Technology invades the modern world

Chapter 451 Where is the Limit? (6K)

This is the last day of 2025. The evening rush hour in Shenzhen has passed, and many migrant workers who had been waiting for the company shuttle bus, dawdling around until 11 o'clock, have already returned to their residences.

At this moment, Lin Wenjie's office light was still on. He was the vice president in charge of supply chain at a chip design company in Shenzhen, China.

Those words just now came from his mouth.

On the other end of the video call was Chen Chih-ming, an account manager at TSMC in Taipei and Hsinchu.

After hearing this, Chen Zhiming looked somewhat dazed. He had heard similar words many times before: we need to reduce risk and adjust orders.

Although the words are similar, there are many subtle differences between the past and the present.

The first is language.

In the past, I would hear such words in English, Japanese, Korean, or even Hokkien, but never in Mandarin.

These global companies, operating under the White House's direction, speak different languages ​​but share the same demands: we need to adjust our supply chains, we need to consider de-risking, and we are diversifying our supply chain orders to other countries.

Chip production orders still go to TSMC, but downstream packaging and testing used to be handled by Chinese companies like JCET, ChipMOS Technologies, and Tongfu Microelectronics, which have a cost advantage.

But now, considering the need to mitigate risk, we want to place orders with packaging companies in other countries and support their growth.

As for customers in mainland China? These Mandarin-speaking customers are completely at his mercy. Forget about risk mitigation, whatever TSMC gives them, they will swallow it all.

But now, with a professional smile on his face, Lin Wenjie's previous statement was: "Our orders will be reduced by 40 percent next quarter."

What followed was de-risking.

Who is at risk?

TSMC?

How dare the Chinese? This thought flashed through Chen Zhiming's mind.

TSMC is also divided into two factions: those from Taiwan and those from other provinces. The former want to strictly abide by the White House's regulations and not to do any risky operations. They want to refuse any orders from mainland China that have even the slightest risk.

The latter is the exact opposite. As long as we can skirt the rules, we will do it, using the slogan that "it's stupid not to make money" to provide certain support for China's semiconductor industry chain.

Chen Zhiming, who was assigned to be in charge of the Greater China business, was of course the latter, but even he never thought that China could really catch up and use the same slogan as foreign customers to de-risk TSMC.

“Brother Wenjie, the extent of your company’s order reduction this time is not a simple inventory adjustment; it’s a fundamental attack on our long-term cooperative relationship,” Chen Zhiming warned.

At the same time, he felt a vague sense of anxiety.

Because it's too fast.

He was so excited that he even wondered if China had already mastered the 5nm process.

Chip manufacturing doesn't appear out of thin air; it requires a process from testing to production, and then gradually becoming proficient in the manufacturing process, a gradual climb up the ladder.

The general process is as follows: first, you need to conduct technical verification and trial production to ensure that your process is feasible and to resolve any core defects that are found.

Even the slightest deviation in equipment parameters, chemical ratios, or temperature control can cause chip failure.

These issues need to be addressed one by one.

Then comes yield improvement and process flow stabilization. In this stage, the chip yield will climb from 5% to 50%. This is stage two.

Phase three involves ramping up production capacity and optimizing costs. Full-speed production will only begin in phase three because by this point, your yield rate will have reached over 70% and stabilized.

At this stage, wafer fabs began to introduce automation systems to reduce human error and shifted their focus to cost control.

Management will identify equipment bottlenecks on the production line. For example, if the processing speed of a certain testing device cannot keep up with that of the lithography machine, then new equipment must be purchased or the process optimized.

The yield rate in Phase 3 will increase from 70% to 85%.

Finally, in stage four, the mature stage, the process has matured, and wafer fabs begin to focus on production rather than research and development, with the technology being invested in the next generation of processes.

Whether it's Chen Zhiming or other semiconductor professionals in East Asia, everyone thought that China's 7nm process wafer fabs were still in Phase Two, only capable of production.

Neither the yield rate nor the stability can be guaranteed.

Now it seems that things aren't quite like that.

Lin Wenjie, on the other end of the video call, leaned back and stared at the screen, saying quietly, "No, it's not about cutting off the source of the problem, it's about going with the trend."

Just like TSMC cut off supplies to Huawei in the past, we are simply following the trend.

After listening, Chen Zhiming knew he had to do something and tried to gather more information from the conversation: "Brother Wenjie, we understand your concern about supply chain resilience."

But you have to admit that in terms of yield, technological ecosystem, delivery stability, and cost, no mainland Chinese wafer fab can compare with us right now.

"With 40% of the orders transferred there, can you guarantee that the yield rate of your next-generation AI chips won't plummet?"

He needs verification.

To verify their judgment, they need to determine how many 7nm wafer fabs China actually has, and whether it's in Phase Two or Phase Three.

It couldn't possibly be Phase Four, could it? A dangerous thought crept into Chen Zhiming's mind.

In fact, he already had a guess in his heart, which was stage four.

It's very simple. In China's semiconductor landscape, Lin Wenjie's company is not small and has a lot of capital backing, but it's not important, and it's not important to China as a whole.

There are too many similar companies. China is nurturing a breeding ground for trouble. As long as one of them manages to escape, that's all that matters. As for who it is, China doesn't care.

From the perspective of semiconductor professionals in East Asia outside of mainland China, the most important semiconductor companies in China are Huawei, SMIC, and Moore's Threads, which bears Lin Ran's name.

If even Lin Wenjie's company can be allocated 60-40% of TSMC's order capacity, it means that China's 7nm capacity is sufficient.

The supply is so abundant that semiconductor companies in the Yangtze River Delta, Pearl River Delta, and Beijing-Tianjin-Hebei regions of mainland China can all benefit from it.

Lin Wenjie sighed, his tone softening slightly, with a hint of apology towards his old friend: "Of course I can't guarantee that."

I even know that this batch of transferred orders will initially result in at least a 15% loss in yield, and that our costs will not be more advantageous than yours in the short term.

But Brother Zhiming, you must see things from my perspective.

He spread his hands and emphasized, "For our H company, the biggest risk is no longer chip yield, but supply disruption."

If one day, the external chains tighten, all our product lines will be stuck, just like Moore's Thread once suffered. The S90 could have been produced, but one day, an update on the White House website caused their S90, which they had designed at great expense, to be destroyed.

This is not what we want, and it is not what we can accept.

This is especially true today, with domestic 7nm wafer fabs already in operation.

Now, Chinese wafer fabs are telling us that they can guarantee an uninterrupted supply, even if the yield is a little lower or the price is a little higher, they can guarantee 100% security.

What choices do you think my boss, the board of directors, and our shareholders will make?

Chen Zhiming remained silent. He had heard Lin Wenjie's long speech, but he paid no attention to it, as if it didn't exist at all.

Only one thought echoed in his mind: it would result in at least a 15% loss in yield initially.

These words were like a spell, instantly easing his anxiety.

That is, stage three.

Phase Three without a doubt.

China's 7nm process has completed its production ramp-up.

Worse still, China's progress is too rapid, and the Taiwan stock market will be harvested even more efficiently.

Any technological breakthrough from the mainland will be exploited to the extreme by the new MacArthur from the Japanese north.

The other party seems to be born to do this, able to seize the best opportunities and achieve the sharpest harvest with the least amount of capital.

On the platform, more than one tweet has expressed suspicion that China and America are colluding to exploit them.

Even those who claim to be part of the Taiwanese community in Taiwan have begun to doubt America.

You should know that this was a topic that couldn't be touched in the past.

Looking at Chen Zhiming's stiff and unsightly face, Lin Wenjie chuckled and said, "Of course, it's not that we can't talk about it. If TSMC can convince the White House to do 3nm manufacturing for us, I think we'll have plenty of opportunities to cooperate."

We can even discuss order transfers for 7nm processes.

Chen Zhiming's heart, which had just been pounding, immediately jumped again: "Brother Wenjie, why is it 3 nanometers instead of 5 nanometers?"

He didn't even have time to nitpick Lin Wenjie's grammatical errors. What did he mean by TSMC persuading the White House? Could such a direct confrontation with higher authorities really achieve that?

Lin Wenjie explained: "Because the difference between 7nm and 5nm is not significant, it cannot be used as a decisive bargaining chip or a reason for us not to support mainland companies and continue to give orders to TSMC."

You need to persuade the White House, and we need to persuade Beijing just the same.

Writing a report is no easy task.

"Thank goodness, thank goodness it's not like you've already broken through to 5nm," Chen Zhiming thought to himself. This was the only piece of good news that wasn't exactly good news. Chen Zhiming then smiled wryly and said, "TSMC convinced the White House? That's the funniest joke I've heard recently."

The two looked at each other speechlessly.

Even though they are both speaking Mandarin, the other person is being swayed by someone speaking English from thousands of miles away.

The following day, TSMC executives from around the world returned to the Hsinchu headquarters by plane.

The meeting was held late at night, and the crisis response room next to the president's office was brightly lit.

Several key charts were displayed side by side on the giant screen:

Huaguo Wafer Foundry's capacity ramp-up curve: The 7nm yield curve has shown an unusually steep increase in the past six months, having crossed the 50% profitability line and is approaching the 75% stability line.

TSMC's revenue forecast for the next two years: The curve has been drastically adjusted downwards, especially at the 7nm and 10nm nodes, with expected revenue losses of billions of dollars.

Both of these are speculations that were hastily put together by TSMC's strategic analysis department.

On the geopolitical risk map, the line connecting Taipei, Washington, and Beijing is flashing red.

Seated in the room were several of the company's top executives: President Wei Zhejia, several co-CEOs, and senior vice presidents in charge of R&D and business.

"The intelligence that Zhiming brought back has given us a clearer understanding of the situation." Wei Zhejia broke the silence: "Lin Wenjie's words have confirmed our worst-case scenario: it will result in at least a 15% loss in yield in the initial stage."

This means that China's domestically developed 7nm process has completed the third phase of production ramp-up.

“We’re not running data in the lab, nor are we doing small-batch trial production,” he tapped the table with his finger, emphasizing his words, “but have already entered the early stages of commercialization with large-scale mass production, and are using the real money from major customers to refine and improve yield rates.”

At this moment, in this present moment, SMIC's identity has changed from a follower to a competitor.

SMIC, backed by the entire Chinese mainland, has already become a threat to us.

Wei Zhejia was deeply exhausted, a weariness stemming from the White House's strangulation.

In the past, China's pursuit did not put much pressure on TSMC. They had not even mastered 28nm themselves. After obtaining 28nm from Japan, it was not a big problem, because TSMC's profit mainly came from advanced processes.

That is, 7nm and below account for 70% of their revenue, while 28nm process accounts for only a negligible 7% of their revenue.

But 7nm is different; it's a threshold, with 7nm accounting for 17% of revenue.

What worries Wei Zhejia even more is the Esserred Capital representative who is constantly causing trouble at the board meeting and cooperating with the White House to urge them to expand their investment in America.

They not only demanded that TSMC increase its investment in America, but also that TSMC send its engineers and their families to Phoenix, Arizona, promising to grant them green cards in America.

They also made even more outrageous demands, namely requiring TSMC to conduct so-called joint research and development with its old rival Samsung and new rival Rapidus.

It's called joint R&D, but it's actually technology transfer. Neither of these companies has any technological bottlenecks; they are both companies in the independent camp. What they need to transfer is the process, TSMC's mature process technology in advanced manufacturing.

The latter, Rapidus, which was only established in Japan in 2022, was involved in a scandal that instructed TSMC employees to secretly take photos of internal process flows.

Finally, there was the coordinated effort from Wall Street and Silicon Valley to continuously short and long TSMC on the Taiwan stock market.

These factors combined to form America's systematic strangulation of TSMC.

This was a situation that Wei Zhejia had never imagined since taking over TSMC from Morris Chang.

Can the US really be so shameless as to use such tactics against a single company? Can China really put this much pressure on the US, forcing them to do everything in their power to secure their competitive advantage in the semiconductor industry?

These facts, one after another, have now provided Wei Zhejia with an answer.

The vice president of R&D, a technical genius who had worked at TSMC for thirty years, appeared extremely angry at this moment.

"This is a betrayal of engineering ethics!" he exclaimed. "It took us five whole years, with the world's top talent and more than $20 billion in capital, to complete the 7nm ramp-up. China, on the other hand, jumped from the 28nm foundation to this stage in just one year! I seriously suspect that there is a problem within us. I think we need to thoroughly investigate the loyalty of our employees."

His tone grew increasingly lower: "We need to investigate even all second and third generations of people from other provinces."

One of the co-CEOs countered, "Oh? According to our intelligence, China is not using ASML's lithography machines, nor photoresists from Sumitomo Chemical or Shin-Etsu Chemical, nor Zeiss's lens systems."

This is the first time I've learned that these suppliers have already completed their own research and development for our 7nm process.

"President Chen has made such outstanding contributions to the company, I must commend you at the board meeting!"

“That’s not what I meant. What I meant was that our processes and equipment are different, but the technical routes are the same. Our suppliers are different, but the engineering principles underlying the processes are the same. Just like when Liang Mengsong left Samsung, he quickly helped Samsung get 14nm working, which caused us to lose $10 billion.”

"The problem behind this is the manufacturing process," the VP of R&D countered.

The co-CEO raised his voice: "Oh? Then we will allow engineers from Japan and Korea to come to our production line, watch our production, attend our production discussions, and review some of our production files."

Is this a leak? Shouldn't we investigate thoroughly? I see the local engineers involved are incredibly enthusiastic about this work; they even learned to bow 90 degrees to the Japanese engineers! They weren't even that enthusiastic towards me.

Seeing that the two were about to continue arguing about whether they were from the same province or another, loyalty or betrayal, Wei Zhejia slammed his fist on the table: "Alright, that's getting off-topic."

What we need to discuss now is how to deal with it, not looking for enemies within.

The Senior Vice President of Business Development quickly stepped in to change the subject: "Company H only shifted 40% of its spending this time, but it's a bellwether."

Next, other local design companies in mainland China will follow suit.

In the near future, we will face a double squeeze: the high-end market will be overtaken by Samsung and Rapidus, while the mid-range market will be swallowed up by Chinese domestic manufacturers. Worst of all, this loss is irreversible.

Wei Zhejia looked at everyone and calmed down: "Now, we must pay attention to the two bargaining chips that Lin Wenjie has put forward."

He pointed to the text record on the screen:
"If TSMC can convince the White House to manufacture our 3nm chips..."

"Because the difference between 7nm and 5nm is not significant, it cannot be used as a decisive factor..."

After an hour-long discussion.

Ultimately, they reached an agreement: on the one hand, to increase investment and accelerate the research and development of advanced processes; on the other hand, to agree to expand investment in America and move one of the most advanced 2nm process production lines to America, but in exchange for permission from the White House to sell 5nm to mainland China.

"Engineer Liang, please have a seat," Lin Ran said to Liang Mengsong, pointing to the seat in front of him.

Sitting opposite me is a true craftsman, a man with the soul of an engineer. If he were Japanese, calling him a chip genius would not be an exaggeration.

Since joining SMIC as co-CEO in 2017, under his leadership, SMIC has achieved an astonishing leap in process development speed. Even without Lin Ran, SMIC has achieved the capability to manufacture 7nm chips without advanced equipment such as EUV, and has produced small batches for specific products.

Most importantly, he helped SMIC build a complete semiconductor manufacturing and R&D system.

Of course, technology is one aspect, but Lin Ran admires the other party's character even more. The other party donated 90 yuan to establish the Meng Ning Scholarship at Hangzhou Dianzi University after only two years of joining SMIC.

Their styles are very different from other 4V semiconductor elites who came to the mainland.

"President Lin, our CG-1 lithography machine is still in the process of technical verification. Theoretically speaking, it is already fine, but its stability is too poor."

The meta-lens system is indeed very advanced and highly precise, but it also has its problems.

That means we can't easily adjust it. This lens system is too precise; any tiny change in temperature, vibration, or particles will cause optical distortion.

Once the distortion occurs, we have to stop production and spend weeks re-aligning and correcting.

The cost of a single calibration is even higher than producing a 7nm wafer for a month.

"Compared to traditional semiconductor manufacturing equipment, the biggest problem with this technology route is that its debugging costs are too high, both in terms of time and money," Liang Mengsong explained.

Just as TSMC was trying to obtain permission from the White House to sell 5nm chips to China, China had already reached the threshold of 4nm.

This lithography machine, named CG-1, is not yet an EUV lithography machine because China has not yet mastered the light source, but it represents the limit of the traditional DUV lithography route.

"Engineer Liang, you've made an absolutely unavoidable engineering mistake," Lin Ran gently reminded him.

Liang Mengsong was taken aback and looked at Lin Ran with a puzzled expression.

In the past collaborations, Liang Mengsong knew that Lin Ran might not be as proficient in the craft as he was, but Lin Ran's imaginative ideas, especially his ability to solve engineering problems mathematically, were absolutely unparalleled on Earth.

Their team jokingly says that if Lin Ran can't solve a math problem, then no one on Earth can.

"We achieved a theoretical resolution of 4nm using the extreme precision of meta-lens."

But you try to tame it using traditional static passive stabilization methods, whether it's isolating vibrations or using constant temperature and humidity.

We spend countless hours fine-tuning and pursuing an ideal, unchanging, perfect focal point.

Lin Ran raised his hand and pointed into the air, as if there was a CG-1 lithography machine right in front of them.

"But the lens group of CG-1 can never be as stable as ASML's pure quartz lens."

Its materials and structure make it inherently more sensitive to the environment.

Both now and in the long term, we must abandon the illusion of absolute stability and turn to dynamic compensation.

　Give me some time, let's slowly climb the ladder. I updated 6 words today. I was planning to write another chapter tonight, but something came up and I had to go out.

　　Last day of double monthly votes! Please give me a monthly vote, waaaaah!

　　
　
(End of this chapter)