Even though microchips have become indispensable in so many products, their development and manufacturing are still dominated by a few manufacturers, who have limited capabilities and appetites and cannot produce the main commodity chips of today’s technology. And because manufacturing chips requires hundreds of manufacturing steps and months of production time, the semiconductor industry cannot quickly adjust to meet the surge in demand driven by the pandemic.
For decades, we have been worrying about how to engrave features of a few nanometers on silicon wafers. The spirit of Moore’s Law-the expectation that cheap, powerful chips will be readily available-is now threatened by something more mundane. : Inflexible supply chain.
Twenty years ago, there were 25 manufacturers in the world producing leading chips.only today TSMC (TSMC) in Taiwan, Intel In the U.S., and Samsung South Korea has facilities or fabs that produce the most advanced chips. And Intel, which has long been a technology leader, is struggling to keep up, repeatedly missing the deadline for producing the latest generation of products.
One reason for the merger was that the cost of building a facility to produce the most advanced chips was between 5 billion and 20 billion US dollars. The chips manufactured by these fabs have functions as small as a few nanometers; in industry jargon, they are referred to as 5-nanometer and 7-nanometer nodes.Most of the cost of a new fab is used to purchase the latest equipment, such as a kind of Extreme ultraviolet lithography (EUV) machines, which cost more than 100 million U.S. dollars. The EUV machine is exclusively manufactured by ASML in the Netherlands and is used to etch detailed circuit patterns with nano-sized features.
Chip manufacturers have been committed to EUV technology research for more than two decades. After billions of dollars of investment, EUV machines were first used in commercial chip production in 2018. “The tool is 20 years late, 10 times over budget, because it’s awesome,” said David Kanter, executive director of the Open Engineering Alliance focused on machines. Learn. “It can even work almost magical. It’s completely like science fiction.”
Such a huge effort has made it possible to make billions of tiny transistors in the Apple M1 chip made by TSMC. It is the first generation of cutting-edge chips that rely entirely on EUV.
It makes sense for Apple to buy the best chips because these chips are used in the latest MacBook and iPhone models, and they sell for millions at luxury brand prices. “The only company that really uses EUV a lot is Apple. They sell a $1,000 smartphone and have very good profits,” Kanter said.
Not only are the fabs that make such chips expensive, but the cost of designing extremely complex circuits is now beyond the capabilities of many companies.Apart from Apple, only the largest technology companies require the highest computing performance, such as Qualcomm, AMD, with NvidiaSri Samavedam, senior vice president of Imec CMOS technology at Leuven International Research Institute in Belgium, said that he is willing to pay hundreds of millions of dollars to design chips for frontier nodes.
More and more companies are producing notebook computers, TVs and cars using chips made with old technology, and the surge in demand for these products is at the core of the current chip shortage. In short, most chip customers cannot afford – or don’t want to pay – the latest chips; today, a typical car uses dozens of microchips, while electric cars use more. It adds up quickly. On the contrary, manufacturers of products such as automobiles have insisted on using chips made using old technology.
More importantly, many of today’s most popular electronic products do not require sophisticated chips at all. “For example, change A14 [iPhone and iPad] The chip is in every computer in our world,” said Hasan Khan, a former PhD researcher at Carnegie Mellon University who studied the impact of the end of Moore’s Law on public policy and is currently working at Apple.” Your smart thermometer at home doesn’t need it, and your car doesn’t need 15 because it consumes electricity and is very expensive. “
The problem is that even if more and more users rely on older and cheaper chip technology, the giants in the semiconductor industry are still focused on building new cutting-edge fabs. TSMC, Samsung, and Intel have all recently announced billions of dollars in investment in the latest manufacturing facilities. Yes, they are expensive, but this is where the profits are-and it has been the future for the past 50 years.
TSMC is the world’s largest chip foundry manufacturer. Nearly 60% of its 2020 revenue comes from manufacturing leading chips with features of 16 nanometers and below, including Apple M1 chips manufactured using a 5 nanometer manufacturing process.
To make matters worse, “no one is building semiconductor manufacturing equipment to support old technology,” said Delford, the chief analyst of the Electronic Component Industry Association of the trade association in Alpharetta, Georgia. “So we are kind of trapped between the rocks and hard spots here.”
All of this is important to technology users, not only because it causes supply disruptions today, but also because it threatens the development of many potential innovations. In addition to being more difficult to obtain, cheaper commodity chips have also become relatively more expensive, because each generation of chips requires more expensive equipment and facilities than previous generations.
Some consumer products will only require more powerful chips. The construction of faster 5G mobile networks and the rise of computing applications that rely on 5G speeds may force investment in dedicated chips designed to communicate with dozens or hundreds of Internet-connected devices.Automotive features such as advanced driver assistance systems and in-vehicle “infotainment” systems may also benefit from leading chips, as electric vehicle manufacturers have proven. Partnership reported by Tesla Both TSMC and Samsung are developing chips for future self-driving cars.
However, when developing products for the “ubiquitous intelligence” future, buying the latest cutting-edge chips or investing in specialized chip designs may not be practical for many companies. Kanter said that consumer device manufacturers (such as Wi-Fi-enabled sous vide cooking) are unlikely to spend money to develop their own dedicated chips to add more advanced features. Instead, they may rely on any chips made using old technology.
He said that the “mean profit” of low-cost products such as clothing leaves little room for more expensive chips, which will increase the price of each item by one dollar—not to mention 10 or 20 dollars. . This means that increasing computing power may hinder the development of clothing, such as clothing that can detect and respond to voice commands or weather changes.
The world may not have more advanced sous vide machines, but the lack of cheaper and more powerful chips will come at a real price: the era of invention driven by Moore’s Law is over, and its decades of promised increasingly cheaper computing power Will be used for the next innovation.
Most of today’s chip customers use cheaper commodity chips, which represents a trade-off between cost and performance. As the global demand for computing power grows, the supply of such commodity chips seems far from enough.
Willy Shih, a professor of management practice at Harvard Business School, said: “The use of semiconductors in cars is still increasing, and the use of semiconductors in ovens and the use of various things are still increasing.” “The problem is here. , Will there be a shortage next?”
In early 2021, President Joe Biden signed an executive order mandating Supply Chain Review Bargaining chips and support Congress’s bipartisan push to at least approve $50 billion for semiconductor manufacturing and researchBiden also held two White House summits with leaders in the semiconductor and automotive industries, including a meeting on April 12, where he displayed a silicon wafer in a conspicuous position.
These actions will not quickly solve the problem of chip supply and demand imbalance. But at least, experts say that today’s crisis provides the US government with an opportunity to try to finally repair the supply chain and reverse the overall slowdown in semiconductor innovation—perhaps it can also enhance the US’s ability to manufacture much-needed chips.
As of 2019, it is estimated that 75% of all chip manufacturing capacity is located in East Asia, while the US share is about 13%. Taiwan Semiconductor Manufacturing Co., Ltd. alone has nearly 55% of the foundry market for processing consumer chip manufacturing orders.
The overriding thing is the US-China competition. China’s national champion company SMIC has been building fabs, but it still lags behind cutting-edge technology for five or six years in terms of chip technology. But Chinese foundries are likely to help meet the global demand for chips built on old nodes in the next few years. “Given the state subsidies they receive, Chinese foundries may become the lowest cost manufacturers because they have established fabs at the 22nm and 14nm nodes,” Khan said. “Chinese fabs may not be competitive in the frontier field, but they can meet more and more demands.”