Hua Guo does not know that in the future, in the chip field, the division of labor will become more and more detailed, and semiconductor manufacturers will gradually specialize in a certain field.
Originally vertical manufacturers, the only semiconductor manufacturers that handle everything from chip design to chip manufacturing are Samsung and Intel.
Throughout the 1990s, China's benchmarking targets in the semiconductor field were first Neon Enterprises and then Korea Enterprises.
Samsung to be precise.
The emergence of Zhou Xin gave China another option, which is to separate chip design and chip foundry. Design, manufacturing, and other technologies upstream and downstream of the supply chain are all supplied by different companies.
As a big country, China will not give up its original strategic line and cultivate a semiconductor company that takes into account both design and manufacturing. But they will not give up exploring new models, especially Shenhai City. They hope to replicate the success of Silicon Valley in Zhangjiang.
Judging from Zhou Xin's experience, there are technical advantages to handling design and manufacturing from one source. When the process reaches below 22nm, this advantage becomes more and more obvious.
Intel has mastered both chip design and manufacturing processes in the 14nm field. This is an important reason why they can use 14nm to create the most powerful single-core performance CPU.
Vertical manufacturing has technical advantages, but its development efficiency is not as good as the commercial division of labor model of chip foundry and chip design.
Later Intel's chip manufacturing department has always been in a leading position in the process, especially in the 22nm process in 2012, far surpassing TSMC and Samsung.
This vertical manufacturing model gave Intel strong confidence. However, after 22nm, the 14nm process was delayed for half a year. Later, Intel's process delays became the norm, and later chip manufacturing was outsourced to TSMC.
So much so that in the end, it was able to single-handedly contract chip foundry and chip design on the most advanced manufacturing processes. Only Samsung still persisted.
As chip manufacturing reaches the 3nm era, Intel needs TSMC’s foundry capacity support.
A company as strong as Intel, which had all the technical advantages in the first decade of the 21st century, would fall behind due to path dependence on technology research and development. Therefore, Zhou Xin is not very sure about this path.
Intel's later decision to find TSMC for OEM was not a sudden decision, but the result of long-term accumulation.
There are two main situations when Intel looks to TSMC for foundry. One is that the departments it originally acquired, such as Infineon's wireless department, already had some RF circuits using the older TSMC 28nm process. This was unnecessary because the original plan required it. Transfer orders for the sake of transfer orders.
Another situation is the problem caused by Intel itself. Capacity planning is a long-term, interlocking process, and problems in one link will have an impact on the future.
Build a factory, prepare raw materials, train engineers, adjust machines, and wait until the technical department debugs the process formula before starting small-scale trial production. During the trial production process, while adjusting the yield rate, the chip design team started designing one to two years in advance.
Finally, at the time agreed between chip foundry and chip design, the design team hands over the design plan to the wafer fab and begins to slowly build the photomask layer by layer, and finally mass-produce, package, test, and ship.
Under the original plan, Intel's 10nm process research and development went smoothly. While the 14nm factory was mass-producing, the 10nm factory was slowly getting ready for testing. When the yield rate reached the standard, it could start receiving large quantities of orders. The chip design team has basically completed the design of the 10nm process.
For Intel, which chip design department will take priority after the 10nm process is implemented? Graphics processors, because the architecture of graphics processors is consistent and the design cycle is short, design blueprints can be quickly produced. Then CPU and Server began to produce 10nm chips. 10nm production continues to increase, and 14nm is slowly freeing up production capacity.
Among Intel's 100,000 employees, in addition to core departments such as TMG/CPU/Server, there are many peripheral organizations affiliated with Intel that also need chips.
Most of these peripheral organizations are collections bought by Intel's former CEO Brian who squandered his capital. For example, the FPGA company Altera, which was originally relatively famous but has gradually been forgotten, bought the Infineon wireless division in order to return to the mobile phone market, and caught up with the popular artificial intelligence Nervana and self-driving Mobileye, which were taken over by others.
After deducting these large luxury goods, there are still some basic IP, IO, memory, and small CPU Atom. Finally, there are some peripheral organizations that are not large in quantity and can be dispatched in an official car. Or it may be a group that purely does research and publishes papers, and is an advanced chip design organization that helps TMG test chips. In short, there are many hills inside Intel, and there are no shortage of surprises.
What's more, there are some wandering samurai organizations that have been quietly floating in Intel for decades for unknown reasons. When Zhou Xin was interning at Intel, a senior in the industry whom he met referred to these groups as the stepmother's unloved children.
In fact, these wandering warrior organizations were once affiliated with formal organizations. It's just that these formal organizations have been disbanded. Like Intel's spur-of-the-moment wafer foundry, the leaders of the design department had no choice but to work part-time within Intel in order to survive for their employees.
For example, if the Server department wants to make a memory control IC and is short of manpower, it will temporarily let the wandering warriors take over the job. If it faces too much competition, such as when encountering Bangalo, India, it uses a sea of people tactics and likes to boast that everything is fine to take away the job. Or if you meet the Chinese engineers who work hard in Penang, Malaysia, you can only scratch your nose and pick up some jobs that others have left behind and have no future.
Intel also makes chips for ZTE, and these groups that help ZTE make chips are the Wandering Warriors. In short, Intel is such a jungle world.
These various internal organizations decide whether they can move into 10nm or continue to use the excess capacity of 14nm based on their importance to the company's profit contribution.
As a result, Intel's 10nm was difficult to produce, and the chip design team was idle. More importantly, what about CPU/Server/GT? We have no choice but to continue to occupy 14nm production capacity.
The worst situation has come. The unit responsible for process research and development in TMG is TD (technology development). TD is already at the extreme state of overwork. In the past, when 14nm was completed, the main force went to 10nm. Leave a small group of engineers to maintain 14nm, and then separate a small group of elites to work on 7nm. Now we will continue to develop 14nm, and we will also develop 14nm+, 14nm++, and 14nm+++, which is the toothpaste squeeze that Intel is criticized for.
There will be fewer people allocated to 10nm, and even fewer to 7nm. He forced himself to sink deeper and deeper into the situation. As a result, one step is wrong and every step is wrong.
Moreover, the 14nm production capacity has been given to the company’s lifeblood CPU/Server. What about other departments? Life still has to go by, and ICs still have to be shipped. In order to survive, everyone started to take action. The relatively powerful Infineon wireless department jumped out and said, we used TSMC in the Infineon era. We have a happy cooperation. Let us Keep using it.
The AI team said that our AI chip cannot wait. How much data is waiting for us to train, NVIDIA has surpassed us by several blocks. Therefore, we must use the best process. If Intel’s 10nm is not available, we will use TSMC’s 7nm.
Some edge departments that help make peripheral chips for the Server system started to speak loudly, saying that if we don't ship the goods, the Server won't be able to ship the goods either, and everyone will die together.
At this point, Intel has no other choice and opening up orders to TSMC is the last resort. As a listed company, Intel has an obligation to shareholders to make profits, but making profits does not mean that it must produce chips on its own.
As for Intel, after it started placing orders with TSMC, it had serious side effects on its own company structure. Inexplicably, Intel began to use TSMC for CPU production.
Zhou Xin is very familiar with this period of history. Hu Zhengming originally went to TSMC as chief technology officer. Later, TSMC chairman Liu Deyin was just Hu Zhengming's follower in the first decade of the millennium.
To be precise, Zhou Xin is very familiar with TSMC’s history. TSMC’s 7nm process foundry was forcibly launched at the beginning.
In the beginning, it was all a push to compete with Samsung. Judging from the OEM price and the yield rate at that time, even the cost could not be recovered. It was completely dependent on government subsidies to smooth out the losses.
Of course, the advantage is that after getting through the initial stage, relying on market monopoly, TSMC's gross profit margin is close to 50%, much higher than Samsung, which only designs its own chips. Samsung also contracts chip OEMs from Qualcomm and Nvidia on a small scale, but Nvidia and Qualcomm are particularly aggressive in keeping prices down.
TSMC and Samsung are backed by the government, and the government would rather provide high subsidies than let them collapse.
This is even more true for Intel. Even if Intel loses money in its chip production and can achieve higher profits by relying on TSMC and Samsung, it is impossible for them to give up chip foundry.
Because Korea and Wanwan, where Samsung and TSMC are located, are too close to China, if there is a future Cold War, these two regions will be on the front line of confrontation. The risk is too high and Amerikan will lose its chip source.
This is why Americen uses various means to either force or lure TSMC to build a factory in Americen. It’s just that TSMC was fooled and stunned, while Samsung relied on procrastination to delay indefinitely.
From a long-term perspective, Zhou Xin prefers a model in which chip foundry and chip design are separated. Even if it is as powerful as Apple, it will also hand over chip foundry to TSMC.
Intel, the only one that has persisted, has a strong advantage in the entire ecosystem. Most of the software in the server, such as operating systems, compilers, databases, etc., are all developed on the Intel chip ecosystem.
These require a long period of running-in and debugging to reach a stable state.
If you switch to a CPU from another company, in theory it should be directly compatible without any problems, but in practice it will definitely require a lot of debugging to reach production status. So unless other manufacturers' CPUs, such as AMD's, are much better, companies have no incentive to replace Intel's CPUs.
In the future, some countries will replace Intel out of national security considerations, which will expand unlimited prospects for their own IT manufacturers.
Therefore, if Intel's model is to be adopted, Zhou Xin will face two problems. The first is that even if Intel is stronger than Intel, it will not be successful if it takes this path, and Zhou Xin also has no confidence.
Without a system, it is impossible for one person to control all nodes from chip design to chip foundry, let alone climb the upstream lithography machines and high-end materials.
There are too many nodes involved, so the innovative enterprise cluster approach is the most feasible.
The second problem is that no matter how outstanding Zhou Xin is in the semiconductor field, he cannot convince all Chinese companies to develop domestic equipment and domestic operating systems to replace Amerikan's IT software and hardware at this point in time.
Therefore, the government appropriately allocates resources and relies on marketization to cultivate a number of competitive companies in the integrated circuit field. Before 18 years ago, these companies were able to occupy a certain market share in the upstream and downstream of the semiconductor industry chain by relying on their own capabilities.
So after 18 years, under the changes in the macro environment, these companies will rely on the Chinese market to grow rapidly and replace foreign industrial chains.
This is why Zhou Xin values Zhang Jiang very much.
Zhangjiang did not play a planning role in the original time and space, and there was a gap between the planning when the Zhangjiang Hi-Tech Park was established.
Zhou Xin hopes to rely on Xinxin Technology, a model project, and financial advantages to attract talents and industries to create economies of scale.
Xinxin Semiconductor and Xinxin lithography machine are only the first step.
Matrix is the second step. In the future, Apple will rely on the high profit margin of mobile phones to improve performance in the field of mobile phone chips regardless of cost. Because no matter what, they can make profits through the iPhone's brand premium.
Qualcomm needs to consider costs when designing chips. Even Android phones using the most advanced processes are not as expensive as iPhones.
One more thing, that is, in the field of chip manufacturing, salary costs only account for a fraction of revenue, and the most expensive cost comes from equipment depreciation.
Only by relying on matrix to succeed and form high profits can we rely on high profits to improve performance in the chip field.
At this stage, SMIC will not be stuck by foreign countries. With the tilt of orders, it can catch up with the most advanced foreign processes.
When SMIC catches up with the most advanced foreign manufacturing processes, it will gradually allow the domestic chip supply chain to enter SMIC's supply chain, starting from testing.
Only in this way can a positive cycle be formed.
When Zhou Xin went to Zhangjiang for research, he thought that if he did this, a large number of semiconductor companies located in Zhangjiang would definitely cause Zhangjiang's housing prices to soar faster than imagined.
In future generations, even people like Zhou Xin who have achieved a little success in entrepreneurship will grit their teeth when looking at the housing prices in Shanghai. If they buy, they will feel that it is not worth it. If they don’t buy, they will feel that something is missing.
The progress of Xinxin Technology in Zhangjiang is smoother than Zhou Xin expected.
Unlike Xinxin Technology's smooth progress, Guan Jianying's promotion of cooperation with SMIC was very unsmooth.
According to the original time and space, SMIC's 250nm process will be trial-produced in September 2001 and mass production will begin in the first quarter of 2002.
In other words, the current 220nm prototype of Xinxin Lithography Machine can fully meet the requirements of SMIC.
As the general manager of Xinxin Technology, Guan Jianying will also be responsible for part of the business of Xinxin lithography machines. In his view, a new chip foundry company has emerged in China, and it was founded by Zhang Rujing, a well-known person in the industry.
Xinxin Technology and SMIC naturally have a lot of complementarities. Xinxin lithography machines can fully cooperate with SMIC. The testing of Xinxin lithography machines can be carried out in parallel with the construction of SMIC's factory.
Moreover, when Zhou Xin was at Shenhai, he also had a good conversation with Zhang Rujing. Both of them expressed that they would strengthen cooperation in the future, especially in the field of lithography machines.
As a result, after Zhou Xin returned to San Francisco, Guan Jianying wanted to push the matter aside, but SMIC did all kinds of Tai Chi and didn't accept the move at all.
Zhou Xin lamented that there were still too few chips, and even SMIC couldn’t handle it.
Contrary to the outside world's belief that SMIC poached people from TSMC with high salaries, SMIC's salary back then was only about 70% of that of TSMC. Most of those who choose to go to SMIC are marginalized engineers at TSMC.
Zhou Xin roughly knows why SMIC is unwilling to cooperate with Xinxin lithography machines. The fundamental reason is that Xinxin lithography machines are immature.
For SMIC, which wants to put into production as soon as possible, where can it find the time and energy to help Xinxin Semiconductor test prototypes?
Zhou Xin thought to himself that he still needs to find opportunities to control SMIC. Although it is divided into different companies to do this, chip foundry and chip design are separated. The impact is not big now. When the process is below 22nm, it will affect performance.
After the 22nm process is reached, chip design must be done together with chip foundry.
The quality of a chip does not only depend on the process. The most typical case is that Nvidia used TSMC's 14nm process to defeat AMD's TSMC 7nm chip (GTX 2080 vs Radeon VII). It has higher performance and even lower power consumption. This is the charm of architectural design.
After reaching 22nm, the design rules of the chip manufacturer will not tell the chip design company. TSMC helps HiSilicon OEM. It is impossible to tell HiSilicon how to set the NAND gate layout of the 7nm chip and where to place the MOS of the inverter on the clock line. In which position.
This is a unique secret that chip manufacturers have accumulated based on experience, and it is TSMC's special skill. Therefore, the chip design company in China will only give TSMC a rough pattern and then add a logical sequence.
These logic sequences require multiple tape-outs to be realized. HiSilicon is responsible for the general direction of the chip, and TSMC makes adjustments in the general direction, including how to design the specific layout and how to create masks.
The effect is naturally not as good as Intel and Samsung, and they can start from the bottom.
This is why Qualcomm and AMD chips are not as good as Samsung and Intel chips.
Seventy billion may seem like a lot, but in fact it is completely different when it comes to actual operations.
When the Nasdaq bubble burst, Zhou Xin acquired ARM and NVIDIA. In fact, what Zhou Xin coveted most was Applied Materials.
Applied Materials has almost a monopoly on all equipment except photolithography machines, but these equipment are not as well-known as photolithography machines.
But even if the bubble bursts, the market value of Applied Materials is still US$100 billion. If you want to hold a controlling stake, you must prepare at least twice the capital.
Of course, the acquisition will not be of much use. American cannot watch you move the R&D and production of Applied Materials to China.
"Newman, I came to the school recently. I have some details about your paper that I need to communicate with you." Chenming.
"Okay, Professor, I'll come over as soon as possible." Newman.
In addition to promoting Enterprise QQ in companies such as Riot Games, Matrix and NewPay, which are owned by Zhou Xin, academic organizations such as CAAS have also purchased Enterprise QQ.
When Enterprise QQ was first launched, members thought: What is the difference between this thing and ICQ?
However, considering that most of the funds for CAAS were provided by Zhou Xin, the purchasing company QQ had no objections. It was Zhou Xin's own money anyway.
Moreover, in order to expand its business, Tencent sold Enterprise QQ to CAAS at a very low price.
At the beginning, most members were not used to using corporate QQ to communicate, and everyone was still used to email.
One is because you have developed a habit after sending and receiving emails, and the other is that you can reply to emails intermittently after writing them, and you do not need to reply to others immediately.
Instant messaging software is different. Communication on instant messaging software is a continuous conversation. It is not easy to stop for too long in the middle of the conversation before replying to others.
When Enterprise QQ was first launched within CAAS, few people used it. It was not until the first anniversary of the establishment of CAAS, after Zhou Xin used Enterprise QQ to create several groups, that the software began to be slowly used.
The most used function is to create a group, and then upload some paper materials in the group for everyone to discuss.
Members of CAAS call this a discussion group, which is mainly a discussion group on topics. Later, various eating, drinking and entertainment groups and local dating groups were also established.
Sometimes software promotion needs an opportunity.
This does not include Hu Zhengming. Hu Zhengming has liked Enterprise QQ from the beginning, because he finally can not think about the email format and can communicate with others at any time when he thinks of it.
Therefore, the most frequently used enterprise QQ is not Zhou Xin, but Hu Zhengming. Many friends who have joined CAAS can communicate with each other through enterprise QQ.
At present, only Enterprise QQ is a social software that uses the real names of acquaintances.
After corporate QQ entered Americen, personal QQ was equally aggressive in Americen.
The main participants in NewPay’s second round of financing are the investment banking departments of two mainstream American banks, Wells Fargo and Citibank. In addition to the banks, they are also Walmart.
As the only non-financial institution to invest in NewPay, Walmart has become the biggest benefit in the Millennium capital market. It rose 7% on the day the news was officially announced.
After Wal-Mart invested in NewPay, Wal-Mart purchased Enterprise QQ from Tencent. Wal-Mart deployed local Enterprise QQ on servers in Amerikan's large and medium-sized cities.
“Walmart can treat the group as an online publishing board. After inviting users to join the group, it will regularly distribute some preferential information and limited-time discount information.
Not only can you send discount information, but you can also send information about the latest products on sale. "
This is also the key point for enterprise QQ to attract Wal-Mart to purchase enterprise QQ.
Compared with offline flyers, it is much more convenient to rely on group chats to send information.
Of course, there will be a problem, that is, the communication cost is too high, because there are too many customers in Walmart stores, and you cannot help customers join the group one by one, and now there is no QR code to scan the code to join the group.
In addition, the number of computers is limited, and it is impossible to put some computers in every store to help users join groups, which is very costly.
Currently, Wal-Mart can only print the group number at the bottom of the receipt, and the cashier will give a reminder when handing it to the user.
At the same time, advertisements are placed in the media to remind users that they can join the group to receive benefits.
This strength is too weak and the conversion rate is limited.
Even after users join the group, they rarely follow the group messages in real time.
Therefore, the explosion of enterprise QQ usage among user groups had to be established during Thanksgiving in December, and the situation only improved after the online supermarket created by NewPay and Walmart was launched.
After the user selects a specific offline store on Walmart's online supermarket, and completes the online payment with NewPay after purchasing the goods, a text message will be received on the mobile phone number registered with NewPay after the online payment.
The text message content includes the order number, mobile phone number and SMS verification code.
Users take their mobile phones to offline stores and report their order numbers to dedicated online service clerks. They will verify the mobile phone number and verification code and then package the goods directly to the user.
That means shopping online and picking up goods offline.
This online ordering and offline pickup model did not become popular in Europe until 2015.
Wal-Mart took the lead in launching this model, and the new e-commerce model makes Amazon feel that it is a formidable enemy.
It turned out that their vigilance was correct. After Wal-Mart launched online ordering and offline pickup, Amazon was the first to be impacted.
Because I choose online shopping to save trouble, online shopping at Millennium Amerikan is not cheap, and many products are even more expensive than offline.
Online shopping is also hassle-free and requires no shipping charges. You can make your purchases before you go out, then stop by Wal-Mart on your way home from get off work, and you can buy what you want in five minutes.
Millennium Walmart's offline stores have a much wider coverage than express delivery sites.
"Newman, the idea you gave Wal-Mart is worth at least 5% of the stock. I'm a little jealous of Wal-Mart.
This plan is enough to support an Internet company with a scale of tens of billions. Before 2000, it would have been possible with a scale of hundreds of billions. "After Wal-Mart's online shopping function was launched, Bill Gates complained while having dinner with Zhou Xin.
“For Wal-Mart, they have a new growth point. In terms of business form, it gives users more choices and puts them further apart from other supermarket brands.
The advantage of online Walmart will last at least half a year. If other supermarket brands cannot access NewPay, the advantage will be longer than half a year.
If I were you, I would choose to make this product myself and cooperate with all supermarkets. You are responsible for the purchase and payment on the Internet, and the supermarket is responsible for offline supply. "
In order to obtain this idea, Walmart signed a long-term exclusive cooperation agreement with NewPay. In the next five years, NewPay will be Walmart’s only electronic payment channel.
If it weren't for this condition, Zhou Xin wouldn't help them design this mechanism.
Because Wal-Mart is a listed company, this kind of exclusive agreement will be announced sooner or later, so Zhou Xin explained it directly to Bill Gates, and then continued:
“The business model of ordering online and picking up goods offline is very simple. The only threshold is electronic payment. However, in addition to NewPay, there are several electronic payment options on the market.
And even if they choose NewPay as a partner, NewPay will not refuse. In this way, the advantage of electronic payment will be lost.
This business model requires extremely strong control and influence. I need to control the updates of goods in each supermarket. I need to connect with different supermarkets and ensure that users can pick up goods offline. This is related to warehouse management and goods supply.
Then there are countless supermarkets in fifty-one states across the United States, and my reputation is too great. If they come to do this, I am afraid that even the San Francisco Bay Area has not even taken over, and a bunch of Internet companies with the same business model have popped up.
Not to mention that traditional supermarkets can also do this business. This is one of those things that looks like a delicious cake but is actually an unpalatable English dish.
Only companies like Walmart have the control and confidentiality capabilities to bring the entire country online at the same time.
And before the commercial was released, there were only rumors, but no more specific news. "
Before the launch of Wal-Mart Online, there were rumors that Newman had reached an in-depth cooperation with Wal-Mart and designed an application specifically for Wal-Mart.
This news is especially popular among Wal-Mart investors, who firmly believe: "If Newman does not cooperate deeply with Wal-Mart, why will NewPay accept Wal-Mart's investment? It is obvious that Newman sees the potential of Wal-Mart."
It was later confirmed that Zhou Xin had indeed helped Wal-Mart design the online Wal-Mart. Sometimes the news in the secondary market was true, but retail investors would lag behind in getting the news.
In an interview with reporters, Wal-Mart's CEO claimed: "Yes, Newman helped us design the online Wal-Mart. He is a genius in the product field. Wal-Mart is honored to cooperate with Newman. In the first month after the online Wal-Mart was launched, it achieved nationwide sales. 1 billion in revenue is a very exaggerated figure.”
Wal-Mart's full-year revenue in 2000 was only US$165 billion, and in the American region it was only about US$150 billion.
As America's largest e-commerce website, Amazon's annual sales in 2000 were only 5.8 billion.
After the interview with Walmart CEO was released, Amazon's stock price continued to plummet. Amazon's stock price, which was originally affected by the bursting of the Nasdaq bubble, shrank by 80%, further shrunk.
Online Walmart's revenue in the first month was 1 billion, and its full-year revenue is expected to far exceed Amazon's when it first went online. So why do investors still invest in Amazon?
Walmart felt that it had made a profit. Not only did it acquire nearly 10% of NewPay’s shares, it also successfully found a second growth point for the company.
It is very difficult for traditional enterprises to integrate with the Internet, and it is even more difficult for traditional retailers.
It’s not a loss for Zhou Xin and NewPay. With the two payment scenarios of Walmart and Riot Games, NewPay can be guaranteed to be invincible.
When the mobile Internet matures, NewPay will firmly occupy the top spot with the help of hardware.
Since current mobile phones do not have browser functions, payment confirmation can only be done through verification codes and there is no way to scan QR codes.
If you have a browser, you can use the SMS link to jump to the webpage to display the QR code.
2G networks can do this.
Therefore, Zhou Xin is eagerly looking forward to the launch of Matrix. When Matrix is launched, online shopping and offline pickup will be more convenient.
The success of online Walmart is just another validation of Zhou Xin’s halo.
This caused Bill Gates to be obsessed with NewPay. As NewPay becomes more powerful in various channels, Microsoft is bound to have electronic payment functions after Xbox goes online.
Bill Gates has fully realized the importance of NewPay, which is by no means as simple as an online bank.
I know the progress is slow, I can only add more updates to make up for the progress, and I updated 8K today!
Within the scope of Crow's ability, Crow will try to update as much as possible.
(End of chapter)
