The global silicon wafer industry is witnessing a resurgence, propelled by the escalating demand for advanced semiconductor components essential for artificial intelligence applications. In the latter half of 2024, worldwide silicon wafer shipments began to recover from the downturn experienced in 2023, as reported by the SEMI Silicon Manufacturers Group . Despite a 2.7% decline over the year, totaling 12,266 million square inches, and a 6.5% revenue contraction to $11.5 billion, the upward trend in the latter part of the year indicates a positive trajectory for the industry.
This revival is largely attributed to the rapid adoption and advancement of generative AI technologies, which have significantly transformed the semiconductor landscape. Hyperscale data center operators and original equipment manufacturers are expanding their server capacities to meet the computational demands of AI model training and inference processes. This expansion necessitates a substantial increase in AI accelerators, leading to a surge in demand for processor units and high-bandwidth memory associated with these accelerators.
The heightened demand for HBM products is evident in the financial performance of key industry players. SK Hynix, for instance, surpassed Samsung in quarterly profits for the first time, driven by a surge in demand for advanced memory chips necessary for AI data centers. In the fourth quarter of 2024, SK Hynix’s operating profit soared to Won8.1 trillion , with sales increasing by 75% to Won19.8 trillion. HBM products, crucial for AI servers, significantly contributed to this profit, comprising 40% of SK Hynix’s DRAM chip revenue. The demand for these chips is expected to double due to investments in AI infrastructure projects, such as the massive $500 billion Stargate initiative by OpenAI and SoftBank.
Similarly, Micron Technology has experienced a positive financial outlook, with shares surging approximately 14% following an upbeat first-quarter forecast. This optimism is fueled by the demand for memory chips utilized in AI computing. As one of the three primary suppliers of HBM chips, alongside SK Hynix and Samsung, Micron is capitalizing on the growing need for semiconductors in generative AI technologies. These DRAM chips are essential for AI-oriented graphics processing units, facilitating the processing of vast amounts of data. Micron anticipates a record revenue of approximately $8.7 billion and an increase in its gross margin to around 39.5% for the first quarter.
The surge in AI applications has also impacted semiconductor manufacturing giants like Taiwan Semiconductor Manufacturing Co. . TSMC reported a significant profit increase, with fourth-quarter earnings rising 57% to NT$374.7 billion . This growth is largely driven by high-performance computing revenue, which surged 58%, offsetting declines in other sectors. CEO CC Wei highlighted the tight AI capacity but reassured that TSMC is working diligently to meet customer demand.
However, the rapid expansion in AI-driven demand has introduced challenges, particularly concerning supply chain constraints. High-bandwidth memory is in even shorter supply than GPUs, potentially hindering the data center industry’s expansion plans. The HBM design provides greater speed and reduced latency, key to the performance of AI processing. During the COVID-19 supply chain shortages, carmakers could not obtain chips for their vehicles, so they simply built cars without the chips and mothballed them until they could shore up inventory. GPU vendors like Nvidia and AMD don’t have that option. No high-bandwidth memory means GPUs can’t be assembled because the HBM has to be added to the GPU package at the manufacturing stage.
In response to these challenges, companies are exploring innovative solutions. NEO Semiconductor, for instance, announced the development of its 3D X-AI chip, designed to replace existing HBM chips and address data bus bottlenecks. This advancement aims to alleviate some of the supply constraints by providing alternative high-performance memory solutions.
The global silicon wafer market’s recovery is further bolstered by strategic investments and collaborations. For example, Foxconn is building a Nvidia superchip facility in Mexico, aiming to meet the growing demand for AI hardware. Such initiatives not only enhance production capacities but also diversify the supply chain, reducing dependency on a single region and mitigating potential geopolitical risks.
