In the face of the high demand for chips in the era of artificial intelligence (AI), this also makes the silicon wafers required to manufacture chips also face a strong demand. However, due to the complex and cumbersome process of producing silicon wafers that are pure enough to make chips, the supply of silicon wafers may be in short supply in the face of the subsequent recovery of the semiconductor industry and the increasing demand for artificial intelligence.
In order to meet the growing computing demand, AI chips are developing in the direction of larger size and higher integration. As the chip area increases, so do the requirements for the size and purity of silicon wafers. Silicon wafers are the basis of chip manufacturing, and their quality and size directly affect the performance and yield of chips.
The process of producing high-purity silicon wafers is extremely complex. From silica sand to silicon wafers, it needs to go through multiple purification and growth steps, and the purity needs to reach 99.99999999% to ensure that there are only a very few non-silicon atoms per 1 billion atoms. This process is not only technically demanding, but also time-consuming and labor-intensive, with strict requirements for production equipment and process control.
At present, the global silicon wafer market is mainly dominated by several major manufacturers such as Shin-Etsu Chemical Industry, Shenggao, Global Crystal and Siltronic. In the face of market recovery and growing demand for AI chips, these manufacturers are increasing their silicon wafer supply capacity by increasing capital expenditures, optimizing production processes, and technological innovation.
The development of AI chips not only increases the demand for silicon wafers, but also brings new opportunities to the silicon wafer market. With the application of 2.5D and 3D stacking technologies, the efficiency and demand of silicon wafers are expected to increase. Silicon wafer suppliers have the opportunity to play a strong role in this transformation through technological innovation and capacity expansion.
Figure: AI is driving the development of silicon wafers
The development of AI chips has put forward larger size and higher purity requirements for silicon wafers. For example, NVIDIA's GB200 AI chip enables the production of superchips by integrating two chips with the largest mask size on a silicon interposer, which requires a further increase in the size of the silicon wafer. At the same time, with the application of new technologies such as high-bandwidth memory (HBM), the demand for silicon wafers has also increased, as HBM technology requires more silicon wafers to achieve its stacking structure.
The supply and demand of silicon wafers in the market, after experiencing the recession of the semiconductor market in 2023, resulting in oversupply and high inventory, the supply of silicon wafers is gradually recovering as the demand in the semiconductor market gradually recovers and the inventory of silicon wafers returns to a healthy level. In addition to the original demand of the semiconductor market, in the face of the booming development of the AI market, the demand for silicon wafers is expected to increase and the price of silicon wafers will increase as the area of AI chips continues to expand and the silicon content is higher. Although the price only accounts for a small part of the cost of TSMC or Nvidia's final product, it will affect the subsequent supply of the entire product.
It is estimated that AI chips with a 2.5D architecture will require 2.5 to 3 times more silicon area than a traditional 2.0D architecture. This not only increases the demand for silicon wafers, but also presents new challenges and opportunities for silicon wafer suppliers. Silicon wafer suppliers need to scale up production, improve production efficiency, and may need to develop new technologies to meet market demand.
