Dai Weimin, Chairman of VeriSilicon: To be a chip design platform as a service (SiPaaS), one must be able to endure loneliness and withstand temptation

Taking the standard cell library (the most basic structural unit required for chip design) as its starting point, VeriSilicon Holdings Co., Ltd., the first domestic chip design-as-a-service platform, was established in August 2001.

Dr. Weimin Dai, founder, chairman, president and CEO of VeriSilicon, received his bachelor's degree in computer science and his doctorate in electrical engineering from the University of California, Berkeley, and was a tenured professor in the Department of Computer Engineering at the University of California, Santa Cruz.

Prior to founding VeriSilicon, Dr. Dai served as co-chairman and chief technology officer of Celestry (acquired by Cadence in 2002), and was the founder, chairman and president of Ultima, one of Celestry's predecessors.

In addition, Dr. Dai Weimin is the founding chairman of the IEEE International Conference on Multi-Chip Modules and the IEEE Seminar on Integrated Design of Chip Packages. He has served as the associate editor of the IEEE Circuits and Systems Monthly and VLSI Systems Monthly. He has published more than 100 papers in various technical journals and conferences, and was awarded the Presidential Young Investigator Award in 1990.

From university professor, international conference chairman to company CEO, Dr. Dai Weimin has had a rich career experience. The biggest and most difficult transition was the first transition from university professor to company CEO. So how did Dr. Dai Weimin complete this transition?

"In my opinion, this change of identity is not a big problem, because American university professors are very different from Chinese university professors." Dr. Dai Weimin explained, "On American university campuses, professors are generally called 'bosses' because American university professors need to pay graduate students' salaries, while they can only receive nine months' salary a year. The salaries of graduate students and the remaining months of salary need to be raised by professors themselves, just like the sales of a company. Many people are not used to this identity at first, and they live very poorly in the first few years. If they cannot get a tenured professorship within six years, they must leave the school. But I adapted quickly, raised enough funds to train more than 10 doctoral students at the same time, most of whom are from China, and obtained a tenured professorship two years ahead of schedule. This experience was very helpful for my later work as a CEO, and the change of identity was naturally not a big problem, and it can even be said to be a natural outcome."

Dr. Dai Weimin believes that his experience as a professor has been of great help in running a company, especially in terms of valuing talent. As the president of the UC Berkeley Shanghai Alumni Association, Dr. Dai Weimin has promoted the establishment of the Shanghai Zhangjiang Berkeley Engineering Innovation Center, vigorously cultivating talents and introducing cutting-edge technologies.

Shanghai Zhangjiang Berkeley Engineering Innovation Center

Dr. Dai Weimin has been present at almost every campus recruitment event of VeriSilicon over the years and participated in the whole process. Dr. Dai Weimin attaches great importance to cooperation with schools and insists on joint talent training between enterprises and schools. Over the years, he has introduced various top industry experts to the school for lectures, and has jointly carried out "VeriSilicon" courses with the school. He has successfully held three fully enclosed "VeriSilicon Cup" circuit design competitions for key Chinese universities based on FD-SOI advanced process technology and with analog and RF circuit design as the proposition. This is not only about bringing advanced technology to campus, but also shows Dr. Dai Weimin's good intentions to give back to the school.

"Core Original Cup" Award Ceremony

This kind of rigorous academic character and the attitude of attaching importance to talent cultivation, just like teaching, are also reflected in the corporate culture of VeriSilicon - "Fair, Care, Share, Cheer". From company executives to front desk staff to drivers, every company employee has stocks and enjoys the benefits brought by the company's development; the company pays attention to the career development of employees. The middle-level leaders of VeriSilicon are basically cultivated independently within the company, and the company also promotes outstanding employees by exception; the company pays attention to the lives of employees. The company arranges a rich breakfast to encourage employees to avoid rush hour to work instead of mandatory clocking in. Every Mid-Autumn Festival, VeriSilicon will send a box of exquisite moon cakes to the parents of each Chinese employee and attach a greeting card written by Dr. Dai Weimin to thank the employee's family; the company arranges a variety of sports activities and competitions to encourage employees to take time to exercise, and holds a "VeriSilicon" sports meeting with rich projects every autumn; the company holds a general staff meeting every quarter, and Dr. Dai Weimin personally tells employees about the company's development and operation, and employees can speak freely at the meeting; Dr. Dai Weimin also personally participates in family day activities held in various places, and completes games and challenges with employees and their families.

There is a strong humanistic spirit in the corporate culture of VeriSilicon. At the annual Spring Festival meeting, VeriSilicon not only invites art masters to bring wonderful performances, but also organizes employees' children to perform their talents, showing their happy innocence and feeling the warmth of "core" family in colorful collective activities. Dr. Dai Weimin himself loves classical music very much, and often gains inspiration from the collision of sensibility and rationality in his exchanges with artist friends. During the 2014 Shanghai FD-SOI Forum, which coincided with the 50th anniversary of the establishment of diplomatic relations between China and France, VeriSilicon hosted the "Night of VeriSilicon--Sino-French Sound and Picture Symphony Concert", which was presented by the famous French painter Gerard Economos, the famous Shanghai-based pianist Song Siheng and the famous Shanghai-based American soprano singer Wang Zuoxin, together with the Shanghai Opera House Symphony Orchestra, for about 1,600 spectators at Shanghai Culture Square. A unique audio-visual feast was presented, perfectly realizing the combination of high technology and art.

VeriSilicon Night Concert

Dr. Dai Weimin perfectly integrates the two identities of university professor and company CEO, complementing each other's strengths instead of simply switching identities. This characteristic is also clearly reflected in Dr. Dai Weimin's thorough analysis and clear views on the development of the semiconductor industry. It can be said that in Dr. Dai Weimin's career, many times are full of academic speculative color and extremely rational and dialectical thinking.

Dr. Dai Weimin's rational and dialectical way of thinking is also reflected in the process of the establishment of VeriSilicon.

VeriSilicon was founded in 2001, when my country's chip foundries were just starting out and faced severe intellectual property difficulties. Without the intellectual property foundation of the standard cell library, it was extremely difficult to do foundry business. At that time, SMIC was also just starting out and was still in the period of digging deep into the foundation.

Why is it called digging the foundation? The beginning of China's semiconductor industry can be traced back to the 1950s and 1960s, when semiconductor products were made using traditional methods. In terms of time, it was much earlier than Taiwan, but it did not accumulate enough core technologies including standard cell libraries.

In 1998, the country held a seminar on semiconductor development strategy and invited Dai Weimin to attend as an international expert. China decided to build SMIC in Zhangjiang. "At that time, I thought that China's IC industry would have a big wave." Dr. Dai Weimin said. It was also after this meeting that Dr. Dai Weimin returned to Shanghai in 2001 to establish VeriSilicon. "We may be the first chip design company to settle in Zhangjiang High-Tech Park." Dr. Dai Weimin said that before that, he was a tenured professor in the Department of Computer Engineering at the University of California, Santa Cruz. "There are only two tenured professions in the United States, one is the Supreme Court Justice and the other is a tenured professor. The US President has a term limit, and the CEO of a company may be fired at any time, but there is no need to worry about these two professions." Dr. Dai Weimin recalled that even so, he chose to give up the tenured professorship and return to China to start a business.

VeriSilicon Shanghai

In Dr. Dai Weimin's opinion, without a standard cell library, it is like building a house with only a frame but no bricks. Even if you have extraordinary martial arts skills, you can only talk about it on paper. When talking about the original intention of the establishment of VeriSilicon, Dr. Dai Weimin said, "So the positioning of VeriSilicon at the beginning was very clear, which was to ' burn bricks and tiles ', because building a house like a chip is impossible without bricks and tiles."

Later, with the improvement of domestic design capabilities, importing standard cell libraries was no longer a problem. How to take IP and design services to the world became the next challenge for VeriSilicon.

"Unlike the early days, our positioning now is to help others add bricks and tiles and build a better house, rather than competing with others. We must be clear that the house belongs to others. This is the main reason why the company later turned to IP and design services." Dr. Dai Weimin said when talking about VeriSilicon's current business positioning.

Dr. Dai Weimin said that more than ten years ago, when VeriSilicon was first established, it first explored the Taiwan market. At that time, Taiwan's IC design companies had a strong ability to control the market and would not do unprofitable business. Basically, the products they made could be sold. Under the fierce market competition in Taiwan, the VeriSilicon team has been tempered into steel and finally mastered "unparalleled martial arts", so that they can handle any project with ease in the future.

With its unique skills acquired in the Taiwan market, VeriSilicon later ventured into mature markets such as the United States and Japan, and also made considerable gains. International manufacturers such as IBM, Google, Microsoft, Marvell, Broadcom, Renesas, Sony and Sharp have become partners of VeriSilicon, and VeriSilicon's international reputation has been established.

After that, VeriSilicon expanded its business to the Chinese market. This is different from the strategy of many companies that first develop the domestic market and then move to North America, from low-end to high-end. VeriSilicon made such a strategy based on the development status of markets in different regions at different stages, which can be described as "going with the flow". At that time, many Taiwanese companies had withdrawn or planned to withdraw from the Chinese market. In their view, after so many years of development, the Chinese market did not achieve the expected explosive growth.

"Enter when you shouldn't enter, and withdraw when you shouldn't withdraw." This is how Dr. Dai Weimin commented on those companies that misjudged the market and missed opportunities.

It is based on rational and calm thinking about objective facts that VeriSilicon chose such a correct market development route, first entering Taiwan, then taking over the North American and Japanese markets, and finally returning to the gradually mature and prosperous mainland market, growing bigger and stronger step by step, becoming a new technological upstart that subverts the chip industry, and thus guiding the current mainstream model of chip development.

Since its development, VeriSilicon has accumulated many advantages in chip design platform as a service, which is also one of VeriSilicon's main sources of income. In terms of design capabilities alone, VeriSilicon has accumulated more than 2,000 14nm FinFET wafers in mass production, and has expanded to the world-leading 10nm FinFET and the 28nm and 22nm FD-SOI process technologies that are taking off. One chip is taped out every week, and more than 50 chips are taped out every year, of which 98% are successfully taped out at the first try.

Currently, design service companies on the market are mainly divided into the following categories:

There is a type of design service company that makes a living by providing design services without having a clear idea of ​​what product to produce at the beginning. Once they find a good field or a good product one day, they will turn around and become a product company.

Another type of design service company was originally a product company, but unfortunately its product failed. However, it has accumulated certain experience in design and temporarily provides design services while waiting for the right time to develop products. These two types of design service companies only use design services as a stopgap measure.

The third type of design service company sells labor and sends its employees out to help clients with design. They have neither offices nor EDA software. It is difficult for employees of such companies to have a sense of belonging, let alone a personal growth path.

VeriSilicon is different from the three design service companies mentioned above. From the beginning, it has been very clear about its positioning: service is service, product is product . VeriSilicon focuses on providing customers with customized and high-quality design services, does not make products, and adheres to the bottom line of not competing with customers. This is the original intention of VeriSilicon.

As a foundry service company, it should be able to "endure loneliness and withstand temptation". Dr. Dai Weimin mentioned that some of the star chip products recently released on the market have attracted much attention, and VeriSilicon has also contributed to this, including providing IP and long-term technical support. However, as an unsung hero behind the scenes, VeriSilicon has been silently guarding behind the success of its customers. In fact, Microsoft XBOX motion sensing game Kinect and every iPhone and Samsung mobile phone have chips designed by VeriSilicon, but you can't see VeriSilicon's name.

Looking at the development history of the chip industry over the past few decades, the emergence of design outsourcing is actually inevitable. Early IC products were systems on a board, with standardization as a prominent feature, and professional fabless and manufacturing outsourcing were born. However, with the advancement of chip design and technology, it has evolved into a system on a chip, with high specialization and systematization as a prominent feature. The design business volume has expanded unprecedentedly, and design outsourcing has gradually become the mainstream. As chip suppliers move up, their core competitiveness lies in systems and architectures. At least half of their engineers should work on software, and IP should become a common technology that can be licensed, and design implementation can be outsourced. Currently, the R&D expenses of chip suppliers account for 25%-30% of their sales, which means that the gross profit must be above 50%; when the gross profit drops to around 40%, this product department may be sold, such as ASR's acquisition of Marvell's baseband department; or it may be closed, such as Avago's closing of its Shanghai HDD department. If you want to preserve the product line, you can either license IP and outsource the design, or form a joint venture, such as Datang Telecom, Jianguang, and Zhilu Capital, which joined hands with Qualcomm to establish Guizhou company Lingsheng Technology. This will become the norm and will be a win-win situation, so there is no need to over-interpret it.

At the same time, as system companies move downstream, they sometimes customize chips and outsource design in order to differentiate themselves and reduce costs.

"Chips in the past were different from those today. A board or a box in the past is now a chip, or a system on a chip. In addition to making chips, chip companies now also have to provide middleware and software. This situation has caused chip manufacturers to do the work of system companies. So what do system companies do? They can only do OEM. When such companies want to launch one or several chips a year, in order to save costs and development time, the best way is to find a foundry design company to design them." Dr. Dai Weimin said when talking about business model innovation.

"After all, there aren't many companies like Apple that can afford to support a team of thousands of people to design chips, which has led to a sharp increase in market demand for foundry design companies like VeriSilicon. VeriSilicon's performance growth in recent years also reflects this point from the side."

From Fabless to Design-Light

So should a design service company develop IP? Because IP development requires long-term investment. It takes many years and a lot of money to develop IP, and it may not be successful in the end.

However, in the eyes of Dr. Dai Weimin, IP is equivalent to the unique kitchen and bathroom technology when building a house. If the kitchen and bathroom technology is good, customers will come to you to build a house again. IP occupies a large proportion in the eyes of customers. If it is used well, it is difficult to change to a new company for redesign. Therefore, over the years, VeriSilicon has always attached great importance to the early investment in IP, established IP moat, and become a first-class IP Power House.

VeriSilicon is committed to improving its own platform and continuously expanding its investment and accumulation in IP. Through years of R&D accumulation and corporate mergers and acquisitions, such as the acquisition of LSI ZSP (DSP, digital signal processor) department in 2006 and Vivante graphics processor (GPU) company in 2016, VeriSilicon has realized a complete set of intelligent pixel processing technology platforms, including internationally leading GPU graphics processing, visual artificial intelligence processing, high-end video encoding and decoding, ISP image signal processing, DSP intelligent voice and audio processing, data compression, display control, etc. VeriSilicon cooperates with world-class companies to establish our technological leadership, and introduces domestic companies to help domestic companies improve product competitiveness, achieve international technology leadership, and implement it domestically.

At the same time, VeriSilicon chooses to deepen its market fields that are suitable for the company's long-term development. Dr. Dai Weimin said that VeriSilicon mainly targets automotive electronics, monitoring, Internet of Things, communications, intelligent computing, AR/VR and other fields, and has accumulated certain core IPs and built a relatively complete platform, namely Silicon Platform as a Service, SiPaaS®, to serve more customers. For example, in the field of automotive electronics, VeriSilicon's graphics processor GPU IP is widely used in automotive digital instrument panels (world's first), in-vehicle infotainment systems (IVI) (world's second) and rear seat entertainment systems (world's third). Seven of the top 10 automotive OEMs in the world have used VeriSilicon's graphics IP for their in-vehicle infotainment systems, and six of the top 10 luxury car brands have used VeriSilicon's graphics IP for reconfigurable instrument panels, occupying an absolute leading position in the market.

VeriSilicon Automotive Electronics Platform

"The automotive market is different from the mobile phone market. It is not easy to enter and even harder to exit." Dr. Dai Weimin described the automotive market in this way. Due to the higher requirements for safety in the automotive market, when VeriSilicon has established a relatively complete platform in this market, it is difficult for other companies to compete with it.

Another important choice for design service companies is whether to tie up with a wafer foundry. This issue has been debated for more than ten years within VeriSilicon. Dr. Dai Weimin said: "Although some wafer foundries do not do design, they may hold a controlling stake in a design service company and appoint a chairman. This is a service model that ties customers to wafer foundries."

Since its inception, VeriSilicon has provided standard cell libraries for China's local IC foundries, objectively contributing to the vigorous development of China's fabless chip companies. Over the past decade, VeriSilicon has designed more than 50 sets of standard cell libraries. In addition to providing 22 sets of libraries to SMIC and 16 sets of libraries to Macronix, it also provides standard cell library design and support for Huahong, Shanghai Advanced, Huali, and China Resources Shanghua. This is the reason why VeriSilicon adheres to the fab-neutral strategy.

Due to the deep relationship between VeriSilicon and SMIC, when the device model of the initial process was not yet complete, VeriSilicon made a huge bet to help it develop a standard cell library, and then revised it again, just to speed up the product launch. Logically speaking, such revolutionary comradeship would easily make it easy to become the designated partner of the foundry.

However, VeriSilicon clearly realized that if it tied to a wafer foundry, although it would be easy to boost performance, it would form dependence and be subject to the wafer foundry's own technical level and production capacity; if it did not tie to a foundry, VeriSilicon would be able to strengthen its technology in many different processes, prepare for battle, and finally accumulate a wide range of technical advantages, and customers would come to it. Moreover, in many cases, it is the customers who choose the wafer foundry. So this is a question of position: should we help the wafer foundry bind customers, or give customers more and better choices?

VeriSilicon will select the appropriate wafer foundry and the best process based on product needs and customer requirements, but the company will never be tied to anyone. This provides customers with a lot of choice and design flexibility, allowing customers to achieve the best match between product performance and wafer foundry process characteristics.

We insist on providing services instead of products, attach importance to IP and make long-term investments, and are not tied to any wafer foundry. Although it was not easy to make decisions on these major issues at the beginning, looking back today, it should be said that every choice was right, and this model of VeriSilicon is destined to succeed.

Not long ago, VeriSilicon received a strategic investment of RMB 200 million from the National Science and Technology Investment Fund, which positioned VeriSilicon as an "IP Power House". In the eyes of Dr. Dai Weimin, this is a great affirmation of VeriSilicon and its business model. He believes that "the next five years will not only be the golden five years for domestic semiconductor companies, but also the golden five years for VeriSilicon".

"The golden five years means that if you want to go public, it is best to do so within these five years." In recent years, the country has increased its investment in the semiconductor industry. In the next five years, Xinyuan must seize development opportunities, closely connect with the upstream and downstream of the industrial chain, win the greatest development, and ultimately achieve the goal of going public. This is the primary goal that Xinyuan needs to achieve in the next five years.

Dr. Dai Weimin also said that the Big Fund favors VeriSilicon in part because of VeriSilicon’s contribution to the entire semiconductor industry ecosystem.

In recent years, many overseas talents, especially those with technical expertise, have returned to China to start a business, but often encounter some difficulties. This is like the relationship between being a chef and opening a restaurant. Many chefs say, I cook all the dishes, why is the restaurant not mine? In fact, there is a difference between cooking and opening a restaurant. Many chefs cook good dishes, but fail to open a restaurant in China. The chip I participated in designing sells well in the United States, why doesn’t it work after returning to China? That’s because he doesn’t know how much sales and marketing investment is involved. After such people return to China, they will inevitably have problems with acclimatization due to their unfamiliarity with the market. At this time, someone needs to guide and help them. VeriSilicon can play this role to try to avoid the failure of such companies and help them avoid detours.

As a design-driven company that values ​​IP design, VeriSilicon will plan to acquire more IPs to serve customers in the future. VeriSilicon does not produce products itself, and the main purpose of acquiring IPs is to build a better platform to serve more domestic and foreign customers. Dr. Dai Weimin emphasized: The pattern determines the outcome, and only with a broad mind can you benefit the world. VeriSilicon's mission is to strive for the "rising tide lifts all boats" of the entire industry ecosystem, rather than "the truth comes out". VeriSilicon improves the overall level of the entire chip industry through services and IP, and does not approve of blindly engaging in price wars to destroy the industry ecosystem, which will not benefit all companies in the industry in the long run.

"The National Fund does not invest much in the design field, only a limited number of companies, especially in strategic investment. The fact that VeriSilicon has been favored is closely related to our long-term accumulation of IP and close connection with the upstream and downstream of the industrial chain, and our ability to serve as a bridge between wafer fabs and chip companies and system companies."

Nowadays, it is becoming increasingly difficult for Chinese capital to acquire foreign capital, especially technology companies. "Buying is one thing, but more importantly, we need to be self-reliant and do our own research and development."

The significance of the Big Fund is to help the entire industry chain establish a healthier and more sound development platform to make up for the shortcomings of the industry chain. With the support of the Big Fund, VeriSilicon will continue to focus on deepening IP and design services and jointly promote the development of China's integrated circuit industry.

The Chinese semiconductor industry often discusses "overtaking on a curve", but where is the "curve"? Sometimes it may not be very clear. Dr. Dai Weimin mentioned that FD-SOI technology is a historical opportunity for China's integrated circuit industry to achieve "overtaking on a curve".

Semiconductors are a very special industry. "In the course of more than 50 years of development, no other industry has been as small as semiconductors, with faster speeds, lower power consumption, and lower costs. If the automotive industry could develop at the same speed as the semiconductor industry, we could buy a car for just a few dozen yuan today, which is unimaginable. However, after the 28-nanometer process, costs have risen instead of falling, and a technological turning point has appeared, like a car's curve," said Dr. Dai Weimin.

More than a decade ago, industry insiders had already seen the technical difficulties that would arise after 28 nanometers, so three professors from the University of California, Berkeley, Chenming Hu (Hu Zhengming), Tsu-Jae King-Liu, and Jeffrey Bokor, studied and proposed two solutions to expand CMOS process technology to below 20 nanometers, one is the three-dimensional structure of FinFET (fin field effect transistor), and the other is the planar structure of FD-SOI (fully depleted silicon on insulating layer). Next, each wafer manufacturer made its own choice between these two routes.

In Dr. Dai Weimin's opinion, there are two main reasons why FinFET has become the mainstream trend of advanced processes today. More than ten years ago, Intel was the leader in wafer manufacturing, and FD-SOI did not meet the needs of mass production at the time, and some core patent technologies were monopolized by a French manufacturer. In order to avoid being constrained by others, Intel chose the FinFET structure for research and development, and TSMC chose to follow Intel's footsteps to avoid risks. Secondly, the FinFET process is very suitable for Intel's high-performance and high-density product positioning at the time. It is based on such historical reasons that Intel, TSMC and other process giants have chosen the FinFET technology route.

VeriSilicon has a clear understanding of the layout of FinFET and FD-SOI process technology routes: TSMC's 7nm FinFET will be a main node with a long life cycle, while 10nm FinFET is a short transition. Samsung will stay at 10nm for a longer time, waiting for EUV lithography technology to mature before making 7nm FinFET. Intel announced that its 10nm PC processor chip will be mass-produced in 2018, and claimed that its 10nm level is equivalent to TSMC and Samsung's 7nm. GlobalFoundries may be a dark horse in the 7nm process. So between the current mature mass-produced 28nm bulk silicon with a better cost-effectiveness and the relatively expensive 7nm FinFET, which advanced process technologies will show their prowess? There is no doubt that FD-SOI process (Samsung's 28nm FDS and 18nm FDS, GlobalFoundries' 22nm FDX and 12nm FDX) will definitely have a place.

When emerging fields such as the Internet of Things emerged, they put forward different performance and power consumption requirements from previous products, which made it difficult for FinFET technology to be perfectly applied to the vast market such as the Internet of Things, especially for products with high power consumption and cost requirements. FD-SOI can well meet the application of low power consumption requirements. If Intel and TSMC chose FinFET for historical reasons, then the FD-SOI process driven by the application side will be a popular choice in the future.

First of all, in applications such as the Internet of Things, 5G, and MMIC, there are many RF circuits that are very sensitive to the process, and the FD-SOI process has a very big advantage. Compared with FinFET, the cutoff frequency Ft/Fmax of FD-SOI is higher, which can support higher RF bands, including millimeter wave bands, and has better RF circuit matching performance. Once the GPS of a mobile phone using the bulk silicon process is turned on, the battery usually drops very quickly. The new watch launched by Huami not long ago includes a GPS designed by SONY and manufactured by the FD-SOI process, which is very power-saving and has become one of the most eye-catching highlights of the product. TSMC released the 22nm planar bulk silicon process in March this year and plans to mass-produce it next year, mainly for RF, especially in 5G, mobile devices, wearables and the Internet of Things.

Secondly, the FD-SOI process supports body biasing technology. IoT devices are different from data centers and may have a long standby time. High performance is required only for 10% or at most 20% of the time. In the FD-SOI process, body biasing can be controlled by software to adapt to different working scenarios. In addition, body biasing can reduce the voltage of the RF front-end circuit, save power consumption, and can also correct the PVT deviation in actual manufacturing later, reducing the difficulty of sign-off.

Dr. Dai Weimin sees the Chinese market as the main driving force for FD-SOI. He believes that now is a good time for China to invest in FD-SOI and use it as a low-power process alternative.

Dr. Dai Weimin pointed out, "Switching from traditional bulk silicon to FD-SOI is as simple as changing a track. However, if you keep squeezing on the FinFET track, it will be very difficult to catch up with the existing leaders on the straight track." Dr. Dai Weimin and Academician Wang Xi, Director of the Shanghai Institute of Microsystem and Information Technology of the Chinese Academy of Sciences, co-chaired the Shanghai FD-SOI Forum and have held four consecutive sessions in Shanghai. Several milestone historical events in the global FD-SOI industrialization started from the docking at this forum. VeriSilicon has cooperative relationships with three FD-SOI foundries, GlobalFoundries, Samsung and STMicroelectronics, to help customers design ASICs and dock with FD-SOI processes.

Dr. Dai Weimin and Academician Wang Xi attended the GlobalFoundries project signing ceremony

China's semiconductor industry is in a special environment. In order to be self-reliant, it is undoubted that while developing FinFET technology, it also needs to promote the FD-SOI industry. In February this year, GlobalFoundries launched a 12-inch wafer manufacturing project in Chengdu with a total investment of US$10 billion. The first phase of the project is a mature 130nm and 180nm process, and the second phase is an advanced 22nm FD-SOI process. After completion, the annual production capacity will reach 1 million pieces. A Chinese FD-SOI ecosystem is taking shape: Shanghai Simgui's substrate wafer manufacturing plant, GlobalFoundries Chengdu FD-SOI wafer plant, and VeriSilicon's FD-SOI IP and design service center. "China should ' walk on two legs ' and develop FinFET and FD-SOI at the same time; in particular, it should seize the FD-SOI, which is a historical opportunity for Chinese wafer factories to 'overtake on the curve' and Chinese fabless semiconductor manufacturers to 'change lanes and overtake'," said Dr. Dai Weimin.

GlobalFoundries Chengdu 12-inch wafer fab

The special environment of China's semiconductor industry is also reflected in the fact that in recent years, the trend of mergers and acquisitions in the global semiconductor industry has intensified, and data from multiple sources show that the gross profit margin of the hardware industry has been declining year by year.

In Dr. Dai Weimin's opinion, the reason for the decline in gross profit margin is that, on the one hand, many companies have too many production lines, their advantages are not concentrated enough, and they are dragged down by non-core businesses; on the other hand, although the smart hardware industry has been booming in the past two years, not many companies have invested in it. The reason is that the hardware industry has a long profit cycle.

So how to achieve a breakthrough in the hardware field?

Dr. Dai Weimin said that for chip suppliers, the first priority is to connect with system manufacturers to ensure that the designed and manufactured chips have a terminal application market. Therefore, seven years ago, VeriSilicon joined hands with the Integrated Circuit Design Branch of the China Semiconductor Industry Association, the Dongguan Municipal People's Government and other units to jointly promote the "Songshan Lake China IC Innovation Summit Forum". "We first surveyed system manufacturers to find out what chips they need most and which domestically produced chip is of greatest significance to you. System manufacturers often answer that it does not have to be the chip with the most cores, the largest, or the most complex. Sometimes it is just a very small chip, such as fingerprint recognition, Bluetooth headsets, etc. After we collect the needs of system manufacturers, we will look for who has such a chip.

At this year's "7th Songshan Lake China IC Innovation Summit Forum", Professor Wei Shaojun, Chairman of the IC Design Branch of the China Semiconductor Industry Association, Director of the Institute of Microelectronics of Tsinghua University, and Chief Engineer of the National Science and Technology 01 Major Project of the National High-tech and Basic Research Institute, pointed out in his speech that the Songshan Lake China IC Innovation Summit Forum is known as the "most down-to-earth forum" in the industry, and the number of fabless IC manufacturers introduced by Songshan Lake has grown to 49. In the past six years, the Songshan Lake Forum has recommended 40 products to the market, of which 36 have been mass-produced, and the total mass production rate of the recommended products is as high as 90%.

The 7th Songshan Lake China IC Innovation Summit

On July 14, 2017, the first "Qingcheng Mountain China IC Ecosystem Summit Forum" initiated by VeriSilicon was held in Qingcheng Mountain, the birthplace of Taoism in China, to create a complete industrial chain closed loop from terminal application/service platform to IC software and hardware design, and then to wafer manufacturing and packaging testing. The forum was hosted by the Integrated Circuit Design Branch of the China Semiconductor Industry Association and co-organized by VeriSilicon Holdings Co., Ltd., and was strongly supported by the Sichuan Expo Affairs Bureau, Sichuan Provincial Economic and Information Commission, Chengdu Municipal Economic and Information Commission and other units. The difference between the Qingcheng Mountain Forum and the Songshan Lake Forum is that the latter only focuses on part of the ecosystem: chips and systems, but not limited to the scope of chip applications; while the former focuses on the entire ecosystem, but focuses on one application scope each year. This forum is dedicated to China's artificial intelligence industry, especially AI terminal chips, and the next session is planned to build China's automotive electronics industry chain. After the meeting, Dr. Dai Weimin wrote the following message for the Sichuan Convention and Exhibition website: "Discussing Taoism under Qingcheng Mountain, the AI ​​Avenue is invisible; Landing next to Dujiangyan is natural, and the IC ecosystem is promising."

Qingcheng Mountain China IC Ecosystem Summit Forum

In addition, as a chip company, we should seize the opportunities and challenges of semiconductors in new fields , such as artificial intelligence.

VeriSilicon's intelligent vision processor uses a 28-nanometer process and only takes up less than 3 square millimeters of area to achieve large-scale artificial intelligence applications, such as high-speed vehicle and pedestrian recognition, facial recognition, and trajectory tracking, and is widely used in assisted driving systems and driver monitoring systems. Such terminal AI processing capabilities will become the standard for future SoC chips. "Because many companies now pre-build artificial intelligence-related modules in their products, even if they are not used now, as the technology matures, they may become standard for products one day in the future, and the chips will be of great use. Therefore, VeriSilicon now helps customers reserve relevant AI modules in the chip, which can be directly activated through the software upgrade system when needed, thereby providing more value."

In Dr. Dai Weimin's view, intelligence is not the highest realm, but wisdom is the ultimate pursuit . "There is a big difference between intelligence and wisdom. Intelligence only has certain advanced functions, while wisdom requires machines to be able to self-learn and adapt, which is a higher level of experience."

Finally, Dr. Dai Weimin also expressed his views on the recent boom in investment and factory construction.

Taking the development of Taiwan's semiconductor industry as an example, Dr. Dai Weimin believes that the success of Taiwan's semiconductor industry is mainly attributed to two aspects. First, it is inseparable from the strong investment and support of the Taiwan government. Second, the geographical limitations of Taiwan have made the distribution of semiconductor manufacturers very dense, mainly concentrated in the Hsinchu area. The emergence of clustering is very conducive to the development of the semiconductor manufacturing industry, so when selecting a site, it is necessary to be relatively concentrated.

It is necessary to attach importance to the development of semiconductor manufacturing, but this does not mean unconditionally and blindly introducing external companies and technologies. "We cannot 'make chips' like the steel-making in 1958. We occasionally see signs of such blind development in China, and this is something we should be vigilant about."

"The key lies in chip design. After the money was invested, the factory was built and the production line was put into operation, but the design problem was not solved and the chips still needed to be imported from abroad. In the end, the factory was just making wedding clothes for others."

Only by being able to endure loneliness and temptation can we run a good business. Only by rationally and objectively finding the real entry point of China's semiconductor industry and looking for good development opportunities can we achieve development. Only by neither blindly introducing nor blindly excluding foreign countries can we promote industrial progress. "Made in China is all over the world, and Chinese design can also do it. After the two are integrated, there will inevitably be amazing explosive power. This is my expectation for VeriSilicon and my expectation for China's semiconductor industry!" said Dr. Dai Weimin.