An expert explains the AI chip ‘supercycle’, and the opportunity it presents for India

Nearly 600 companies from 52 countries converged at SEMICON India, the flagship conference under the India Semiconductor Mission (ISM), which is governed by the Ministry of Electronics and Information Technology (MeitY). Prime Minister Narendra Modi inaugurated the event, highlighting the rapid expansion of India's semiconductor ecosystem.

The conference unfolds during an era of unprecedented global demand for semiconductors, primarily driven by artificial intelligence (AI) and the necessary infrastructure. Experts refer to this growth as a semiconductor "supercycle," a multi-year phase of investment and expansion resulting from a fundamental technological shift. This is reminiscent of the supercycles witnessed during the computer boom in the 1990s and the smartphone revolution in the 2010s.

The driving force behind the chips boom is the surge in data centres, which serve as the physical infrastructure for storing and processing digital data. With the growing adoption of AI technology, AI data centres have become crucial, necessitating specialized components like AI chips or accelerators to handle the intensive calculations required for model training and execution.

These processors rely on high-bandwidth memory (HBM) chips, which stack layers of memory close to the processor, enabling rapid data transfer between memory and processor cores. However, the production of HBM chips is complex, requiring advanced packaging techniques to integrate processors and memory stacks effectively. While data centres dominate current demand, emerging industries such as robotics and autonomous vehicles are also expected to drive future growth.

These sectors require advanced chips with significantly higher memory and storage capacities compared to traditional vehicles. To secure a stable supply chain, memory manufacturers are increasingly shifting towards business-to-business sales, forming long-term contracts with major tech companies such as Microsoft, Amazon, Google, and Meta.

These agreements, often spanning several years, guarantee chip purchases regardless of demand fluctuations. Despite this growing demand, India currently plays a secondary role in the AI chip boom. India has approved 12 semiconductor projects under the ISM 1.0, focusing on assembly, testing, and packaging facilities. However, the country lacks a large-scale wafer fabrication plant for producing chips domestically.

Micron's Sanand facility in Gujarat will process imported wafers, while Sahasra Semiconductors will handle packaging for NAND flash storage. While these projects propel India into the later stages of memory chip production, wafer manufacturing remains an external responsibility. This situation leaves India primarily benefiting from the cost and availability of imported HBM and advanced accelerator chips, rather than capturing the profits generated by these components in the AI ecosystem.