In the first part of the global semiconductor industry outlook (1), the market size of the semiconductor industry was briefly introduced, AI has been hot in recent years, and generative artificial intelligence has also pushed the semiconductor industry to a new stage of development. According to industry analysis, AI-related applications will contribute about 10% of global semiconductor sales growth in 2024.
Based on the development of the market, Deloitte predicts that artificial intelligence may bring new changes to the semiconductor industry:
New manufacturers
Artificial intelligence gold rush, AI chips are in short supply. Nvidia, the largest AI chip company on the market, has captured 80% of the market share, and AMD's sales soared after the launch of the MI300X chip last year, which is expected to exceed $3.5 billion. While some industry giants are seizing the AI chip market, AI chip startups have also sprung up like mushrooms after a rain, grasping the tuyere of AI chips and continuously launching competitive AI chip products. For example, Etched, which is currently preparing to design the Sohu chip, has raised $120 million.
New models and architectures
New architectures and models for AI chips are moving towards greater efficiency, flexibility, and lower power consumption. Now, some new progress has also been made: a research team from the University of Pennsylvania and Nokia Bell Labs in the United States has proposed a new AI chip that uses light waves to perform complex mathematical operations to increase processing speed and reduce energy consumption. This technology takes advantage of the high-speed transmission and low energy consumption of light waves, and is expected to solve the energy consumption problem of current chips.
Edge chips
AI chips can be divided into two categories: cloud and edge according to their deployment location and application scenarios. Edge chips are chips designed for edge computing and edge AI applications, emphasizing solving problems on the data source side to meet real-time decision-making and processing needs. Edge chips have a wide range of applications in many fields, such as smart manufacturing, smart security, smart home, smart transportation, smart city and other fields. According to Gartner statistics cited by Dexin Semiconductor, the size of China's edge AI chip market in 2022 will be about US$4.99 billion, and it is expected to grow to US$11.03 billion by 2025, an increase of 121% over 2022. Globally, the edge AI chip market is expected to reach $68.8 billion by 2026. With the rapid development of technologies such as the Internet of Things, intelligent manufacturing, and AR/VR, the demand for edge computing chips will continue to grow.
Figure: market size of global edge chip
The whiplash effect
When the supply is insufficient, the chip industry is very prone to insufficient inventory. Inventory accumulation and supply-demand imbalances can lead to large fluctuations in the prices of semiconductor products. When there is a surplus of inventory, suppliers will adopt a price reduction promotion strategy in order to destock; And when there is a shortage of inventory, suppliers will raise prices to make higher profits. The "long whip effect" not only leads to problems such as inventory accumulation and price volatility, but also increases the overall risk to the supply chain. Due to the delay and distortion of information transmission, it is more difficult to collaborate among all links in the supply chain, which can easily lead to problems such as production delays and extended delivery times. The "long whip effect" will exacerbate the cyclical fluctuations and uncertainties of the semiconductor industry, affecting the stable development and sustainable development of the industry. In order to alleviate the impact of the "long whip effect", the entire industry needs to strengthen cooperation and collaboration to promote the digital transformation and intelligent upgrading of the supply chain.