October 08, 2020
2019 is the most sought after year of EDA in China for 20 years in my career. The strategic importance and importance of EDA have been pushed to unprecedented heights. The media has paid great attention. The companies in the industry are also sought after by capital. Every week, there are calls from strange investment institutions. Find me, this makes us practitioners who have been bowing our heads very unaccustomed.
As early as 2018 before Huawei was embargoed, I once talked about the competitiveness and competitive strategy of local EDA companies in IC think tank. The content in the video was also quoted by many media. The company and product competitiveness mentioned include competitive strategies. I think It is still established so far. If you are interested, you can go to IC think tank to look through the original video. The following picture is one of the core content of the video. It lists several relatively large-scale companies in China (which I personally know very well) and the locations of their main products. These companies still maintain a good development trend this year and their revenue While growing, new products and new technologies are introduced. In addition, two new entities have emerged this year supported by the efforts of all parties and capital. They have strong strengths in the directions of digital integrated layout, hardware simulation (Hongxin Micronano, S2C) and foundryEDA (full core intelligent manufacturing). The EDA ecosystem is becoming more and more perfect.
However, no matter the strategy is still stuck, EDA is still a part of integrated circuits, and the competitiveness of integrated circuits cannot be separated from the market. As a practitioner for many years, our focus is still on the market to make competitive products and technologies, so the focus of this article is to talk about the "domestic substitution" of EDA, talent issues, local company competitiveness and development strategies around the market. In addition, the above picture was originally based on a page where I discussed the DTCO method (DesignTechnologyCo-OpTImizaTIon) at an academic conference. I personally think that DTCO has a lot to do in China. In addition to market needs, Chinese companies have key formulas, so they are competitive as local EDA companies. The supplement, let’s talk about DTCO by the way.
Many are all, why are you stuck?
The world's leading EDA companies all claim to have the entire process, but none of them can truly realize the dominance of first-line customers. In addition to the market's need for competition, the core reason is that they can't make all the tools the best.
If you let the CAD department of a leading domestic design company do statistics, the most used tools (the ones that really get stuck) may be a few tools, so it doesn’t make sense to “complement” if you leave the market. The "domestic substitution" of EDA cannot be compared with the domestic substitution of chips. The users of chips are system manufacturers, and the best system manufacturers are in China, which can help design companies to iterate products quickly, and the market capacity is large enough.
However, the EDA user base and the best design and manufacturing companies are all overseas. The tools that really get stuck on the neck need to be iterated by advanced technology and design. This cannot be solved by money. At the same time, the local market is relatively small and several leading companies. EDA's sales in China account for only about 10% of its global revenue, which is much smaller than the proportion of mainstream chip makers in China.
So the one who really got stuck
1. EDA tools to solve advanced manufacturing and design. Local companies lack high-end design and manufacturing companies, and local companies lack product iterations for high-end users. To make breakthroughs, technology and R&D need to go overseas and enter the high-end user market. In the past, HiSilicon has begun to take the initiative With local EDA companies iterating products, local EDA manufacturers benefit the most not from orders, but from rapid product iteration and competitiveness enhancement, but there are still too few advanced companies that can help us iterate.
2. The global market share of EDA in China is not high, and the market capacity is not large. It is not easy to make money due to the presence of international competitors. Without the support of the market and profits, it cannot be sustained by policy support alone.
The shortage of talents is constantly mentioned-why there is a shortage
There is a shortage of integrated circuit talents, and EDA talents are even more lacking. Many people may have heard on different occasions or various statistics. However, there is no particularly good answer to why there is a shortage and how to solve the problem of talents. It’s basically back to the word “market”. Popular industries rarely complain about the lack of talents. China is not short of AI talents (I know a lot of EDA people who are engaged in AI). Because of the hot market, the core reason for the lack of EDA is China’s lack of talent. A good EDA company and market size are also our own problems for practitioners, and we cannot blame others.
To solve the problem of talents, we cannot rely solely on university education or return to the market. The current is a good time for our practitioners to develop rapidly. Under the premise of not violating fair competition, we can use our strength to develop and obtain returns and rapid expansion through the market and capital. Several high-quality EDA companies create new markets and expand the scale of their industries, so that talents come from themselves.
Competitiveness of local EDA companies
In 2018, I once said that local companies have first-class world competitiveness in the data terminal (parameterized testing), turning data into simulation input (modeling simulation) and circuit simulation (SPICE and memorySPICE) (for details, please refer to IC Think Tank Video ), especially the data testing and modeling end, is also rarely deployed by the world's leading EDA companies. These are also key factors for the implementation of the DTCO method.
The industry is familiar with the leading enterprise Huada Jiutian with more than 20 years of accumulation from the earliest panda system. In terms of data, modeling and simulation, Chinese companies have also accumulated very deep.
Taking modeling and simulation as an example, the global market share of EDA tools and services in this field by Geelun Electronics and BDCOM is more than 70%. Starting from the predecessor of BDCOM, Aikesaili, they have fallen in love for nearly 20 years. Since Professor Hu Zhengming created the BSIM compact model, it has undergone 3 generations of product technology accumulation to obtain its current industrial status.
Data terminals Guangli Micro and Boda Micro have maintained rapid growth in mass production and general parameterized test products. Rapid testing can provide massive data, which is also the basis for process development, design optimization and chip yield analysis.
The company's strength in RF, IPD and system simulation products is also increasing.
In addition to the two newly emerging companies this year with FoundryEDA and digital processes as their main competitiveness, the local EDA ecosystem is becoming more and more perfect, and the related products of these several companies are competitive in the international market.
DTCO has a lot to do in China
DTCO is the abbreviation of DesignTechnologyCo-opTImizaTIon. The core goal of this method is to help advanced Foundry achieve faster design-in by co-optimizing chip design goals and optimize PPA to improve yield.
This article focuses on concepts, so I will not go into the technical details. I will have a technical evolution on the landing of DTCO on the ASP-DAC on January 15, 2020. Those who are interested can pay attention.
DTCO is not a new concept. In advanced processes such as 7nm, high-end design companies such as ARM, Qualcomm will intervene in the very early stage of process development and iterate with Foundry.
Most Chinese design companies rely on EDA and process to obtain product competitiveness. Except for some high-end design companies, they cannot benefit from the performance and yield improvement brought by related methods. Therefore, if a real and effective DTCO method can be implemented at the EDA level, The value generated in the Chinese market is huge, and it truly solves the problems of China's integrated circuit industry:
1. It can help China's Fab/IDM speed up advanced process development and shorten TTM
2. Make the high-end mature products of Chinese design companies faster porting and mass production in the local foundry
3. Significantly shorten the process/design optimization cost and iteration cycle
4. Improve yield and reliability
China’s manufacturing and design market is growing very fast. In addition to strong demand, Chinese local EDA companies also have all the key recipes required by DTCO. At the same time, the correlation network and regression technology in recent years have become more and more perfect. System/chip/device/process The cost of establishing an inter-correlation network is becoming more and more feasible. Personally, I am very optimistic that the DTCO method will be implemented in China and become an EDA comprehensive solution that can generate value and make money. This is also a new market that we can create and lead.
A complete DTCOflow requires early-stage OPC, TCAD, mass testing, rapid device modeling and circuit simulation, and the Chinese ecology is basically complete. Take our own business that we are good at. Test solutions can quickly generate massive amounts of data. At the same time, any process adjustment/optimization (split) leads to device performance changes, we can use the unique bidirectional regression network (BDN) for automatic device modeling Based on the 50-core processor, the modeling of a single 14nm FinFET device can be completed almost instantaneously, that is, it provides instantaneous input for circuit simulation instead of waiting for engineers to manually take several weeks.
In the later stages of SPICE, FastSPICE, highsigma simulation, etc., the strength of local manufacturers is world-class. In addition, from the landing of DTCO products, the storage industry provides a very good location. Compared with large-scale SoCs, the changes in the back-end storage are relatively shrinking, and the completion process and the actual value cycle are also much faster. Large space.
In terms of alternative feasibility, compared with the whole process of analog or digital EDA, I personally prefer Memory's whole process of EDA, which has strong tool convergence, obvious market growth, complete local formulas, and some formulas are better than international leading EDA companies. Competitiveness.
Summary and industry outlook
China's EDA industry is developing vigorously, developing competitive products with international competitiveness, and products entering the first-tier international factories. Driven by local design/manufacturing factories, targeted complements, capital compliance and legal integration and introduction Advanced technology and team, as well as creating new product ecology and new markets based on the Chinese market, are the development roads that I recognize as a practitioner.
Finally, make a simple industry outlook for next year
1. There will be more than one leading company, profitable EDA companies with the potential to go public;
2. Although it is considered a "stuck neck", Chinese EDA will be more open, with more organic integration and more cooperation with international EDA giants;
3. There is no need for policies to catalyze the birth. The market-certified technology will eventually succeed, and more products will go to the international market.