Apple, AMD, and Nvidia are vying for TSMC chip orders, and TSMC is the reason for the dominance.
On January 30, IT Home reported: Apple, AMD, and Nvidia are competing fiercely in the field of artificial intelligence, and they placed orders with TSMC some time ago, and related chips will be produced after April.
It seems that TSMC's performance in 2023 is stable again.
After losing its major customer Huawei, TSMC did not fall, but marched all the way, becoming more and more courageous. Not only did the revenue reach a new high, but the market share also further increased.
According to relevant data, TSMC's revenue in 2022 will be 50258.4 billion yuan, a year-on-year increase of 426%, and the market share reached 60%, reaching a record high.
So the question is: how did TSMC grow into a leader in the field of chip foundry?Why are IC design companies more willing to place orders with TSMC?
How TSMC became a leader in the chip foundry industry.
Founded by Zhang Zhongmou in Hsinchu, Taiwan in 1987, TSMC is the world's first professional chip foundry and the company with the highest level of chip manufacturing.
Today, TSMC is a global leader in chip manufacturing, which means that if TSMC sneezes, the entire chip industry will catch a cold.
TSMC's current achievements are inseparable from its own efforts, but it is also inseparable from Intel, ASML and many other companies.
TSMC's success is first and foremost due to its founder, Zhang Zhongmou.
Zhang Zhongmou is a typical academic talent, he studied at Harvard, MIT and Stanford, three top universities, ordinary people can be admitted to one of them, even if they are the best among the people.
After a year at Harvard, Zhang transferred to the Massachusetts Institute of Technology (MIT) and became a student in the MIT Department of Mechanical Engineering. Because he felt that Harvard did not have any advantages in terms of employment.
In 1955, Zhang Zhongmou graduated from the Massachusetts Institute of Technology and joined a semiconductor company called Sylvania.
After entering the company, Zhang Zhongmou began to gnaw"Knowledge of semiconductors"It took three years to finally put it"Knowledge of semiconductors"Gnawed over.
In 1958, Zhang Zhongmou left Sylvania and joined Texas Instruments, the largest semiconductor company in the United States at the time. At the same time, he was joined by an employee named Jackkilby.
Jack Kilby is"Integrated circuits"and was awarded the Nobel Prize in Physics in 2000.
Zhang Zhongmou often drank coffee and chatted with Jack Kilby, and witnessed the birth of integrated circuits.
During his time at Stanford University, Zhang's career became clear, and eight years later he was promoted to the third director of Texas Instruments.
In 1972, Texas Instruments began to transform, the semiconductor business was gradually abandoned, and Zhang Zhongmou was naturally let go.
In 1983, Zhang Zhongmou was forced to leave Texas Instruments.
Creation of TSMC.
After leaving Texas Instruments, Zhang Zhongmou lived a life of idle clouds and wild cranes, and soon he received an invitation from Sun Yunxuan to serve as the president of Taiwan's Hsinchu Industrial Technology Research Institute.
Zhang Zhongmou gladly accepted the invitation.
In 1987, ITRI signed a contract with Philips to establish TSMC, a company specializing in chip manufacturing, with Zhang Zhongmou as the chairman of the new company.
Zhang Zhongmou decided that TSMC would only do one thing, that is, chip foundry, and not other businesses.
This model is very different from that of Intel, Samsung, Intel, and Samsung, which are typical IDM-type companies, i.e., companies that specialize in chip design, manufacturing, packaging, and sales.
Since Intel was the strongest chip company at that time, it looked down on TSMC, a professional chip foundry, and refused to invest in shares.
Certified by Intel.
At the beginning of TSMC's establishment, it was struggling. Poor technology, few talents, no orders at all.
Zhang Zhongmou had no choice but to find Grove, the former CEO of Intel. finally tricked Intel's technical team into guiding her.
The team came to TSMC and pointed out more than 100 problems on the spot and hinted that TSMC should know what to do.
Zhang Zhongmou had a smile on his face, but he was very uneasy in his heart. As a result, he began to govern with an iron fist, and finally rectified all the more than 100 problems as required.
After the rectification, TSMC passed the audit of Intel and obtained Intel's professional certification. With Big Brother certification, orders don't have to worry about it. TSMC has also begun to develop slowly.
Merged with Shida. In 1997, Zhang Rujing also resigned from Texas Instruments and began to embark on the road of entrepreneurship.
During his time at Texas Instruments, Zhang Rujing built ten semiconductor factories in one go, hence the name"Factory King"。These experiences led Zhang Rujing to decide to create a professional chip foundry.
Soon, Shida Semiconductor was born and became a strong competitor to TSMC in just three years.
Don't let anyone snore around the couch", is Zhang Zhongmou's consistent strategy to deal with competitors, so he used his connections to negotiate and finally acquired the world's largest semiconductor company for $5 billion.
Zhang Rujing didn't know about it in advance, but when he heard the news, he angrily approached Zhang Zhongmou and negotiated with him. The two parties reached an agreement to build a state-of-the-art biscuit factory on the mainland.
However, this agreement was never realized, and Zhang Rujing had no choice but to leave TSMC and start a new business, which led to SMIC some time ago.
130 nm against UMC.
In the 180nm process, TSMC is close behind UMC. But at 130 nm, TSMC is back!
UMC's chip manufacturing business was helped by IBM and Infineon, but the development progress was much lower than expected.
TSMC found the right partner, and with its own efforts, it developed a 130-nanometer chip ahead of schedule, surpassing UMC.
Work closely with ASML.
In 2002, the leading lithography machine at that time was still Nikon in Japan, and ASML was just a second-rate company.
In the process of developing the 193-nanometer lithography machine, Nikon encountered many problems that cost billions of dollars, but still failed to overcome them. At the time, TSMC's Benjaminlin came up with a solution.
Using a liquid instead of air as a medium, the wavelength of the light source is reduced by refracting the liquid, and finally a more advanced chip is made. "
But Nikon, which is working on dry lithography, rejected Hayashi's proposal. The shrewd ASML felt that this was a once-in-a-lifetime opportunity and contacted TSMC for cooperation.
A year later, the world's first immersion lithography machine was developed, surpassing Nikon in manufacturing process, and ASML was a huge success.
Subsequently, ASML pursued the victory and successively developed 157nm and 132nm lithography machines. TSMC also used ASML's lithography machine to upgrade its chip process to 32 nanometers, advancing Moore's Law by three generations.
Soon after, ASML began to develop EUV lithography machines, which were more difficult from a technical and capital-intensive point of view. TSMC once again invested in financial support and acquired part of ASML's capital, becoming a shareholder of ASML.
In 2009, the EUV lithography machine was successfully developed, which means that chips with 7nm and below processes can be manufactured.
TSMC became ASML's most valuable partner and shareholder and, as a matter of course, took the lion's share of the EUV lithography machine.
With a large number of EUV lithography machines, TSMC has conquered 7nm, 5nm, 4nm and 3nm processes, far behind Samsung in terms of yield rate, and has become a veritable chip foundry leader.
TSMC's chip manufacturing process is the best.
In the entire chip foundry field, TSMC's level is undoubtedly the best, leaving competitors far behind.
At present, the world's top ten chip foundry companies are TSMC, Samsung, Goldfront, UMC, SMIC, AutoNavi, Hua Hong Semiconductor, VIS and Dongfang Hi-Tech.
These ten companies are second only to TSMC and Samsung in terms of mass production of 7-nanometer chips, but Samsung's processing level is much lower than TSMC.
Built by Samsung"Fire Dragon", built by TSMC"Ice unicorn"。
Qualcomm has already teased because of the heating up of mobile phone processors"Fire Dragon", but this fire dragon is basically built by Samsung.
The famous Snapdragon 888 is jokingly called by netizens"Fire Dragon 888"。
On New Year's Day 2021, Xiaomi debuted the Snapdragon 888, which made rice fans cheer and scream, but after the phone was in their hands, it encountered a serious problem --"Hot hands"。Especially when playing games, it feels like holding a hand warmer.
After testing, the power consumption of the Snapdragon 888 CPU reached 78W, the single-core power consumption is 1W higher than that of the Snapdragon 865, and the multi-core power consumption is more than 19w。As the power consumption increases, so does the natural heat generation.
Some netizens said that after running "deusoriginal" with Xiaomi 11 for 20 minutes, the surface temperature of the mobile phone reached 48 degrees Celsius. What is this concept?8 degrees Celsius warmer than bath water!
The reason for this is that Samsung's manufacturing process is too procrastinating, with a low yield rate and a high leakage rate, and in the end, Snapdragon becomes a fire dragon.
In addition to the Snapdragon 888, the Snapdragon 821 and Snapdragon 8Gen1 also have serious heating phenomena, so it's no wonder that Qualcomm will transfer some orders to TSMC.
Take the Kirin chip created by TSMC and Huawei as an example, it is called by industry insiders because of its excellent heat dissipation ability"Ice unicorn"。
In August 2018, Huawei launched the Kirin 980 mobile phone chip, which is equipped on the Mate20.
This chip uses TSMC's 7nm process, which is the industry's first 7nm SOC. The CPU is designed with 8 cores, and the maximum frequency is 26GHz, the overall performance is 20% higher than the previous generation.
The biggest highlight of the Kirin 980 is the heat dissipation, even if you play large online games, it is not easy to heat up, so it is called by netizens"Ice unicorn"。
The reason why the Kirin 980 has excellent heat dissipation is not only Huawei's design skills, but also TSMC's excellent manufacturing level.
Comparing the manufacturing levels of TSMC and Samsung, if you were the CEO of Apple, AMD or NVIDIA, who would you place an order with?
TSMC 5nm vs Samsung 5nm.
TSMC and Samsung have achieved mass production of 5nm chips, although they are both 5nm, but they are actually very different.
Transistor densities are different.
TSMC 5nmn5 has a transistor density of 1731mtr mm (million transistors per square millimeter), specifically logic circuits.
Samsung's 5NMLPE process has a transistor density of 1265MTR mm, which is a difference of 470,000 transistors per square millimeter compared to TSMC.
The transistor is the most basic unit of a chip, and naturally the more the better.
Because the higher the number of transistors, the faster the clock speed, the richer the features, the wider the bus, the larger the cache, and the more cores the chip has.
Overall, Samsung's 5nm process is"Discounted", resulting in chip performance lagging behind TSMC.
The key values are different.
TSMC and Samsung are not quite the same in terms of gate pitch, metal pitch, and these key values.
TSMC's 7nm process has a gate spacing of 57nm, a 5nm process shortened to 50nm, and a metal spacing that has been reduced from 40nm to 30nm.
Judging from Samsung's data, the 5nm process is basically the same as the 7nm process, no wonder TSMC will say that Samsung's 5nm is not much different from its own 7nm.
FinFET process, TSMC has more advantages.
Both TSMC and Samsung's 5nm chips use the FinFET process, and TSMC has an advantage in this process.
FinFET is a new type of gold-oxygen complementary semiconductor transistor. It is called a fin-FET because it is shaped like a fish's fin.
The FinFET is an innovative technology that designs an otherwise flat transistor into a three-dimensional shape with the gate surrounding the channel.
Simply put, finFETs have better channel control, faster switching speeds, higher drain currents, lower switching voltages, and lower power consumption.
Therefore, despite the high cost, chipmakers have opted for this superior transistor.
FinFET is developed by C. from the University of California, Berkeleym.Hu proposed, he is also called"Father of Finfet"。But you know cm.Was HU used to be the CTO of TSMC?
1999, cm.HU proposed FinFET technology, 2001, cm.It's no coincidence that HU became TSMC's CTO, right?
For 4 years, Hu Zhongmou was responsible for TSMC's FinFET technology, which is one of the reasons why TSMC's chip technology is stronger.
If you were in charge of Apple, AMD or Nvidia, would you give the order to TSMC or Samsung?I believe the answer is already there!
Apple, AMD, and Nvidia are the three major customers of TSMC's orders, and there is another reason, that is, these three companies are all American companies, and TSMC has also built a 5nm factory in Arizona, USA.
This move may also hint that Samsung will build a factory in the United States if it wants to get more orders.
Thinking of this, I am worried about the development of domestic chips again, when will I be able to build my own 5nm chips?3 years, 5 years, or 10 years?