Technology

Op-Ed: Using AI to Help Solve Today's Semiconductor Problems

Jonathan Ernst | Reuters
  • There seems to be a critical component missing in the ongoing discussions around the global semiconductor shortage now impacting the auto industry as well as others.
  • A lack of supply chain strategy, a forecasting of raw material shortages even outside of the pandemic, and sheer risk management could have helped mitigate this global crisis.
  • The chip technologies that exist today make it possible for cars to be functional and capable.  

There seems to be a critical component missing in the ongoing discussions around the global semiconductor shortage now impacting the auto industry as well as others. While saddled with the burden of previous trade policies, the U.S. needs an answer that can correct the short-term, supply-demand imbalance. Building $20 billion chip factories — or fabs — over the course of five years is a long-term solution to a near-term, crushing dilemma.

The automotive industry is feeling the strain the most. However, it is worth noting that a lack of supply chain strategy, a forecasting of raw material shortages even outside of the pandemic, and sheer risk management could have helped mitigate this global crisis. Keep in mind also that while cars are moving toward becoming electronics themselves, the chip technologies that exist today make it possible for them to be functional and capable.  

The manufacturing industry needs to quickly take note that what has been hyped as a necessity due to Moore's law — 5 nanometer nodes — is false 90% of the time. Building a $20 billion fab is simply expensive (though companies with near-monopoly status, such as Taiwan Semiconductor Manufacturing Co., would be served well by this approach.) TSMC admittedly cannot keep up with the industry's applications. Car companies and the majority of electronics companies that have technologies that are sub-micron, but larger than 10 nanometer nodes, are enough. And with clever design and iterative artificial intelligence approaches, the process will be less expensive and produced in a significantly shorter timeframe.

The shortage is a crisis that requires immediate action

Using the legacy equipment from the last 20 years to make products that the manufacturing industry needs now can be accomplished more efficiently through the utilization of intelligent control that is possible with AI. This will allow for older machines to make chips in ways they were not originally designed to do, but will allow for optimizing production in a way that results in better yields.

These fabs are prime real-estate for manufacturing things that can mitigate the chip shortage by focusing on:

  • Optimization with a reinvestment in established technologies that are available such as Nvidia's GPU's, the latest camera sensors, and more. Development of new materials and device designs are accelerated when paired with computation and AI.
  • Rethinking inspection and ensuring alignment with the manufacturing process at the highest level of detail possible, as opposed to relying on a high throughput surface scan on manual inspection during wafer level inspection.
  • Utilization of AI and predictive systems to continuously scan for defects in real-time to significantly improve product quality, increase yield, and prevent billions of dollars from being wasted from scrapped materials that are defective.

Production yields must remain high and in order to do so, manufacturers must perfect the manufacturing process and ensure quality control. Improved production, decreased waste, and optimizing what we currently have would make a tremendous impact. In fact, it could be the game-changer that makes higher levels of inspection details the standard. Incorporating AI in production processes eventually will allow us to secure our domestic manufacturing supply and prevent nation-state hacking.

Julie Orlando is the chief product officer at Nanotronics, and a member of the CNBC Technology Executive Council.

Copyright CNBC
Contact Us