AI is fast revolutionizing our world as we know it, making it efficient. Over the past few decades, innovation in silicon manufacturing has been confined mainly to shrinking transistors to include more of them in integrated circuits. However, this innovation has significantly changed over the last few years due to the advent of AI, which has encouraged innovators to think differently and reinvent the basic chip design and build a new type of semiconductor chip built for AI, powering the next generation of computing. 

Over the next several years, growth in the semiconductor industry is expected to grow substantially. Aided by artificial intelligence (AI), the growth is expected to be five times more than the non-AI related semiconductor growth, with the global AI-related semiconductor market growing from USD 32 bn in 2020 to USD 65 bn in 2025. 

AI market dynamics are changing, and intelligence is moving towards the edge, and one can derive its effects on the chipset revenues. While assessing the deep learning chipset revenue, ‘Inference’ revenue is expected to grow almost 10X from ‘Training’ by 2025. While assessing the deep learning chipset revenues from a power consumption point of view, it is expected that the ‘sub 5W’ segment is expected to witness maximum growth, more than 2X the growth of high (>50W) and medium (5-50W) power domains by 2025. In terms of the market sector, ‘Edge’ is expected to grow more than 2.5X than the ‘Enterprise’ revenue by 2025. 

Many organizations are performing all sorts of complex AI and ML applications. Mobile, high-performance computing (HPC) is most important from a pure semiconductor consumption perspective. Therefore, the next generation of semiconductor chips will have to be intelligent to incorporate efficiency and specialization. Chip architecture will be designed to optimize the software running on it and run more efficiently. 

As companies look to manufacture these chips, they must ensure that the new technologies will have more options for processing stacks that can evolve and change as the demands of ML and AI models rises. This will enable the companies to stay competitive and not stagnate when software is limited by hardware. 

The next generation of chips will shape the field of Artificial Intelligence in the coming years. Effective collaborations between peers, academia, corporations, and the government will lead to the creation of the hardware which will power the upcoming era of AI. 

Edited by Anjali Raja K 

Sources:

Community- By NASSCOM Insights

Data and some of the insights are from leader sessions in SemiconIndia 2022