The Millennium Semiconductor Survival Guide

“Before the emergence of Xinxin, there were only three players in the chip foundry field, TSMC, Intel and Samsung. Because Intel and Samsung both have products for individual consumers, among these three companies, only TSMC’s customers come entirely from external companies.

Previously, our practitioners did not think there was any problem with this. Because of the highly professional nature of chip foundry and chip design, IDM was limited by risks and operating costs and was unable to compete with the Foundry model.

The risk of the IDM model is that you need to do chip design and manufacturing at the same time. Once the sales are not as expected, you will face high losses. Then you need to support two types of talents at the same time, which is also a cost.

Foundry's model only requires you to do a good job in chip design. While reducing costs, you can also adjust production strategies based on sales expectations. The cost of production is borne by the foundry.

From the beginning of the 1990s to the new millennium, TSMC's Foundry model was welcomed by the industry because of its low cost. The industry gradually believed that this model would be the main development direction of chips in the future.

So much so that Intel, the industry giant, hands over all non-critical wafers to TSMC, and it is only responsible for the production of core wafers. Xinxin has once again brought the industry's attention back to the IDM model. IDM has to bear more risks. Likewise, once your products can be sold, your profits will also increase a lot.

This change in concept has led IDM-model factories like Intel and Samsung to choose to hand over wafer manufacturing to their own factories, while small and medium-sized chip design companies are more willing to cooperate with new chips with low unit costs. TSMC’s market share continues to increase Being cannibalized.

TSMC not only faces strong competition from Xinxin on the revenue side, but its talent attraction is also getting worse year by year. In the past, TSMC was able to attract senior talents from China Semiconductor, but now their first goal is new chips, followed by TSMC."

This blog is a relatively in-depth analysis of the problems faced by TSMC. Xinxin has hit the chip industry in the Wanwan region in an all-round way.

TSMC originally walked on two legs, Intel and Apple, but now Intel has been poached by Xinxin. A large part of TSMC's orders are either made by Intel itself or for Xinxin. Not to mention Apple, which has nothing to do with TSMC.

It is almost impossible for TSMC to take advantage of the mobile Internet to dominate the chip foundry field.

As for the issue mentioned at the end of the blog, TSMC’s declining attraction to talents is crucial.

After Zhang Zhongmou retired and Liu Deyin took over as CEO of TSMC, they relied on their strong position in the chip foundry field to attract talents from all over Asia except Neon and Korea.

Like Zhang Xiaoqiang, who graduated from Tsinghua University. The Tsinghua University in Yanjing is not the one in Hsinchu. He has a PhD in electrical engineering from Duke University. After leaving Intel, he joined TSMC and became the vice president of the technology department. After Jin Pingzhong retired, Zhang Xiaoqiang took over the responsibility of the business department. person's position.

There is also Huang Hansen, who has a bachelor's degree from the University of Hong Kong, a Ph.D. from Lehigh University, and a professor in the Department of Electrical Engineering at Stanford. In 2013, he built the world's first computer assembled from carbon nanotubes and appeared on the cover of Nature magazine. After joining TSMC, he also became a technology Vice President of Department.

After the rise of Xinxin, talents like Zhang Xiaoqiang and Huang Hansen will be attracted to Xinxin first, not TSMC.

“During the general election, we like to use the golden cross to describe the fact that the support rate of the candidate who was originally lagging has overtaken that of the leader, and the gap will become wider and wider after this overtake is achieved.

Recently, with the completion of Xinxin's acquisition of ASPLA, a large number of chip industry practitioners and financial media on the island have used this word to describe the golden cross between Xinxin and TSMC. Mr. Liu, what do you think of this? "

In order to stabilize the morale of the military, Liu Deyin personally came out to support TSMC and participated in a business interview program on Zhongtian.

It is also interesting to say that Liu Deyin expressed his views on Wutong in an exclusive interview with the New York Times in the future. The headline of the island media was: "Liu Deyin said that if Wutong happens, the world order will collapse."

The headline of the mainland media is: "TSMC Chairman personally refuted Silicon Shield's statement."

As for the specific New York Times interview and Liu Deyin’s true views, the media doesn’t want you to know.

To put it bluntly, Liu Deyin is the old blue man's point of view. It is best to maintain the status quo and not delay TSMC's making money.

Now Xinxin is putting pressure on him from all directions, making it impossible for TSMC to resist. The financial crisis has made it very difficult for TSMC to raise funds at this time, and it needs to bear huge capital costs.

“I don’t quite agree with the term golden cross, because it is not a zero-sum game between us and Xinxin, there is competition and cooperation between us at the same time.

The wafer foundry market is huge and cannot be completed by one manufacturer. It can accommodate us, Xinxin, Intel, Texas Instruments, etc. at the same time.

The golden cross describes only two competitors in the market, and then it is an absolute zero-sum game. In the end, the winner takes all, so I don't agree with this statement.

Not to mention Xinxin and Intel, they are of the same type, and we and SMIC are of the same type. "Liu Deyin needs to refute this kind of remarks.

Because comparing TSMC and Xinxin together, TSMC's stock price in the U.S. stock market will not be able to turn around in a short time, and may even fall.

He knows very well that in the face of Xinxin, TSMC has no advantage at all. If investors benchmark TSMC and Xinxin, no one will choose TSMC.

For the primary market, we can just invest in Xinxin's newly built factory. Basically, we will give you a fixed dividend of about 7% every year. Isn't it great?

Those investors in the secondary market who believe in value investing don’t even want to invest in TSMC. Not being a leader means they won’t enjoy the added value.

Therefore, Liu Deyin wants to reshape the market's view of TSMC and Xinxin. TSMC's competitors are companies such as SMIC, Huahong, and Shenhai Belling. We are a competitor on the track.

As long as your competitors change quickly enough, you will always be number one.

"Oh, this is an interesting statement, so the market view is wrong. Xinxin and Intel are in direct competition. Xinxin's biggest competitor is Intel?" the host asked curiously.

To put it bluntly, he is also here to cooperate with Liu Deyin in creating this image, so he only needs to be a good supporter.

Liu Deyin nodded sharply: "Yes, before the birth of the new core, Intel was the only company that survived to this day with the IDM model, which is what we call design and manufacturing.

At the beginning of its birth, Xinxin's goal was to be a chip foundry. It was a chip manufacturing company established to compete with TSMC. However, as Mphone became popular around the world, Xinxin's main business goal became to help Mphone design and produce chips. .

Later, with the launch of Xinxin's own brand Xiaomi and Bluetooth chips, this concept was further strengthened. Xinxin cannot even meet its own internal chip supply, let alone receive orders from outside companies.

On the contrary, part of the demand for Xinxin's internal chips will be given to TSMC. "

What Liu Deyin said is true, but he hid part of the information, that is, Xinxin's orders to TSMC are getting fewer and fewer. As Xinxin's production capacity increases, sooner or later they will grab TSMC's orders.

Also, Xinxin’s priority cooperation partners are Huahong and SMIC, followed by TSMC.

The host suddenly realized: "It turns out that there is still a cooperative relationship between Xinxin and TSMC, which means that the better Xinxin develops, TSMC can also benefit from it."

Liu Deyin wanted to refute in her heart, but on the surface she still had to pretend to agree: "Yes.

Models like New Core and TSMC are equivalent to a single company with fewer restrictions on data sharing, so Intel can optimize each other's manufacturing and design more closely than AMD.

Their final chips can be priced lower than AMD's because there are no additional foundry fees to pay. "

Most of AMD's foundry is done by TSMC.

“Similarly, Xinxin is also a company that can take full advantage of data sharing to optimize design solutions and manufacturing processes.

Precisely because Xinxin and Intel are the same type of companies, Xinxin made Intel very uncomfortable in the field of mobile chips. Intel's leading advantage over AMD disappeared when facing Xinxin. "Liu Deyin wants to drag Intel down.

His words will be spread to Silicon Valley by the media. The connection between TSMC and Silicon Valley is much closer than the connection with Zhang Jiang.

Intel started to lag behind in the 14nm process. It was originally scheduled to break through the 10nm process in 2016, but the result was a complete collapse. TSMC took the lead in 16nm, 10nm and 7nm processes, directly leading Intel by three years.

After Intel lost its leading position, the problem of inefficiency was exposed. Their yield rate of manufacturing wafers was significantly lower than that of TSMC, and more fabs were needed to achieve the same output.

From cost to performance to yield rate, it lags behind in almost all aspects. No one can even explain why Intel lags behind.

So much so that when Intel itself carried out reforms later, it adopted an internal foundry model, and the internal design department had to pay to lock in orders.

Use a large-scale internal foundry model to save costs. For example, different chip design teams, how much was the sales of the chips your team designed last year, and then Intel will allocate the money you have at your disposal based on your team's sales. Take this money to lock in the order.

Instead of the previous situation where all teams were lined up according to their importance, Team A took up all the production of advanced processes. As a result, the things they created could not be sold at all depending on the defense. They could just sit on their laurels and finish taking up all the advanced processes.

And because Zhou Xin knew the shortcomings of the IDM model, Xinxin adopted an in-house foundry model from the beginning.

Even because of the popularity of Yanque and Honghu, Xinxin has internal chip designers to form a separate team to design smartphone-related chips.

"Then who do you think will have the advantage between Intel and Xinxin?" The host also watched the excitement and didn't take it too seriously.

Liu Deyin said: “Intel controls the present, and Xinxin seems to have an advantage in the future.

We all know that personal PCs have been the mainstream electronic product in the past few decades. Intel firmly occupies the monopoly of personal PC chips, but Xinxin's leading position in smartphone chips is equally stable. Smartphones are the future. "

The host asked: "Mr. Liu, when you were interviewed about ten years ago, you said that the Fabless model is the future and IDM is a backward model. Now with the emergence of Xinxin Technology, the IDM model seems to be revitalized.

Everyone has discovered that in addition to Intel, other companies can also do well in the IDM model. Do you want to overthrow your own views ten years ago or continue to insist that Fabless is better than IDM? "

Liu Deyin said: "Xinxin seems to be an exception. Its strength is based on its strong position in smartphone chips. When other companies cannot occupy the absolute lead in a certain type of chips like Intel and Xinxin, Fabless It’s better than IDM.”

The host laughed: "So you still insist on the view that Fabless is better, just add a patch to this view."

Liu Deyin nodded: "Yes."

In the future, Intel will completely fall behind TSMC after 14nm. After ten years of hard resistance, they can no longer resist and split chip manufacturing and chip design.

Intel-branded foundries can take orders for Apple's A-series chips and M-series chips, and orders from Intel's design department can be given directly to TSMC instead of covering them up with various excuses.

When Intel made the decision to split, it expected that it would save US$3 billion in costs and contribute 6% to profits in the first year after the split.

By 2023, the foundry's revenue is expected to exceed US$20 billion, surpassing Samsung to become the world's second largest foundry. With these cost reductions and efficiency improvements, Intel aims to catch up with TSMC by 2026 and potentially win business from big players like Apple and Nvidia.

The result is that this is just a beautiful fantasy, and the actual situation is far worse than their own expectations. Intel's backward technology is related to bloated organizational relationships and the loss of organizational vitality. It is not necessarily the IDM model that is not working.

"What do you think of Xinxin's acquisition of ASPLA?" the host asked, "This is also a topic of great concern on the island recently. In such a general environment, Samsung is expanding production and Xinxin is making mergers and acquisitions. Is TSMC Are you going to take advantage of this opportunity to expand against the trend?”

"To be honest, although I don't want to admit it, Xinxin's acquisition of ASPLA is not good news for TSMC." Liu Deyin smiled bitterly: "Because this means that Xinxin's technology can be improved again, its production capacity can be increased, and it can be given to TSMC. There are even fewer orders.

Samsung's expansion of production and the expansion of Xinxin both indicate one thing, that is, they have enough confidence in their own companies and believe that they can survive the financial crisis and survive until the day when the situation is reversed and the cycle goes up.

TSMC also has the same confidence, but compared with these two companies, TSMC's financial strength is indeed not that strong. "