Smarter engineering in semiconductor manufacturing - the exciting road ahead!

DT's journey in the semiconductor manufacturing vertical started more than 3 years ago! It has shaped up to be an exciting journey with strategy to execution engagements solving numerous high-value challenges with Machine Learning (ML) based solutions across BUs and functions - Product digital catalogs, Upgrade recommendations engine, Obsolescence, Installation & Warranty.

AI/ML has the potential to generate huge business value for semiconductor companies at every step of their operations, from research and chip design to production through sales. Based on a recent McKinsey survey about 30 percent of semiconductor device makers are already generating value through AI/ML while the rest are also piloting solutions in that space.

As you can see from the use cases below, there are several value chains to explore like the demand forecasting use case which spans manufacturing, procurement, sales, and operations planning, or the chip-design automation and verification which extends from design into manufacturing. The present-day challenge is now compounded by the chip shortage impacting global manufacturing which is expected to last atleast until 2022!

Semiconductor firms have high capital requirements so operate in a winner-takes-all environment and have a need to shorten product lifecycle and aggressively pursue innovation to introduce new products quickly cutting down on cycle times.

The research and design costs for the development of a chip increased from about $28M at the 65nm node to about $540M for the leading edge 5nm node (Apart from fab construction costs which increased from around $400M to $5.4B). There is an immediate need to bring some of those costs down significantly with the automation of chip design and verification.

Some AI/ML use cases that can yield immediate cost benefits are:

• Tool parameters adjustment - specifying constant timeframes for various process steps in manufacturing individual wafers which can experience fluctuations that can damage or waste the chip. Machine learning models can be employed which use live tool-sensor data, metrology readings, and tool-sensor readings to capture nonlinear relationships and adjust the processing accordingly.
• Visual Inspection - Ensuring wafer quality by detecting defects during production is typically conducted using cameras, microscopes, or electron microscopes which are inspected manually subject to errors and backlog. ML driven computer vision tools can be trained to detect and classify defects more accurately and consistently on those images leading to early detection and reduced costs.
• Research & Design - Firms can avoid time-consuming iterations and errors during the research and chip-design phase through automation related to physical layout design and verification process. ML can identify patterns in component failures, predict failures in new designs and propose optimal layouts for better yields. Catching IC design errors early on can help reduce/eliminate costly iterations during manufacturing

Would love to hear your experiences conceiving and delivering value driven by smart engineering solutions in manufacturing!

Source: https://www.mckinsey.com/industries/semiconductors/our-insights/scaling-ai-in-the-sector-that-enables-it-lessons-for-semiconductor-device-makers

Sid Nandi is an entrepreneur, business transformation leader, coach, technology strategy & management consultant experienced in helping organizations, business and technology undergo digital transformation with agility delivering customer centric products and solutions.

?His venture (DT) provides management, strategy & delivery consulting specializing in digital transformations to help accelerate organizations transform their business, technology & operations in creating sustainable value in a digital world. He is passionate about connecting and collaborating with fervent transformation professionals from all industry verticals committed to building sustainable enterprises!.