Expert Q&A: What's driving the semiconductor shortage?
Last week the White House announced its support for legislation and funding to address recent shortages in the semiconductor supply chain. Concurrently, the President signed an Executive Order to begin a "100-day review of supply chains for four critical products: semiconductor chips, large-capacity batteries for electric vehicles, rare earth minerals and pharmaceuticals." These initiatives respond to shortages that have negatively impacted automakers and other manufacturers who rely on the semiconductor industry.
We asked Illinois ECE Professor John Michael Dallesasse to share his insights. Prof. Dallesasse has over 20 years of experience in the Optoelectronics industry and has held a wide range of technology development and management positions, including Vice President of MicroLink Devices and Senior Director of Engineering and Technology for Emcore's Fiber Optics Division. Most recently, he was the Chief Technology Officer, Vice President, and co-founder of Skorpios Technologies, Inc., a venture-capital funded startup that is developing and commercializing silicon photonic ICs based upon a wafer-scale process for selective integration of III-V materials on SOI substrates.
Q: Is the shortage of semiconductors a result of the pandemic, or has this issue been building for a longer period of time?
A: I would say that the shortage we're seeing is a symptom of a bigger problem in the semiconductor supply chain. While the current shortage may have been precipitated by the pandemic, the cause is the result of deeper changes that have been happening for some time. The cost of building a semiconductor fabrication plant (fab), and the resulting expense associated with idle time, has created a business model where running at or near full capacity is critical. Consolidation in the industry has also led to both a small number of suppliers and a small number of geographical areas where fabs are located. This creates a significant risk for disrupting the supply chain by events such as pandemics, political upheaval, natural disasters, business decisions to capitalize on a near-monopoly status, etc. Diversifying the number of companies and geographical location of fabs would de-risk the supply of integrated circuits (ICs), but it may impact the economies of scale and ability to run at or near full capacity.
Q: Automakers were singled out as an industry facing severe shortages, but does the problem extend to other industries as well? If so, which ones?
A: Anyone purchasing a large number of chips that need to plan their production well in advance could have been affected by this. If one customer cuts their projections, an IC supplier is going to want to sell that capacity to someone else, so the fab isn't idle. Idle time is incredibly expensive because of the cost to build and maintain a wafer fabrication facility. I don't specifically know who else was affected, but anyone who decided to cut their future deliveries because of the expectation that the pandemic would hurt their sales would be in this category. Lead times are long, both because of capacity constraints and because these are very complex things to make. We may not know what other industries are affected because they could be keeping this confidential - such news could have a significant impact on stock price, valuation, etc. They may not want to disclose it until they have to. There could also be national security issues.
Q: Is the supply shortage a short-term problem, or will it take longer to fix? What steps do you think need to be taken?
A: This gets into the question of how much customers are willing to pay to de-risk their supply chain. More fabs could be built around the US or around the world to improve supplier and geographic diversity, but if more fabs mean more idle time or some other form of reduced efficiency, costs will go up.
Q: How can academic institutions contribute to a solution? Are there research initiatives that have the potential to help?
A: At least right now, this is more of a business issue than a technical issue. Perhaps the research question to ask is, can we produce something with the functionality of a complex IC that doesn't need to be made in a billion-plus dollar fab. Perhaps the answer to the question of what comes after Moore’s Law is the same as the answer to the question of how to deal with the IC supply chain problem. At the same time, ICs are very good at what they do, and anyone who says they can displace them should be viewed with a healthy dose of skepticism. For a long time, people have believed that we can solve any problem with more complex silicon and better algorithms. With performance scaling now at its limits, we may need a paradigm shift on the order of the tube-to-transistor shift to move past these problems, at least for some applications. Elegance often beats brute force, and novel thought and openness to change might define the path forward.