Certify Alexametrics gif

    Semiconductor Forecast 2022

      Dec 15, 2021

    VLSIinsiders' cloudside chat - December 15, 2021

    This week Andrea Lati, Dan Hutcheson, Jason Abt, and Risto Puhakka discuss what insiders are saying about …

    • SEMICON West and the new tools introduced there
      • Attendee take on next year’s business
    • The 2022 Forecast for semiconductors and equipment
      • Backlogs and Inventory
    • Qualcomm Snapdragon 8CX predicted it would be N7 but it was N5
    • DRAM Technology Forecast as cell sizes get smaller
    • China’s back in the news with SMIC back on the entity list

    About VLSIinsiders: Every week our analysts have a cloudside* chat to discuss current events and key issues of concern while sharing what they’ve heard over the past week from the semiconductor industry insiders. 
    * cloudside chat: A fireside chat without the fire and is across the clouds of the internet

    Other VLSIinsiders

    Transcript

    VLSIinsiders' Cloudside Chat — December 15, 2021

    DAN HUTCHESON: Welcome to VLSIinsiders. This is Dan Hutcheson.

    ANDREA LATI: This is Andrea Lati.

    JASON ABT: Jason Abt.

    RISTO PUHAKKA: This is Risto Puhakka.

    DAN HUTCHESON: So, we're closing out the year. I want to wish everybody happy holidays and but last week, we were all at SEMICON West. And then we've also been just digging into the forecast and finalising it going into next year. So, we'll be talking about that. And then some other exciting things like what's going on in China and some of the process technology. So, Risto, why don't we talk about SEMICON West since you and I were both there. What was your impression?

    RISTO PUHAKKA: Well, I mean, first of all, it was nice to be out there having meetings. Even though it was a little bit spares, but I think the important thing was to actually, it was done, you know, and people got out, some people flew in from, especially from Europe. So, it was really good. Couple tool announcements, you know, good conversations. So, I was actually really happy to get kind of good old-fashioned networking going.

    DAN HUTCHESON: Yeah, yeah, that was a really nice piece of... Ah, what tools did you see you really liked?

    RISTO PUHAKKA: I think, you know, it's a bit of a niche tool. But, Infinite Test has strong plants have kind of far, you know, let's call it fully automated in the line type of AFM, you know, metrology tool. It was really nice to see them actually really think from basically putting it online rather than the classic. Hey, let's do a better lab tool. So, that was really interesting.

    DAN HUTCHESON: So, that was…

    RISTO PUHAKKA: Yeah, definitely. I mean, let me put it this way. It's definitely something worth paying attention to over the next couple of years once they unroll it, then because it may have an impact on the metrology space.

    DAN HUTCHESON: You know, when I first started was when Peter Jenkins went there to run it. They pulled him out of retirement. I said, “Wow, great technology, but it’s not ready for the fab.” It looks like something that’s ready for the fab now, doesn't it?

    RISTO PUHAKKA: Yeah. That's exactly the thought process there. So, and they have a first tool build up right now and that's all they have goods to show very quickly.

    DAN HUTCHESON: Anything you saw that you like?

    RISTO PUHAKKA: Well, I mean, then there really the issue was more of the meetings and getting kind of the overall forecasts update, you know, when people view next year in which is released. I mean, it's funny, you know, the comments are that everything looks fantastic, great. We could ship more, we have visibility, but then on the other side, everybody's worried we are in the cyclical industry and say, how long can this last?

    DAN HUTCHESON: Yeah, I heard the same thing. I heard 20 plus percent people were generally more bullish than we were. And then when we but then they all said it was based on backlog. And yeah, you know, they could next year's good, just if you just as long as the backlog doesn't get cancelled. In fact, some people had some fears that it was another there was like the SEMICON of Y2K…

    RISTO PUHAKKA: Yeah, yeah!

    DAN HUTCHESON: …which having been a moment event and podium. I remember that very well, Y2K, for this young investment analyst got up and said, “This time, it's different. We're not going to have a downturn.” Just as it was unfolding. But yeah, the fact that everybody's really cautious is really different from past downturns. So, when you get to this point, where all of a sudden, you're on the precipice, and nobody sees it, it seems like everybody's worried about it. So that just tells you.

    RISTO PUHAKKA: And you know, that if you especially, you know, talk to equipment supply chain companies, you know, they just pretty much all the major ones invested really heavily so. Basically, they got their boards approve $100 million a year ago, now it’s spent, and now they're going another $100 million. So, of course, they're getting a lot of questions, you know. How far? How much? You know, and obviously the boards want to keep the money that they made and do something else with it. So yeah, it's a very classic challenge, you know, on management.

    DAN HUTCHESON: In fact, a lot of the All-Stars that made it into this year’s Hall of Fame/All-stars that's coming up later this week, early next week, who were there because of just really preparing for the upturn and being ready.

    RISTO PUHAKKA: Yup. Yeah, exactly.

    DAN HUTCHESON: Let's, why don't we switch over? You've got the real numbers. You just finished, you just shipped yesterday. Tell us what next year is going to, how it's going to turn out?

    ANDREA LATI: Well, yeah, let's start with equipment actually since that's always the more difficult one. And for this year actually, we ended up downgrading the equipment forecasts a little bit from 40% to 35. Mainly because, you know, companies could not ship in the second half; they just did not have enough parts to finish their systems. So, all that is rolling into 2022. And for 2022, we're now expecting equipment to increase about 20%, whic, again, will be a pretty good year. And in terms of, you know, the activity, it’s as hot as ever been. And if you look at it, there are a lot of projects currently ongoing. Intel is, you know, building in Arizona. You have TSMC in Arizona. TSMC also is planning to build in Japan; you have Samsung in Texas, Global Foundries, in both Singapore and New York, and then you have Texas Instruments, you know, also in Texas. So, again, there is a lot happening in the equipment market in terms of activities. So, I think the momentum is there. You're right about the backlog because everybody's hoping that the backlog clears next year. But given the level of activity, I think that 2022 should be another very good year for the equipment suppliers. And switching gears...

    DAN HUTCHESON: Fab expansion is going on right now.

    ANDREA LATI: Yes, yes, it's, it’s incredible; actually, it's incredible. So definitely, there is a lot of momentum out there. So, I think that should carry us throughout 2022 without too much of an issue. Then 23’ and 24’ become a little bit more debatable, but let's get 22’ out of the way first. So then, in terms of ICs, we didn't really change the forecast very much because most of the numbers came in line with our expectations. We're basically forecasting this year's market to grow about 26% and next year to grow about 11%. With ICS, it is a little bit more tricky. Because again, the inventories are becoming a little bit of an issue, especially at the OEM level, they're actually increasing quite a bit. So, then it becomes more of a question whether they are going to tolerate a higher level of inventory or not. But given what's happening, you know, at the macro level, and also with this new variant of COVID-19, my sense is that the inventories will probably stay elevated throughout the year for 2022. So, and if that's the case, then I think that 11% growth is more than doable for semiconductors. But again, inventory is something to watch, because especially at the OEM level, they're increasing.

    DAN HUTCHESON: It's interesting because a month ago, it didn't look too bad.

    ANDREA LATI: Yeah, and but when we got. Yeah, you're right, because, but they've been increasing, especially when we got 3rd quarter data. You could actually tell that, you know, especially in the consumer side and the computer side, especially PCs, they are actually holding a lot more than they have done in quite some time.

    DAN HUTCHESON: Yeah, and a lot of that is that they're holding the inventory they can get and then they've got all this stuff they can't get, so they can't ship which just shows you how everything is discombobulated.

    ANDREA LATI: Correct, correct. And it's across the board, Dan, it's not just you know, the consumer. We have a similar story in the automotive market, even there is quite a mess, actually. Because, you know, the only way they are waiting for a few more parts that they cannot get. And sometimes, it's more of a supply chain issue than a capacity issue. So, and that's what makes the ICs forecast a little bit more difficult this year because, again, the pressure points are intensifying a little bit more than what we've seen for quite some time.

    DAN HUTCHESON: Yeah, that's the funny thing about auto is that if you look at it, there's all these chips for going into the auto industry, but then the shipments of autos continue to be weak.

    ANDREA LATI: Yep.

    DAN HUTCHESON: So, anything else of interest in, with your forecast?

    ANDREA LATI: No. And again, this was pretty much in line with our previous update. So, it was more, you know, the equipment timing, actually, that impacted this year's numbers. But as far as the overall demand for equipment, it remains very strong. It's just a matter of getting those tools for customers.

    DAN HUTCHESON: Yeah, and I can see on the semiconductor analytics side that the growth is definitely slowing. Yes, it's coming back down again. And it is still way above historic trends, but it's heading it's hunting that trend down.

    ANDREA LATI: Yeah, and that's to be expected because we, you know, we cannot grow it, you know, 25, 30% forever. So, it will have to come close to that trend line. And then the question is, when and how far do we go below the trend line?

    DAN HUTCHESON: So, Jason, what's going on with technology? Where do you see things going next year and what’s this year like? You were telling me about Qualcomm Snapdragon?

    JASON ABT: Yeah, you know, we made a prediction. This goes back several months now. So, Qualcomm has a number of different processor flavours, of course. And we tend to, of course, watch the most advanced the 888 right now that we're looking at. But you know, take a little bit of a step back, they also have a processor family called 8CX. It's their so-called mobile computing platform processor, and their Gen3, just kind of came out. We were expecting it to be fabricated with TSMC’s N7, for a number of reasons. One is we didn't actually think from a performance standpoint, the five nanometre was needed. And of course, I mean, just capacity at five nanometre is, well it’s constrained only so much that each vendor will get allocated to, but we were wrong. In fact, the Gen3, the 8CX, is indeed N5. And, you know, it's maybe not new news, if you will. But it's, you know, it really strikes me how much capacity some of these big players are trying to capture as much as they can for their processes. Qualcomm, of course, being one of the major consumers for TSMC's leading edge. And of course, it also makes me wonder, you know, how much capacity is available at the leading edge for some of the second tier, second-tier players. And my guess is not very much.

    DAN HUTCHESON: Although most of the investment goes into the leading edge so the shortages this year have mainly been in the trailing edge, the More than Moore stuff, which last week just finally went to balance in the More than Moore foundry space, but you know, to me, it also speaks to the fact that Moore's law is not dead. You know, last week, and you know, you see things like this, where they're pushing straight through and skipping, you know, N7 going to N5, tells you that there's still a lot of their value playing leading edge, you know.

    JASON ABT: Oh, yeah, for sure. On the flip side, though, and this is something that Dr. Choe, who's really a domain expert in the memory side, and Dan you had a chance to speak with him yesterday.

    DAN HUTCHESON: Yeah, we had a great, there's a video up that you can watch right now.

    JASON ABT: And one of the other things that he has been working on is kind of forecasting where DRAM cell sizes are going over the next, actually, the next probably 10, almost 15 years. So, some forecasting there. And what we're seeing, or what is forecasting is actually kind of a slowdown in the DRAM cell size shrink. If you go back to 2010, we were at a very steep decline in DRAM cell size. In fact, in the span of about three years, DRAM cell size, it shrunk about 60 to 70%. I mean, it's just unbelievable. And, of course, it speaks to the price per bit. And, you know, we're following how expensive it is for DRAM. Now, if you move forward to today and you take a look at where we're forecasting, that rate has slowed down tremendously. So, where we are today, to do another 60 to 70% shrink, we're looking at seven or eight years out. And a lot of the challenge, of course, is around EUV. So as more and more process steps require EUV and we can see it's quite limited right now in DRAM, but that's going to continue to grow. There is a constraint there. And we're going to see a slower progression, I think, over the next roughly 10 years.

    DAN HUTCHESON: You know, what about the physical? As you know, there's two things that limit scaling. One is, can you make it smaller? And, you know, the other one is, the physics, will it work?

    JASON ABT: Will it work? And so one of the...

    DAN HUTCHESON: How much of its physics because I know there was a real slowdown and DRAM because you couldn't store enough electrons in a capacitor.

    JASON ABT: Yeah. And in fact, if you talk to Dr. Choe, he actually sees a pretty significant change in capacitor materials in the DRAM and DRAM stores charge capacitor, probably towards the end of this decade. So, somewhere between 2025 and 2030, there needs to be a fundamental shift of materials used to create the capacitor so you can actually get, not only get the charge but maintain the charge, and of course, that goes back. There's an access transistor that helps hold that charge in, and, you know, we're expecting that to actually switch to a real FinFET design as we go forward in this decade.

    DAN HUTCHESON: Yeah, yeah. Well, I mean, it goes back to the real part of Moore's laws; it's just about innovation. And it's cool, you know, we've been doing materials enabled scaling, you know, we really dropped off lithographic scaling, after around 90 nanometres or so. And since then, there's been a lot of materials enabled scaling, which this is cool, you know, especially for company at some of our customers who are on the process side; they deliver the materials part piece of this game.

    JASON ABT: Yep.

    DAN HUTCHESON: So anyway, one final note that I forgot to cover was sort of China's getting hot again. Andrea, you brought it up, but SMIC being a...

    ANDREA LATI: Yeah, back in the end of the list. Yeah. And it feels like they're trying to bed any equipment that is capable of 40 nanometres and below, and of course, that throws another kink in the forecast for next year. Because again, China has been a pretty big driver of the semiconductor equipment market. And on both sides, on the multinational sides, that have fabs there and also on the local manufacturers there. So, of course, they cannot get these leading-edge tools that's probably going to, you know, have a negative effect on the semiconductor equipment market for next year.

    DAN HUTCHESON: By the way, the interesting thing I noticed is that SMIC’s stock price varies directly with being on the entity list and being off the entity list, and often in the last week or two, it's been getting beat up pretty badly. So Risto any comments. I know you...

    RISTO PUHAKKA: Yeah, I mean, basically, that the issues of, you know, export licences and things like that is read, let's call it the re-emerging from the status quo that it has been for about one year now. So, you know, the inquiries are kind of increasing in the challenges around it or increasing. It seems to be that the US government is getting a harder position on that.

    DAN HUTCHESON: Yeah. Okay. Well, Jason, all I can say is you're lucky because the technology really does have no national boundaries. An electron is an electron no matter where it is.

    JASON ABT: There it is.

    DAN HUTCHESON: On that note, I want to wish everybody happy holidays. Thank you, guys, for being here and we'll probably come back next year sometime early in the year.

    RISTO PUHAKKA: Happy holidays, everybody.

    ANDREA LATI: Happy holidays.