological culture, detesting the inefficiencies and departmental silos brought by processes and KPIs.
Martin despises employees who flatter and practice submissive upward management, encouraging challenge and debate. He knows that only in this way can various talents maximize their potential in their areas of expertise, and in such an environment, pretending to understand while presenting PowerPoints would be questioned to death.
Seven, Link to heading
Finally, the book’s content about China is not convenient to discuss here, so let’s briefly describe how ASML mutated in its struggle.
In this book, we see ASML and the Netherlands constantly striving to resist U.S. export controls, making every effort to fulfill China’s surging DUV orders. To cope with the doubling of shipments in recent years, ASML’s workforce has also expanded rapidly, with a quarter of employees being new hires within a year by 2023. However, in terms of technical confidentiality, ASML has made a sharp turn.
ASML was originally a company that valued an open culture and did not strictly manage the security of its technical documents, as they needed to closely collaborate with many technical partners, customers, and research institutions.
ASML takes a distinctly pragmatic approach to confidentiality. They believe that in the technological race, there is only one way to win—you must run faster than your competitors. Of course, they are also quite confident, as countless know-how is hidden in different components and even on the supplier side, making replication indeed difficult.
In our article “ASML’s Last Stand,” we mentioned that the PAS5500, which made Big Martin famous, later became a legendary machine that remained popular for decades, with nearly 2,000 units still in operation worldwide today. ASML grew somewhat weary of maintaining these old machines, and about five years ago, they seriously considered granting all PAS5500 technology to China’s SMEE. Then came the events involving Trump, and that was the end of it.
In 2014, some Chinese employees from ASML’s Silicon Valley branch started anew with optimization software. This is the XTAL incident, which we won’t describe in detail here. This incident triggered ASML’s anxiety about industrial espionage, and as the trade war escalated, ASML gradually shifted from being open to becoming more sensitive.
In recent years, ASML has also claimed to have frequently suffered from espionage and hacking activities, recording 2,800 cyber incidents in 2022 alone, ranging from ransomware attacks to attempts to steal intellectual property.
The number of Chinese employees at ASML is astonishing, partly thanks to China’s excellent university engineering education and enrollment expansion. ASML indeed finds it challenging to find enough engineers from other countries to replace them globally. Perhaps because of this, it wasn’t until 2021 that ASML began conducting background checks on new hires. Later, the Netherlands followed the U.S. in imposing restrictions on the “Seven Sons of National Defense” studying abroad.
Eight, Link to heading
We can’t elaborate much more on this topic. Finally, let’s take a glimpse at the energy issue described in the book.
“The lithography machine draws 1.5 million watts from the grid to produce a 30-kilowatt laser, generating 100 watts of EUV light. Of this, about 1 watt of light ultimately falls on the wafer.” (Note: The figures in the book were from an earlier time.)
Readers can easily calculate the extremely low energy efficiency, which is unimaginable in other fields. It has been estimated that TSMC, with 100 EUV machines, could consume about one-tenth of Taiwan’s entire electricity supply.
Although scientists are still striving to recover the massive heat and energy loss from lithography machines for heating or other purposes, this power-hungry beast continues to evolve and grow. Moreover, the AI chips it produces will lead to even more terrifying power consumption. Even the boldest predictions now may underestimate the massive future energy usage.
Of course, there’s a paradox: without these EUVs, the chips produced would consume even more power.
Overall, this book is still worth reading. The author’s ambition is greater, attempting to outline ASML, a company forced into the spotlight, from a broader perspective, including geopolitics, energy and environmental protection, human resources, cultural business warfare, and more.