Friday, 4 March 2022

TSMC Executive Says Chip Shortage to Last Several More Years

One of the top executives at TSMC gave an interview recently on the chip shortage, and surprise, surprise, there’s no silver bullet that will end it any time soon. This isn’t exactly breaking news, but others industry heavyweights have been ruminating recently that with COVID-19 transmission rates dwindling and lockdowns easing up, there might be a chance that the situation would start to improve in 2022. According to Dr. Y.J. Mii, who is TSMC’s Senior Vice President of R&D, that’s not going to happen.

In the interview, Dr. Mii offered a very simple explanation for both the cause of the chip shortage, and why it won’t be ending any time soon: supply and demand. According to Mii, demand for semiconductors has been surging the past few years, and it  caught the industry as a whole somewhat flat-footed. In response, many companies have taken action by announcing new fabs and other attempts to expand their wafer production capacity, but those efforts will take years to bear silicon fruit.

“Right now, the industry is investing a tremendous amount of capital into building extra capacity to solve this chip shortage problem,” Mii says. “We have a much clearer picture of future demand today than we had two years ago.” Mii does say that he thinks COVID-19 played a role in crashing the global economy, but he thinks the increased demand for semiconductors is more to blame for today’s woes. It should be noted that these are his opinions, and not those of TSMC as a company.

Dr. Y.J. Mii of TSMC began working for TSMC when it was just starting to make 90nm wafers. (Image: TSMC)

Interestingly Mii’s assessment of the situation differs slightly from recent comments made by Intel CEO Pat Gelsinger, who predicted the industry would start to rebound in 2023. Gelsinger said demand only surged during COVID-19’s onslaught, going from a typical five percent growth rate annually to 2o percent almost overnight as everyone was stuck at home around the world. However, Gelsinger line of thinking seems to be in harmony with Dr. Mii’s, as Intel is currently investing billions of dollars on new fabs all around the world.

Intel has already revealed Ohio as the location for its next fab, in addition to adding two more in Arizona alongside its current operations, and it’s still yet to announce a new fab somewhere in Europe. That announcement should be made soon, with the rumor mill suggesting it’ll be somewhere in Germany. These efforts to diversify the location of its silicon fabs will help it mitigate future supply chain disruptions in one location due to a natural disaster, political unrest, or a future COVID-19 outbreak. These types of concerns have taken on new urgency in the wake of the current situation in Ukraine, with people pondering the repercussions of a potential similar action by China against Taiwan, which is where TSMC is headquartered.

Dr. Mii, who is a 26-year veteran of TSMC and was transferred to the R&D wing of the company when it begun making 90nm wafers, also discussed the difficulty companies are facing as manufacturing processes continue to shrink. Mii is in charge of the company 3nm node, which is scheduled to go into production later this year for products such as Apple’s next-gen M2 or M3 chips. He says he’s already working with his team on TSMC’s 2nm plans.

“We are approaching atomic scale,” Mii told the IEEE. “Before, we could achieve the next-generation node by fine-tuning the process, but now for every generation we must find new ways in terms of transistor architecture, materials, processes, and tools.” He summarized this thoughts by noting, “In the past, it’s pretty much been a major optical shrink, but that’s no longer a simple trick.” Interestingly the 3nm node is just now coming online two and a half years after TSMC ramped up production at 5nm, giving us a glimpse into future time tables. Currently both Nvidia and AMD are using TSMC’s 5nm process for its next-gen GPUs, and Intel is a customer as well as it is creating its new Arc GPUs on the company’s 6nm process. Intel is also hoping for some of that 3nm action for its Meteor Lake GPUs, which are scheduled for 2023.

Now Read:



No comments:

Post a Comment