The Semiconductor Industry at a Crossroads: Finding New Growth Drivers in the Age of AI, Digital Twins, and Carbon Neutrality
The semiconductor industry is at a crossroads in search of new growth drivers.
”It must be acknowledged that ‘Moore’s Law’, in which chip performance doubles every two years, is slowing down. However, the pace of technological innovation has not slowed down. As the expansion capability of a single chip slows down, the semiconductor industry is looking for other innovative methods to maintain the index. “Grade growth.”
Sassine Ghazi, president and chief operating officer of Synopsys Technologies (Nasdaq: SNPS), summarizes the current development status of the semiconductor industry. He will serve as CEO of Synopsys Technologies on January 1, 2024.
The reason why the semiconductor industry has become an important driving force for the development of human society is due to the improvement in production efficiency brought about by chip performance, and at the same time, it has given rise to applications such as the Internet that have completely changed human life. Fundamentally speaking, performance is not the ultimate goal of chip innovation, but improving production efficiency and changing human life is.
Artificial intelligence, high-performance computing and the new energy revolution are becoming the key to changing human life, and these fields have put forward more innovative requirements for chips, in addition to performance, more complex functions, higher security, and faster delivery, Lower energy consumption and cost are urgent needs for chips in the new era. Moore’s Law, which is gradually slowing down, can no longer meet these requirements. Semiconductor innovation needs new impetus.
Synopsys is the world’s leading electronic design automation software EDA company and interface IP provider, and is also a leader in global software security companies. Standing at the upstream of the chip industry chain, Synopsys has discovered during its cooperation with various companies in the industry chain that digitalization and low-carbonization have become the development trends of the chip industry. Under the new trend, the new driving force for chips to maintain exponential growth will be From three aspects: 1. Improved efficiency of the entire chip industry chain driven by artificial intelligence; 2. AI+ digital twin technology promotes the accelerated implementation of chip innovation; 3. Chip innovation in the carbon neutral era will have a broader boundary.
EDA+AI, a multiplier for the efficiency of the entire chip industry chain
The innovation model of the chip industry in the past few decades is very clear. Due to the existence of Moore’s Law, chip performance continues to grow regularly, and developers develop applications based on chip performance. Chips are the starting point of the development chain. There are chips first, then new applications, and then new electronic products.
However, as consumer demand is stimulated by the rapid growth of chip performance, Moore’s Law begins to slow down again, and this model that worked well in the past begins to gradually fail.
However, the semiconductor market has not slowed down. The Internet of Everything has expanded the scope of chip applications, and artificial intelligence has created demand for high-performance computing. The semiconductor market continues to grow at a rapid pace. It took the global semiconductor market 60 years to reach US$500 billion, but Synopsys predicts that the next US$500 billion will only take seven years, that is, the global semiconductor market will reach US$1 trillion in 2030.
GPUs and NPUs that support artificial intelligence, low-power communication chips and functional chips that support the Internet of Everything, etc. Customers need more complex functions, higher energy efficiency, faster delivery, and lower costs. To meet these needs, the overall chip must be improved. efficiency of the industrial chain.
Synopsys took the lead in introducing artificial intelligence into chip design four years ago and launched DSO.ai, the first AI application for chip design. It has been adopted by many leading chip companies around the world and has been successfully taped out more than 260 times. The scale of AI in chip design applications become a reality.
In early 2023, Synopsys upgraded and expanded DSO.ai into a full-stack AI-driven EDA solution, Synopsys.ai, covering the design, verification, testing and manufacturing of advanced digital and analog chips. AI has improved the automation and quality of EDA tools, reduced the amount of procedural work in chip design, and significantly reduced the manpower and time required to complete the same work.
Sassine said, “Artificial intelligence-driven chip design was pioneered by Synopsys. We see the potential of artificial intelligence to improve chip manufacturers’ productivity. Customers who use our artificial intelligence EDA solution have reported that the design time has been shortened from several months to a few.” Weekly, the test cost has been reduced by more than 20%, while the performance and power consumption of the chip are better.”
Innovation Accelerator—AI Digital Twin
The rapid expansion of application scenarios is an important difference between current chip innovation and the past. Artificial intelligence + the Internet of Everything continue to create new application scenarios. In these scenarios, chip innovation requires continuous trial and error. If it is tested with physical objects, it will cause a lot of time and The waste of resources and high-simulation virtual environment have become a necessity for chip innovation.
Virtual technology has long been used in chip innovation. However, previous virtual technology had many problems such as incomplete parameter coverage, limited virtual conditions, and low test simulation. Even if virtual technology developed to the digital twin stage a few years ago, the simulation still could not achieve the goal. People were satisfied. It was not until the introduction of artificial intelligence that AI digital twins made high-simulation virtual environments possible.
AI digital twins are fully used in fields such as artificial intelligence, the Internet of Everything, and new energy revolution. Among them, the application value is particularly outstanding in new energy-related industries. These industries are closely connected with reality and have greater and more complex environmental impacts, such as smart cars, wind energy, etc. , photovoltaics, are all at the forefront of digital twin technology applications.
Take smart cars as an example. The demand for semiconductors in automobiles has exploded in recent years, and automobile chip innovation presents the three characteristics of “large quantity, many types, and many functions”, making automobile chip innovation a complex and typical current situation. Scenes. Taking the amount of software code as an example, the current amount of software code for smart cars is about 100 million lines, and will increase to 300 million lines in 2030. For comparison, Synopsys, which has been deeply involved in the software industry for many years, has a total code amount of about 300 million lines.
Sassine said, “In order to cope with the complexity of software-defined cars, Synopsys’ virtual ECU technology moves testing and verification from prototype vehicles and physical test platforms to virtual environments. This verification method can significantly improve test simulation and safety, Reduce risks and detect errors early, thereby reducing development time and costs.”
New technologies have indeed improved the efficiency of vehicle development. According to feedback from car companies, the development cycle of smart cars has been shortened to three years, which is more than half the time of traditional fuel vehicles. Among them, AI digital twins play a key role.
Especially in the vehicle development process, AI digital twins can be called “cheating devices.” Traditional vehicle testing involves manufacturing multiple prototype vehicles and testing multiple subsystems simultaneously. But now it is possible to conduct simulation tests on the digital twin of the entire vehicle several months in advance, which is equivalent to opening an additional account in advance. The advantage is comparable to a “cheat device”, saving time, increasing efficiency, and reducing costs.
In the process of verifying the simulation test results of the prototype vehicle, the simulation test and verification data can also be continuously compared to optimize the digital twin simulation degree, further improve the reliability of the simulation test, form a virtuous circle, and significantly improve the efficiency of vehicle development.
Similar to smart cars, highly complex systems such as artificial intelligence, high-performance computing, and multi-chip packaging require the support of AI digital twins. In addition to accelerating the implementation of chip innovation, it can also conduct low-cost trial and error and explore more innovation possibilities.
Carbon neutrality opens up the boundaries of chip innovation
The goals of semiconductor practitioners in the past few decades have been very clear. They are looking for various ways to continue Moore’s Law. Extreme performance is the ultimate pursuit. But after Moore’s Law slows down, carbon neutrality may become an important driving force in opening up the boundaries of chip innovation.
First of all, in order to be carbon neutral, chip energy efficiency needs to be improved. Sassine said, “Power consumption, performance and area are the three key indicators of chip design. Currently, power consumption is becoming the ultimate limit for advanced technologies such as artificial intelligence, high-performance computing and smart cars. Factors. Coupled with people’s increasing concern about the impact of energy consumption on the climate, reducing chip power consumption is becoming key.” Ge Qun,
global senior vice president of Synopsys and chairman and president of Synopsys China, took ChatGPT training data as an example to explain the advanced The energy consumption level of the chip: GPT-3 has 175 billion training parameters, and a single training consumes 1,287 megawatt hours, which is approximately equivalent to the daily electricity consumption of 250,000 Chinese households. The next-generation GTP-5 will have 100 times the parameter volume of GTP-3, the calculation amount will soar to 200 to 400 times, and the power consumption of training may be hundreds of times that of GPT-3. The figures are staggering.
Behind computing power is electricity, and low-carbon green is becoming a challenge facing the semiconductor industry. To address this challenge, Synopsys has launched an end-to-end low-power EDA and IP solution that covers the complete process from architecture, RTL, implementation to signoff. Its Synopsys.ai solution can shorten the time by 25 days with the help of artificial intelligence technology. % of the design cycle, reducing energy consumption by 30%, and reducing chip power consumption by an additional 25%.
In addition to reducing chip energy consumption, Synopsys is also exploring the application of chip scientific knowledge and methods to broader scenarios. Ge Qun believes: “To achieve carbon neutrality, the key is to ‘decarbonize electricity’. On the one hand, we must change the energy structure and increase the proportion of clean energy; on the other hand, we must do a good job in energy conservation and emission reduction in key areas. Both aspects require Rely on the power of semiconductor innovation.”
Efficiently deploying energy has been the focus of chip scientific research in the past few decades, and it is also the direction in which chip innovation can make a greater contribution to carbon neutrality. Develop more advanced and energy-saving chips to empower the carbon neutralization process of thousands of industries; help energy conservation and emission reduction through “computing power decarbonization”; actively participate in the construction of energy networks, and use chip scientific knowledge and methods to help new energy resources better Integrating into the energy system will fundamentally help “power decarbonization”.
While promoting chip innovation with the goal of energy efficiency, and expanding the application of chip scientific methods in other fields, carbon neutrality is opening up new boundaries for chip innovation.
Whether it is EDA+AI improving the efficiency of chip innovation, digital twin technology accelerating the implementation of chip innovation, or carbon neutrality opening a new boundary for chip innovation, innovation has always been the core driving force for the progress of the semiconductor industry. In the past, the direction of innovation was to tirelessly pursue extreme performance following Moore’s Law, but now the direction of innovation tends to be diversified. Chips will penetrate into every corner of life, and the knowledge and methods accumulated in chip science over the past few decades will be extended to more fields. , looking for a new “Moore’s Law” and making greater contributions to the progress of human society.
