Divine Translation Bureau is a compilation team under 36Kr, focusing on science and technology, business, workplace, life and other fields, focusing on introducing new technologies, new ideas and new trends from abroad.
Editor's Note: Once upon a time, Intel was the banner of the semiconductor industry, with leading technology, rich profits and extremely moist life. OEM for others is something it has never considered. After all, it takes too much investment and a long construction period, and the technology may fall behind before you recover the cost. But things have changed recently. The shortage of chips caused by various reasons has made the foundry headed by TSMC earn a lot of money, and the technology of the foundry has also been continuously improved by undertaking different customer needs. Not only can Intel not make that much money now, but its technology is also backward. So the semiconductor giant plans to attack for the second time and do OEM business. This time, can it succeed? The article comes from compilation and we publish it in two parts. This is the second part.
Image source: Intel
Focus on:
It is easier to slip if you take too many steps at once than a few small steps.
The real challenge in the OEM market is to ensure that every customer's needs are met.
Making customers trust is as important as manufacturing ability.
For Intel and the United States, adopting OEM mode represents a major shift.
Intel in the sunset: Can making chips for others make a comeback?
It's easy to take too big steps.
At the same time, TSMC is rising. Throughout the 2000s, TSMC was well funded, which enabled them to invest in larger and more advanced fabs. This company proved that OEM business is not only feasible, but also profitable.
Soon, TSMC began to extend an olive branch to Apple, which finally put the company in a leading position. At that time, Apple had begun to design its own chips for the iPhone. Although its early version was made by Samsung, patent disputes led Apple to start looking for alternative manufacturers.
Apple has been testing TSMC chips for several years. In 20 15, they began to directly compete with TSMC's chips and Samsung's chips on the iPhone6s. Samsung still has an advantage in name-its transistors are smaller, so it should be more efficient than TSMC's. But in most cases, this is not the case. TSMC's chip maintained its position. When the iPhone7 was launched, Samsung was out. Analysts praised the phone as "very thin and powerful, and TSMC has been in the iPhone since then.
TSMC has mastered the art of low-power processors, partly because it has improved its technology in the process of OEM chips for hundreds of different customers. Hu Hu, a professor and former chief technology officer at the University of California, Berkeley, said: "Once you have a large number of OEM customers, it is impossible for all of them to synchronize their product cycles. Almost at any time, as long as a new technology appears, some customers will be willing to pay for it.
"Compared with taking a few small steps, it is easier to slip if you take too many steps at once.
-Professor Hu from the University of California, Berkeley.
A foundry that can take care of multiple customers and technologies can quickly surpass competitors. For example, TSMC's scale enables it to master extreme ultraviolet lithography technology faster than any other company, thus reducing the steps required to manufacture advanced chips and increasing the output of the wafer factory.
Nonsense, large foundries usually have dozens or hundreds of customers, which encourages them to take small steps because there will always be interested customers. He said: "It is easier to slip if you take too many steps at once than a few small steps." Companies are also more likely to recover from their mistakes.
Historically, Intel usually used the Great Leap Forward to try to reflect Moore's Law. According to Moore's Law, the density of transistors will double every 18 to 24 months. For most of the company's history, Intel has been very successful and introduced impressive updates, which made the company one step ahead of its competitors. But then at 20 15, Intel slipped away. The company announced that the time to market of its 10 nanometer process chip will be delayed. 20 17 announced the extension again. Soon, the industry giant not only failed to share the points with his rivals, but fell behind.
At the same time, other companies began to follow Apple's example and design their own chips instead of buying ready-made components like Intel. With the lead of TSMC, more and more companies began to hand over their designs to TSMC for manufacturing, which gave the company more opportunities to improve its technology. Today, about 90% of cutting-edge chips are made by TSMC, and the rest are made by Samsung.
Nonsense: "It is extremely difficult to develop a new generation of technology. In terms of extremely advanced technology, Intel now lags behind TSMC and Samsung, which can be traced back to the fact that Intel did not participate in OEM.
Trust problem
PaulTriolo, head of geo-technical practice at EurasiaGroup, said: "The real question is whether Intel's entry into the OEM business will help them return to the leading position.
The success of the semiconductor industry depends largely on the scale. The more you sell, the more opportunities you have to improve the process, which will help you enter the next process node. Trioreau said: "This is the case with TSMC. To some extent, they can reach the critical scale of their ability and customer relationship. Then a virtuous circle is formed-that's what you need.
Nonsense: "Without OEM, can the United States maintain its leading position in technology? This is indeed the problem that Gersinger tried to solve. If we don't get involved in OEM business, Intel can't stay ahead in technology-I certainly agree with this conclusion.
For Intel and the United States, adopting OEM mode represents a major shift. Historically, leading American companies have either acted as IDM to design and manufacture chips or as fabless designers to outsource production to another company. Part of the reason is that most of the profits of computer chips come from design and sales, not manufacturing. Jennifer Guan, deputy director of innovation and research at California State University at Monterey Bay, said that success as a pure OEM "is almost a special case in history. Not only that, she added: "TSMC has proved that this is actually a profitable business. Many people didn't expect this to happen.
Now, Intel seems to be putting its enthusiasm into OEM business. So far, Intel has announced that Qualcomm and AWS have become its own customers. Klaus Schugraff, vice president in charge of OEM strategy and planning, said that 65,438+000 other companies have expressed interest. "Our orders are very long and customers come from all walks of life. At present, Intel seems to focus on the high-performance market. Shugraff said: "We found that the business growth in the next five or ten years will mainly come from frontier areas.
However, the real challenge of OEM market is not whether it can attract customers, or even whether it can develop better technology-but whether it can ensure that every customer's needs can be met. Hunter, a researcher at Georgetown University, said that one of the reasons why TSMC is different is that it "can meet the needs of a wide range of customers, especially chip design companies." Intel's experience in this area is not so rich.
As a pure foundry, TSMC's customers don't have to worry that the unique elements in their designs will be stolen by manufacturers and used on their own chips. Professor H.-S.PhilipWong of Stanford University led the research and development of TSMC from 2065438 to 2020. He said: "When communicating with TSMC, one of the basic principles they mentioned was customer trust. This is as important to them as manufacturing ability.
Intel seems to have learned from its last failed attempt, which was shelved five years later in 20 18. Different from last time, the new Intel OEM service is an independent business unit, reporting directly to Gelsinger. Shugraff said that the company has been building key relationships with suppliers for many years, so that fabless companies don't have to cram their designs into Intel's way of doing things. Moreover, in terms of process settings, customers can also avoid worrying about their secrets appearing in other chips.
MargaretHenschel, an employee of Intel Fab32 mass manufacturing plant, is walking through Chandler, Arizona. Tim Herman/Intel
Shugraff said: "We can isolate customers' confidential information and provide services to these customers while using Intel core. Any process customized by Intel for a specific customer "can only be known to those who need to know. He added that the company is also committed to "being fair and transparent to customers in the way of allocating production capacity." This looks like dealing with other foundries.
This acquisition may help Intel acquire some of these skills quickly. Earlier this year, there were rumors that Intel was negotiating with Grofonder, but so far there has been no progress. Meyer said: "Intel's previous attempts to do OEM failed mainly because it did not focus on customers, nor did it focus on manufacturing a large number of components for a large number of different customers. But Grofonder did it.
Shugraff admits that adapting to a wide range of customer requirements may be a test for the company, because the company has always relied on the close relationship between its internal designers and its manufacturing department. But he thinks Intel can overcome this challenge. "This will make us a little nervous. But fundamentally, we are ready for this.
America at the Crossroads
It will take some time for anyone to judge whether Intel's ambition of OEM is successful or not. Observers believe that it will take at least three years to reach a conclusion, and it is more likely to take five years. It takes several years to build a fab, and it takes months or years to test and produce new chip designs. Professor Shi of Harvard University said: "These things always take some time.
In the long run, Intel's success depends on American industrial policy. TSMC's cost is lower than Intel's, partly because of its strong support. The Semiconductor Industry Association of the United States estimates that the cost of building and operating a wafer factory in the United States in 10 is 30% higher than that in Asia, and about half of the cost difference is attributed to government subsidies.
The US Congress is considering injecting about $50 billion into the semiconductor industry to encourage R&D and construction of fabs in the United States. This can promote fair competition to a certain extent, but it will also bring new challenges. Jin Zhijie, dean of the School of Engineering at the University of California, Berkeley, said that if this bill is passed, the semiconductor industry will create tens of thousands of new jobs every year. She said: "This means that we need 5000 to 10000 fresh graduates every year. No university can meet the demand of this kind of labor force development, not even a university system like the University of California.
Jin Zhijie, Dean of the School of Engineering at the University of California, Berkeley, is talking with freshmen who are taking management, entrepreneurship and technology courses. David palmer Morris Bloomberg Co.
Hunter said that in the short term, highly skilled immigrants will help fill this gap. But in the long run, the United States needs to find ways to "increase the number of talents," he said. "Talents at all levels are needed, from those with professional degrees to those with doctoral degrees. According to these ideas, Jin Zhijie and her colleagues have been developing a project that spans industries, universities and community colleges to meet the demand. She said: "We hope to operate the best practices of semiconductor education and promote these practices throughout the United States.
Both the United States and Intel still have a steep mountain road to climb if they want to regain their leading edge in semiconductor manufacturing, but the respondents have no intention of giving up the opportunity. Hunter said: "I think there is a view that letting the' whole company' focus on manufacturing improves the possibility of TSMC surpassing Intel. But I don't think what happened is inevitable. The United States has talents from all over the world, we have strong intellectual property protection, and we have an ecosystem of design companies eager to cooperate with American foundries.
Professor Hu of Berkeley agrees. He said: "We have a leader, and that is Intel. We have a technical foundation and can basically compare with the best universities in the world. We still have the best universities in this field. Maybe what they lack is a sense of urgency. Hu added that success "actually depends on whether you think you have to do it."
At present, many people think that the United States, especially Intel, has no choice. Meyer said: "We have seen that this kind of * * has aggravated the relatively small shortage of semiconductors and caused the global economy to fall into chaos. The success of the United States depends largely on this.
Translator: Bossi.