Huawei, TSMC, Intel turn to big data and AI to optimize advanced manufacturing

Amanda Liang, Taipei, Staff reporter, DIGITIMES Asia 0

Credit: AFP

With the rise of Artificial Intelligence (AI) and Big Data, an increasing number of major companies have started showcasing their data processing capabilities and AI integration technologies for applications in advanced semiconductor manufacturing processes. Considering the demand for computing chips impacting the semiconductor manufacturing process, the 2023 World Semiconductor Conference Expo held in Nanjing recently saw Huawei, TSMC, and Intel emphasizing the application of AI and Big Data at the process level.

According to a report by Chinese media outlet Sci-Tech Innovation Board Daily, under the current AI trend, Huawei, TSMC, and Intel have each presented solutions for semiconductor manufacturing, demonstrated their Big Data platforms, and advocated for investments in computational power.

Interestingly, at the 2023 World Semiconductor Conference Expo, while TSMC's exhibition area was undoubtedly focused on the semiconductor manufacturing segment, Huawei was also allocated a spot in the manufacturing zone, presenting EDA tool solutions related to the development of computing chips.

Industry insiders cited in the report pointed out that, due to ChatGPT's heavy reliance on computing power, the complexity and quantity of computing chips have rapidly increased, subsequently increasing the demand for EDA tools.

They explained that in the current stage of chip development and simulation, large files ranging from kilobytes (KB) to terabytes (TB) need to be processed, and the read/write functionality of these files in memory affects the computing speed. As chip development and simulation require higher computing power and DRAM demands, specialized software like EDA increasingly requires support for cluster computing capabilities. Therefore, Huawei focused on showcasing its research and production solutions in EDA simulation.

At the conference, Su Hua, a manager in TSMC Nanjing responsible for lithography, mentioned that due to TSMC's gigafab with a monthly production capacity of up to 100,000 wafers and generating approximately 70 billion pieces of data per day, coupled its global expansion, TSMC relies on its big data platform to achieve high levels of productivity and quality standardization.

TSMC stated that each wafer is thoroughly planned before production, including determining the process flow, process parameters, and optimization strategies. During the production process, dynamic corrections can be made based on deviations from the previous steps, establishing the relations between various steps at the big data level to the process flow to ensure dynamic stability in the production line. This is a key aspect of TSMC's quality control.

As for Intel, China media reports suggest that its AI deployment in semiconductor manufacturing involves defect detection on production lines, tool/equipment/wafer fab matching, and multivariate process control. It is believed that AI will bring transformative changes to semiconductor manufacturing.