Identifying anomalies in the vast amount of data generated by fabs.
AI holds great promise for our industry. It will help close the divide between design, manufacturing and test that is required to effectively produce today’s most advanced hybrid devices.
One way to look at the potential impact of AI in the semiconductor industry is to realize that it is more and more driven by software engineering.
When designing a chip, it’s essentially like writing code. AI, with tools like Microsoft’s Copilot, is having a dramatic impact on the way code is now produced. An engineer can now synthesize code, allowing the computer to generate routine and mundane parts. Semiconductor testing is also a form of programming. On the manufacturing side, factories generate gigabytes to terabytes of data daily containing valuable insights hidden within. Humans can’t sift through all this data to find anomalies, but AI can. Using AI to identify these “needles in a haystack” will revolutionize every aspect of the chip industry over the next five years.
Both AI and ML have already been deployed in advanced testing for complex systems and packages to track the number of test insertion points. Engineers are then able to decide which chiplets belong with which other chiplets in a specific package to optimize testing, add and subtract tests to tune the production flow.
Today, these techniques are deployed by only a small number of experts inside leading companies. Anecdotally, a large Japanese semiconductor company with tens of thousands of employees has 30 employees with the know how to use AI/ML. For AI to have an impact, it must be pervasive.
Not all use cases will equally benefit from the application of AI. When the semiconductor industry applies AI to a narrow problem where the width and breadth of information is narrow enough for a human to track, it is hard for the algorithms to do better than a human. One example is when an engineer wants to predict when a piece of equipment will need preventative maintenance. It’s hard to beat a human’s understanding of the equipment.
A better place to apply AI is with a wide field of information, such as consuming all the product lifecycle management (PLM) data that went into a design, all the vectors that were generated and those that were not, as well as all the equipment control variability data and in-line inspection data. The space of that data is broad and encompasses several different specialized domains. While different types of engineers focus on each part, the integration across those dimensions is something that can be done much more effectively in software and algorithms, and much better than individuals could. I call it human-limited yield improvement, as humans can’t understand and uncover all the underlying relationships across this huge data space.
AI’s power to transform semiconductor design and manufacturing will be showcased during the AI Executive Conference hosted by PDF Solutions Thursday, December 12, in San Francisco. PDF Solutions and other industry experts will be on hand to point the way toward the industry of the future and tout the democratization of AI for everyone in the chip industry, not just the few specialists today. Speakers will explore how AI is revolutionizing semiconductor design and manufacturing and discuss tangible applications of AI in the semiconductor industry.
The conference will offer insights into the power of AI to transform semiconductor design and manufacturing with examples of successful AI applications. Presentations will highlight how industry experts are deploying AI and ML to make a difference in their business. Others will detail how they are enabling the infrastructure to reduce the effort barrier, eliminate the barrier of expertise required to apply these techniques and create the democratization of AI.
PDF Solutions AI Executive Conference
Date: December 12, 2024
Location: St. Regis Hotel, San Francisco, Calif.
Agenda and Registration
Leave a Reply