Kate McAlpine:
Welcome to the Michigan Minds Podcast, where we explore the wealth of knowledge from faculty experts at the University of Michigan. I’m Kate McAlpine, engineering news editor for the Michigan News Office. I want to welcome Valeria Bertacco, the Mary Lou Dorf Collegiate Professor of Computer Science and Engineering who explores hardware solutions for next generation computing and security. Valeria is also the vice provost for engaged learning at the University of Michigan, supporting international partnerships and initiatives.
We’re here to talk about semiconductors and strictly a semiconductor is a device that can switch between whether it’s transmitting electricity or not but, semiconductors and computer chips kind of get muddled together in the news. Can you define them both and explain how you get from semiconductors to computer chips?
Valeria Bertacco:
Yes. Let’s start with semiconductors. Your definition is very accurate. However, I would say that there are additional type of devices that are part of the semiconductor family. An important one are semiconductor devices that can store information, memories made of semiconductor materials. And, there are even more other types of devices like sensors for instance sometimes are made of semiconductor material.
Now, how do we go from this transistor, the semiconductor devices that can switch, to a computer chip? Let’s think about your house. You probably have some rooms in your house where there is one light but two different switches in different part of the rooms where you can turn on the light or turn off the light. And, as long as you turn one switch, the light turns on. So, that is a OR gate and this is one of the key pillars of building a computer chip. And then, also in your house you may have some fuses and certain lights can only turn on if the fuse is on and if your light switch is on and that’s called an AND gate. And, it’s the other pillar of how to design a computer chip. And then, when I have these pillars, which usually take four to six transistors each, I can build all sorts of other complex computing solution. Semiconductor, I would say, are the building block of computer chips and that’s the relation.
Kate McAlpine:
Okay. So, how dependent are we on semiconductors, chips, and everything else you can make with semiconductors?
Valeria Bertacco:
Let’s see. I wake up in the morning and I use an alarm for that and that is chips. And then, I brush my teeth and my electric brush has chips. Then I open the fridge, which has chips, and my stove, which is induction, has chips. And, my car has thousands of chips. And, I have to tell you, I have a pretty old car. Then, the traffic lights have chips to navigate to work and the door opener of our buildings in the university have chips. I spend the rest of my day surrounded by chips as I work with my computer and do meetings online. So, I don’t know if there is any part of life that doesn’t entail chips and semiconductors. Even when you go camping, the GPS to make sure you don’t get lost and then you probably would have a phone in case of an emergency.
Kate McAlpine:
So, they’re all over in our lives. How are they made, in a nutshell?
Valeria Bertacco:
So, making a chip entails several, when I say several, its several tenths, of steps of deposition and etching. Now, deposition is a step of putting some material on top of your chip in a specific pattern. So, think about decorating a cake where you may put a stencil on top of your cake and then you sprinkle the powdered sugar and then you have the pattern you want on the top of your cake.
So, in a similar way, when I’m trying to build a semiconductor, I will put down a stencil, it’s called a mask, and then I use UV light, which is used to make the exposed surface in my stencil sticky. And then, I can use a gas or I can spray some material right on top, that’s my powdered sugar, and then I can remove the stencil and I have the pattern made. In between deposition and etching, often you have to wash the surface of the chip because you want to make it extremely smooth to prepare it for the next step of deposition.
Kate McAlpine:
Why do we need like tens or hundreds of layers in these semiconductor chips?
Valeria Bertacco:
The reason why you need to do this many times is because every deposition and etching step only puts one layer of material on your chip, and you need to do this for every type of material and sometimes even many times for a single type of material. So, for instance, in building a transistor, I have to do the position for the N part and the P part and the insulation in the middle.
Kate McAlpine:
Can you say a little bit more about what the P and the N semiconductors are? What do they do?
Valeria Bertacco:
It’s a way to modify the crystals in the silicone material to favor the electrons going from one end to the other of the transistor.
Kate McAlpine:
Okay.
Valeria Bertacco:
The set of steps that is repeated many times is to depose the metal layers. The metal layers are the connection between transistors so they can talk to each other. And, once upon a time, in the olden days, you just needed a single metal layer to create all the connections, basically drawing a map that connects all the houses. But, nowadays a flat surface is not longer enough to do all the connection than one needs. And so, you have to do a deposition with one metal layer, then you put insulation in, then you do another metal layer, and you keep doing this many times to build 3D structure. Think about on the freeway when you see spaghetti connections, you need to build a road in the third dimension so that you can have all the connections you need.
Kate McAlpine:
I mean, it sounds complicated, but do you know why we stopped making them in the US?
Valeria Bertacco:
We never really stopped. I would say that today there are fewer companies making chips in the US than we used to have in the past. And, I think that one of the key aspects that led us to the situation that we have today, is that in the past, many companies in the United States were making chips. They were making both the design and then producing their own chip. And then, over time, as the cost of producing chips became higher and higher, they were more constrained in their finances so they had to choose and they went with the design because the design allowed more innovation because the design allows more flexibility and this lower cost. And so, they kind of abandoned their chip manufacturing aspect and they focused on the design.
While we have other companies in other countries who were born only producing chips for other people, like TSMC is called pure-play because they don’t design any chip. They serve companies who do design and they produce a chip for them. And so, they really specialize in that and they never stop doing that.
And that, I think, has created a pull-pull situation where the company thought, “Well, I don’t need to have my manufacturing and do this billions of dollar investment to keep up with the technology advancement. I can use this pure-play company and it doesn’t have this big initial step of investment.” And so, over time, more and more companies started to rely on dedicated companies that do manufacturing.
But, I also like to tell you that things have been changing the past three, four years. A number of companies both domestic and international companies have been investing in developing manufacturing plants in the United States, for instance, TSMC and Samsung and we know about Intel. So, there are a number of companies who are developing more plans to bring more manufacturing of chips to the United States.
Kate McAlpine:
What is driving that move?
Valeria Bertacco:
In the immediate term, I would say the government has issued the CHIPS Act in 2022, which entails a lot of investment to support companies who want to open manufacturing plants in the United States. But, if I were to look at the years before that, I would say the Department of Defense was concerned about reduced manufacturing in the United States even 10 years ago. Usually the Department of Defense who needs chips that have to be secure, they would use a US design team and then they would use a US-based manufacturing fab.
And, over time, about 10 years ago, they started really observing that if they want to do the frontier technology, they couldn’t use any US-based manufacturing place. And so, there was a challenge. They were thinking about, “How do we address this problem? How do we do research and innovation that helps us leverage commercial products?” But then, I think what really tipped the scale was the pandemic and the shortage of chips in the United States due to the reduced supply chain that took place during the pandemic has really made everybody aware how dependent we are on these chips and how we need to have a more robust supply chain. We shouldn’t be depending on countries that may go offline at any point in time for the US economy. And, I would say we are in the state of Michigan. A big industry that really suffered from that was the automotive industry who could not put cars on the road because some component was missing.
Kate McAlpine:
I see. And so, what does it take to rebuild semiconductor manufacturing in the States?
Valeria Bertacco:
As I was saying before, we are already seeing some being built. It takes about four to five years to build a fab in terms of the time from when you invest the money to when you actually have the fab online ready to produce. It takes developing the workforce. We have a lot of engineers and technicians that work in semiconductor broadly but, of course, we had a decreasing number of experts that work in the manufacturing because the manufacturing had moved away.
So, it takes that, but I would say it also takes the investment in the innovation because you cannot build a fab and it’s a very beautiful fab and it does a most advanced technology node, and then you stop there because, if you stop there, two years from now, you are in the trailing node again. So you need to just build a fab that is at the frontier and have the workforce, but you also need to invest in the people who are going to figure out what’s the solution five years down the road and then 10 years down the road so you can build the momentum and keep it up.
Kate McAlpine:
And, what is the role of U of M in trying to maintain that momentum?
Valeria Bertacco:
I love this question. So, at the University of Michigan, we have one of the largest computer and electrical engineering programs in the country surrounded by also very strong programs in material science and in chemistry. So, the entire system stack of producing chips. And, when I say one of the largest programs, I mean both by number of students, we’re talking about 700 plus graduate students that work in one of these fields and also the number of faculties involved in doing research and innovation in this space. So, we have counted over a hundred faculty in this research segment. And, I would say literally just the raw number of people working in this space is among the largest in the country.
Not only that, but the type of researchers that we have at the University of Michigan are very amenable to bringing their ideas to market. So, we see a lot of startups, we see a lot of collaborations with companies to bring innovations into a product that that company is interested in. And so, I think we are really positioned to make a leadership contribution and the leadership role into this reborn semiconductor industry in the United States.
Kate McAlpine:
All right. So, we have the CHIPS Act, we have workforce development. What else is it going to take to keep this going?
Valeria Bertacco:
Well, the CHIPS Act is a one time and really other countries and sometimes even other companies have really learned that you need to sustain investment to keep up the momentum. So, the CHIPS Act started in 2022. There is still funding being distributed but, I really hope that there will be another investment down the road and not maybe exactly the same. Five years from now, we are going to start to really have some clarity on which are the subsegments where the United States has really a key advantage and where really we would like to invest. But, if it’s one time, it will be difficult to build enough momentum that is sustained.
Kate McAlpine:
Where is semiconductor technology headed? What are some directions that the next big thing could go in?
Valeria Bertacco:
A direction that people have always loved very much is making transistors smaller. The VLSI revolution took place in the early eighties and was sprung by a book that was published by a Michigan faculty. At that time, she was not a Michigan faculty yet, but then she became Lynn Conway. And, basically taught systematic techniques and provided systematic algorithms to design microchips at scale. What happened is, when in the early seventies people started designing chips, they would take a piece of paper and draw by hand every single transistor in that chip and they got very good at it. They could do thousands of transistors in a chip.
However, as technology evolved, it became possible to fit millions of transistors in a chip and it wasn’t really possible to draw millions of transistors in a single chip all by hand. And, what she did is she came up with an algorithm so that one could describe the chip at the high level, and the algorithm will provide all the drawings with the correct distances between transistors, the correct spacing between metal lines. And so, basically that unlocked the ability for engineers to design much more complex chips. That was possible just a few years prior.
Kate McAlpine:
Wow.
Valeria Bertacco:
Back to our question.
Kate McAlpine:
Yes. Where might things go from here?
Valeria Bertacco:
Yes. So, transistor scaling has slowed down significantly in the past decades. There are some ideas on how to build the physical structure of a transistor to try to push it farther out. What I’ve seen in the past, I would say, five years, is a proliferation of memory types of devices, devices that can store information, devices that can retain the information, even if the chip loses power. And so, I can operate that chip in a solar power environment or I can drop it in a field. And, when there is energy, it can compute. And, when there is no energy, it goes to sleep and retains all the information. So, there are many different architectures that have been proposed in recent years in the space of memories and some of them have become commercial, some are still in the research stage.
And then, there are lots of innovations in the space of moving data. So, connecting memory with computers or connecting memories with memories. There are innovations on other types of materials that maybe can be used to develop transistors that are at a smaller scale or have different characteristics even if they’re not smaller. I would say those are the key spaces. And then, in the horizon there is a quantum computing promise, always there.
Kate McAlpine:
All right. So, we have to figure out which ones make the most sense to manufacture in the US?
Valeria Bertacco:
No, we don’t. I think you don’t want to wait first to find out which is the one that makes most sense. It would be best to invest broadly and see which one takes off. Sometimes it’s hard to predict these things because the success or the failure of a solution is not only based on the technological innovation it provides, but how costly or viable it is to produce that at scale.
So, sometimes it’s hard to predict these things. I would say investing in a broader set of solutions and then seeing what becomes successful would be great. I also think that, given that this is in the United States, which is known worldwide for creativity and innovation, solutions that are disruptive are the ones that are best suitable for the United States. Engineering away little details to try to make a transistor maybe a tiny little smaller, I’m not sure that that’s the best use of the US engineering mind.
Kate McAlpine:
And, is there anything you’d like to add about semiconductors, chips, manufacturing in the States?
Valeria Bertacco:
I’m really excited to see how… The general population thinking about semiconductors has really changed. You see semiconductor and micro electronics in the news so much more. You see all these announcements of new companies, and expansion of the workforce, and I think it’s paying off. One thing I’ve been noticing is our own enrollment in classes where you learn how to make chips and how to design chips has doubled over the past three, four years. So, we really see that it’s becoming top of mind for many people how to participate in this industry and how important it is to participate in this industry. And so, I’m honored that I’m right in the middle of this.
Kate McAlpine:
Thank you very much Professor Bertacco. It’s been great to hear about the semiconductor industry.
Valeria Bertacco:
Thank you very much, Kate. It was great to have an opportunity to chat with you today.
Kate McAlpine:
Thank you for listening to this episode of Michigan Minds, produced by Michigan News, a division of the university’s office of the vice president for communications.