Top 10 market and application trends in the electronics industry in 2024

After several years of trough, the global semiconductor industry will finally usher in a new round of upward cycle by 2024. In this issue, the analyst team of ESM-China has made trend analysis and market outlook on hot topics or fields such as generative AI, accelerated computing, intelligent driving NOA, automotive chips, wide bandgap semiconductors, memory chips, display panels, brain-computer technology, satellite communications, and chip distribution.

Trend 1: Generative AI is developing rapidly and moving further to the edge

2023 is undoubtedly a hot year for generative AI represented by ChatGPT and Stable Diffusion, and OpenAI’s corporate gossip can even become an absolute hot topic in this year. Google, the search giant, even felt an unprecedented “white terror” because of the large-scale application of the GPT model.
Many industry giants, including Meta, Baidu, Microsoft, Alibaba, etc., are scrambling to push their own large models, and Nvidia has specially added a Transformer engine for large model computing acceleration for this generation of GPU architecture, and various AI chip companies with large computing power have reversed the direction of market promotion in 2023, actively emphasizing the chip’s ability to large-scale training or inference of generative AI models.
Enabling generative AI to write code, do design, and create productivity has all staged one miracle after another in 2023. Even by talking to generative AI, generative AI does a lot of the work of designing the digital front end of the chip without writing a single line of code. At the 2023 NVIDIA GTC Fall Developer Conference, Huang demonstrated “Giving Generative AI a PDF of 2D Drawings of a Factory,” which delivers a physically accurate digital twin of an entire factory in a fraction of the time.
We believe that the future market value of generative AI will never be the model of ChatGPT charging fees for its services to the public. For example, in the early stage of chip design and digital twin construction, generative AI enters different industries and vertical market segments, which is the direction of its value creation. In 2023, some companies have already started to launch enterprise-grade generative AI solutions and services, and 2024 will be the key to the development of this model.
In addition to AI technology research, generative AI applications in vertical markets also mean that generative AI is not just running on the cloud, but continues to move to the edge – especially when there is more fine-tuning and market-based customization of models. Design, architecture, scientific research, healthcare, manufacturing, and other fields will all be profoundly transformed by generative AI.
Not just at the edge of the enterprise or data center, generative AI inference will play an increasingly important role in more devices, such as PCs, mobile phones, and even embedded applications. For example, Intel is about to promote the AI PC concept on a large scale, and MediaTek has also added a generative AI acceleration engine to the mobile phone AP SoC. In short, market participants in different locations are eager to get a piece of the new era of generative AI.

Trend 2: Accelerated computing continues to penetrate all walks of life

Especially with the gradual increase in the demand for accelerated computing in data centers, the position of general-purpose processors such as CPUs in the system architecture is no longer as absolute as in the past. The best argument to support this proposition in 2023 is that in the Q2-Q3 natural quarter of this year, NVIDIA’s quarterly revenue surpassed Intel (Nvidia FY2024 Q2 vs. Intel FY2023 Q2) for the first time – although this comparison may be related to the rapid development of AI, and from the enterprise level, this quarter is the trough of Intel in recent years, and it is also the peak of Nvidia.
Previously, ARK Investment predicted that the value of accelerators, including GPUs, ASICs, FPGAs, etc., in data center servers will grow at a CAGR (compound annual growth rate) of 21% over the 10-year period from 2020 to 2030. By 2030, the accelerator market will be worth $41 billion, while CPUs will be $27 billion. In other words, accelerators significantly compress the market space for CPUs. This is also the reason why the traditional CPU giants represented by Intel and Arm are working hard to promote the XPU heterogeneous strategy.
If we want to talk about the vertical market segment, let’s still take the semiconductor industry as an example: at present, the computing power supply of the data center of the semiconductor manufacturing foundry factory is still dominated by CPUs. But in reality, some of the jobs in semiconductor manufacturing, such as OPC image formation, require a lot of matrix multiplication — jobs that lend themselves well to GPUs or other accelerators.
At the NVIDIA Developer Conference in 2023, NVIDIA specially released the cuLitho acceleration library for the lithography step, which truly allows GPUs to participate in computing, and its performance and energy efficiency are dozens of times higher than that of traditional general-purpose computing processors. It is said that only 500 servers including GPU acceleration can do the work that used to be done by 40,000 CPU servers, and the data center occupies 1/8 of the space and consumes 1/9 of the power consumption of the previous one.
Over the past few years, these kinds of stories have played out in many areas. With the slowdown of Moore’s Law, the performance and energy efficiency dividends obtained by the manufacturing process are gradually not as bright as those in the earliest years of semiconductor technology, but the general trend of social digital transformation, energy conservation and carbon reduction still requires further improvement in performance and energy efficiency. To meet the market demand, it must be accelerated computing in addition to general computing, which is also the theoretical basis for our prediction.
In fact, the implementation of AI technology in all walks of life is essentially a form and manifestation of accelerating the development of computing. The supercomputer market will soon be dominated by accelerators, and HPC applications such as simulation, digital twins, and quantum computing have begun to apply accelerated computing on a large scale, and in a broad sense, supercomputing is part of the data center – the digital transformation of society and life will continue to apply accelerated computing to all walks of life.

Trend 3: Autopilot Navigation is a big game

Navigate on Autopilot (NOA), also commonly known as “pilot assisted driving” or “advanced intelligent driving”, is a combination of navigation and assisted driving, which is an autonomous driving assistance system based on vehicle sensors and high-precision map data, designed to help drivers drive more safely and efficiently on highways and urban roads.
According to the application scenarios, NOA can be mainly divided into high-speed NOA and urban NOA. At present, high-speed NOA has been implemented on a large scale, and urban NOA is entering a stage of rapid advancement. From January to September 2023, the high-speed NOA penetration rate of passenger cars in China was 6.7%, an increase of 2.5 percentage points year-on-year, and the urban NOA penetration rate was 4.8%, an increase of 2.0 percentage points year-on-year. For the whole year of 2023, the penetration rate of high-speed NOA will be close to 10%, and the urban NOA will exceed 6%.
The development of NOA in China began in 2019 when Tesla pushed the NOA function to Chinese users, and then new forces such as Ideal, NIO, and Xpeng also entered the game one after another to launch high-speed pilot assistance functions. At present, high-speed NOA has gradually become the “standard” function pursued by various car brands, and perception, regulation and control algorithms and product function definition have become the key to the quality of NOA function experience of various brand models.
Starting from 2023, the call for “heavy perception and light map” is getting higher and higher, and local manufacturers in China are mainly based on BEV+Transformer technology to optimize and upgrade the system perception ability to reduce the dependence on high-precision maps, thereby reducing costs and promoting the rapid implementation of urban NOA.
However, the diversification of all scenarios, especially the complex traffic conditions in China, has brought challenges to the development of urban NOA, and participating enterprises need to not only have algorithmic logic solutions to deal with complex scenarios, but also consider new technologies such as large models, multi-modal data, automatic annotation, and intelligent computing centers. In general, NOA is still in the early stage of the introduction of the consumer market, and it will not be able to reach commercial maturity in the short term, and local car companies still need to provide users with full and correct guidance and education while developing their functions.

Trend 4: Traction of key applications, and multi-point flowering of wide bandgap semiconductor applications

When the semiconductor industry is gradually entering the post-Moore era, wide bandgap semiconductors have stepped onto the historical stage and are regarded as an important field of “lane change and overtaking”. It is expected that in 2024, wide bandgap semiconductor materials represented by silicon carbide (SiC) and gallium nitride (GaN) will continue to be used in communications, new energy vehicles, high-speed rail, satellite communications, aerospace and other scenarios, and achieve rapid growth in the global application market.
The largest application market for silicon carbide (SiC) devices is in new energy vehicles, which is expected to open a market of tens of billions. The ultimate performance of silicon carbide substrate is better than that of silicon substrate, which can meet the application needs of high temperature, high voltage, high frequency, high power and other conditions, and the current silicon carbide substrate has been used in radio frequency devices (such as 5G, national defense, etc.) and power devices (such as new energy, etc.). And 2024 will be a big year for SiC expansion. IDM manufacturers such as Wolfspeed, Bosch, Rohm, Infineon, and Toshiba have announced accelerated production expansion, and it is believed that SiC production will increase by at least three times in 2024.
Gallium nitride (GaN) power electronics products have been widely used in the field of fast charging, and the next step is to further improve the operating voltage and reliability, continue to develop in the direction of high power density, high frequency and high integration, and further expand the application field. Specifically, the increasing use of consumer electronics, automotive applications, data centers, and industrial and electric vehicles will drive the GaN industry to grow by more than $6 billion.
Gallium oxide (Ga₂O₃) is being commercialized, especially in electric vehicles, power grid systems, aerospace, and other fields. Compared with the previous two, the preparation of Ga₂O₃ single crystals can be completed by a fusion growth method similar to that of silicon single crystals, so it has great potential for cost reduction. At the same time, in recent years, Schottky diodes and transistors based on gallium oxide materials have made breakthroughs in structural design and manufacturing process, and there is reason to believe that the first batch of Schottky diode products will be launched on the market in 2024.