The number of people working in the U.S. chip industry has hit a record low. What is the future of integrated circuits?

The Wall Street Journal reported that by the end of 2015, Alphabet Inc. (Google's parent company) and Facebook had a total of 74,505 employees. The combined market value of the two companies is twice that of Microsoft, but the workforce is one-third less. Currently, the combined market value of the top five technology companies in the United States (Apple, Alphabet, Microsoft, Facebook, and Oracle) is $1.8 trillion, 80% more than the market value of the top five technology companies in 2000 (Cisco, Intel, IBM, Oracle, and Microsoft), but the total number of employees is 22% less.

The Semiconductor Industry Association (SIA) pointed out that chips are still the third largest export item of the United States, second only to automobiles and aircraft. In 2001, the total number of employees in computer and electronics companies in the United States reached 1.87 million, which dropped to 1.03 million in August 2016. Take Micron as an example. In 2013, there were only 11,300 employees in the United States, lower than 14,000 in 2000. At the same time, the number of employees outside the United States jumped from 4,800 to 19,600. Apple Inc. has a total of 80,000 employees in the United States, accounting for two-thirds of the total, but half of its American employees are engaged in retail work.

The U.S. Bureau of Labor Statistics recently announced that employment in the semiconductor and electronic component manufacturing industry fell by 1,700 people from the previous month to 358,600 in September 2016, the lowest number since statistics began in 1985 and a 50% drop from the highest record in history (714,500 in January 2001).

(Source: FRED)

The author recently attended the "ICCAD2016" in Changsha. In a chat with Wu, the general manager of a semiconductor company in China, he joked that "the semiconductor industry is not a job for humans to do. It is to produce products for humans." Considering the recent trend of EE switching to CS, does EE really have no future?

Attachment: What is the future of microelectronics (integrated circuits)? : Li Hongjiang

Apology statement: When I first published this article on October 14, I was negligent and directly copied the content of the search engine without the author's authorization. I am extremely sorry for the content Li Hongjiang posted on Zhihu!

Thanks to the author Li Hongjiang for his wonderful analysis and answers. This article was first published on Zhihu. The original link is as follows: https://www.zhihu.com/question/26498266/answer/84698109. It is recommended to follow the author’s personal WeChat public account: IC Gas Station

This question is about the future of integrated circuits, so it must be divided into two sub-questions to answer.

Question 1: Is there a future for the development of the integrated circuit industry?

Question 2: If a person chooses the integrated circuit industry, does his personal development have a future (money)?

Short answer to question 1:

The development of the integrated circuit industry has entered a bottleneck period, or a slow development period. However, in the long run, there is still great potential for development, and after the next technological revolution, there may still be relatively rapid development.

Short answer to question 2:

For individuals, entering the integrated circuit industry is a high-investment but mediocre choice. The so-called good career development is to work for a leading giant in the industry. The possibility of entrepreneurial success is very small, and it is difficult to become rich.

Long answer to question 1:

The integrated circuit industry is a miracle in the history of human industrial development. Since Gordon Moore proposed this law in 1965, countless scientists and engineers have continued this exponential growth law for nearly 50 years. Although the growth has slowed down in recent years, it still maintains exponential growth. The tiny transistor concentrates almost all the essence of human natural science and engineering. Doubling in a short period of time is not uncommon, but the continuous exponential growth for 50 years is difficult. In these 50 years, humans have achieved unprecedented growth in computing power. The figure below is the computing power growth curve obtained by humans per $1,000 per second over the past 100 years.


Microelectronics practitioners can say that they both love and hate Moore's Law. They love it because it tells us the direction to move forward, that is, to keep reducing costs and improving performance, so that we can remain competitive. In the business world, knowing which direction to move forward is more important than being uncertain about the direction of development. They hate it because it puts everyone in a prisoner's dilemma, that is, if they cannot double the performance (cut costs by half) within two years (18 months in the early stage), they will be defeated and eliminated by their competitors. The consequence of this is that electronic products will be reduced in price in a short time after they are launched on the market. Everyone is used to the fact that electronic products must be reduced in price, and consumers are spoiled invisibly. Faced with this situation, manufacturers can only hope to obtain as much profit as possible in the early stage of the launch of new products, and then quickly invest funds in the research and development of the next generation of products, in the hope of continuing to lead in the next generation of products, so as to have certain benefits. This approach worked in the first 40 years of Moore's Law, but when the process was reduced to 22nm, an important change occurred. As the R&D investment in new processes increased, the prices charged by foundries for new processes also increased. People discovered at this time that the cost of a single transistor at 22nm was actually higher than that at 28nm!

Although people can continue to make transistors smaller and their speed can still be increased, one of the statements of Moore's Law, "the cost of chips is reduced by half", has completely failed. What does this mean? It means that the path of chip manufacturers to reduce costs by pursuing new processes in the past forty years has been blocked. In the past, using new processes can not only bring performance improvements but also reduce costs, but after 22nm, manufacturers are faced with a choice, whether to spend a high cost to pursue new processes, or to be conservative and use old processes. At the same time, the cost of a newer process is getting more and more expensive, and many small and medium-sized manufacturers are already unable to do so. Only the giants in the industry have the capital to pursue more advanced processes. For example, the full layer mask price of the 22nm process when it first started mass production was about tens of millions of US dollars (I made it up), and tens of millions of dollars is a considerable investment even for a medium-sized fabless, not to mention the increase in the complexity of front-end and back-end design brought about by advanced processes, the increase in the cost of EDA tools, etc.

In order to cope with the rising R&D and operating costs, major companies have realized that the era of fighting alone and making a comeback is over, and the right choice to survive the winter is to join forces with strong companies. Therefore, in 2015, there were more mergers and acquisitions in the integrated circuit industry than in the past six years combined, including the largest merger in the history of integrated circuits between Broadcom and Avago (more than 30 billion US dollars), Intel's acquisition of FPGA giant Altera, NXP's acquisition of Freescale, and other mergers and acquisitions that were unimaginable ten years ago. This is completely consistent with the law of industry integration in many traditional industries after entering the mature stage of development. In 2015, according to statistics from the Semiconductor Industry Association (SIA), the sales of the entire semiconductor industry fell by 0.2%, and the expected industry growth in 2016 is only about 1%. And what was the GDP growth rate of the whole world in 2015? The IMF estimated it to be 3.1%. Yes, the semiconductor industry is already an industry that cannot even outperform the average GDP.

So where will the next growth point be? Some say it is the Internet of Things (IoT), some say it is VR, and some say it is automotive electronics. I can’t make a judgment. But it is obvious that if the entire semiconductor industry wants to develop rapidly again, it needs a broad market and irreplaceable demand like the smartphones of the past. As for connecting light bulbs and refrigerators to the Internet, or connecting cars to 4G, they will not be successful.

But I still have confidence in the semiconductor industry, and this confidence comes from human pursuit of technology and convenient life. Back in the 1980s, many of the novel things in the future world imagined in the famous movie "Back to the Future" have been realized in 2015, 30 years later, and are even more awesome than what was imagined in the movie. Just like the demand for mobile phones was created, perhaps N years later, another demand will be created and become a part of our lives, and I believe that such demand must require the progress and support of semiconductors. This is my conclusion: in the long run, there is still great potential for development, and after the next technological revolution, there may still be relatively rapid development.

Long answer to question 2:

It is an indisputable fact that the current treatment in the hardware industry is not as good as that in the software industry. And with the integration of the industry, various companies laid off employees last year, and the market is full of experienced engineers. And what I heard from the school is that at present, in many universities in the United States, EE students are turning to software, including famous schools like MIT. So the first conclusion is that for fresh graduates, there are fewer job opportunities than in previous years (China still has a relatively large demand due to national support), and there are fewer companies to choose from. After entering the company, there is also a certain risk of being laid off. The income level is lower than that of software, but it still has advantages compared with other traditional industries such as machinery manufacturing and chemical industry.

In terms of career development after entering a company, it is difficult for hardware companies to expand as fast as software companies, so the salary increase is not high. In the United States, it is good enough if the salary increase can resist inflation.

Another thing is that it is too difficult to start a business in the chip industry. The cost of tape-out is high, EDA tools are expensive, and VCs are not interested. As a result, the number of startups in the chip industry is far less than that in the software industry. A software startup may survive for two or three years after raising a few million dollars. In the chip industry, sorry, the license fee for EDA tools and the cost of tape-out are all used up. So more and more people have to stay in big companies, work hard for years, accumulate experience, and take the technical route of engineer->senior->staff-->principle engineer, or take the management route. If you want to get rich overnight like a software company going public, I think the probability is not much different from winning the lottery.

As for investment, this industry is, after all, a technology-intensive industry. A master's degree is the stepping stone to enter the industry, and the courses studied are also quite difficult.

Finally, I would like to add two sentences. Many people argue about the difference between software and hardware. In fact, the difference is very obvious. https://zh.wikipedia.org/wiki/中国工业分类

Integrated circuits belong to the secondary industry, which is manufacturing, while software belongs to the tertiary industry, which is the service industry. If you understand this, many problems will become clear at a glance.

This article is a combination of relevant reports from technews and the wonderful analysis and answers from Zhihu’s Li Hongjiang. If it is inappropriately reproduced, please contact the backstage in time. Thank you.

[About reprinting]: Reprinting is limited to full text reprinting and the title and content of the article must be completely retained. No deletion or addition of content is allowed to circumvent the original protection, and the beginning of the article must indicate: Reprinted from the WeChat public account "Semiconductor Industry Observation icbank". Thank you for your cooperation!

【About submission】: Semiconductor elites are welcome to submit articles. Once accepted, they will be published with your name and rewarded with a red envelope! Please indicate "Submission" in the title of the manuscript email, and indicate your name, phone number, unit and position in the manuscript. Welcome to add my personal WeChat account MooreRen001 or send an email to jyzhang@moore.ren

Click to read the original text and join the Moore Elite