GLOBALFOUNDRIES Inc. (GFS) Earnings Call Transcript & Summary

Good morning, everybody. My name is Raji Gill. I'm the Head of Global Semiconductors and Automotive Technology Research here at Needham & Company. Welcome to our 24th Annual Growth Conference. We're very excited to be hosting a fireside chat with GlobalFoundries. With us from GFS is David Reeder, the Chief Financial Officer; as well as Sukhi Nagesh, Director of Investor Relations. So we're very excited to be hosting a chat following a very successful IPO. For investors that are listening on the webcast, there is a question box. [Operator Instructions] I'll be monitoring that throughout the session. I'll relay those questions on to David as well as Sukhi. And then if you have any questions, you want to e-mail me. I'm also checking my e-mail as well at [email protected]. So with that, welcome.

Thanks for having me. Happy New Year.

Happy New Year to you as well. David, maybe we could first touch base on some of the near-term trends that you're seeing. Yesterday, TSMC had a very strong quarter, raised their CapEx guidance, did mention that they're seeing in the industry a higher structural growth. They specifically talked about end market demand may slow down in terms of units, but the silicon content is increasing. And so they've essentially also said that capacity is remaining very tight throughout 2022. I don't necessarily want you to specifically comment on TSMC per se. But as a comparable kind of foundry in a different area of the market, wanted to get your thoughts on how you're seeing the end markets progressing this year and how you're thinking about capacity for the industry.

Sure. Look, I think coming out of fourth quarter into 2022, we're seeing a lot of the same trends that we saw throughout 2021. I think we're going to continue to live in an environment in 2022 in which supply and demand is still out of balance. And what I mean by that is more demand than supply is available. I do think that we are getting healthier as an industry in terms of being able to satisfy some of that demand. But I don't believe that in 2022, that supply and demand imbalance, it's resolved. And so what does that mean for the industry? Well, I think it means that we continue to operate in an environment in which largely most end markets are on some type of allocation throughout the year. I think it's going to be dynamic in terms of who and which markets you're satisfying at any given time, given the puts and takes that are occurring in the kind of global supply chain, if you will. I think it leads to a firm pricing environment. I think you saw price increases from all of the major industry players in 2021, I think, because of the supply and demand imbalance that pricing continues to remain firm. And I believe that pricing gets passed along up and down the stack. I think that you get those increases in indirect costs and direct materials. I think that gets passed along to fabless and IDMs. And I think that ultimately gets passed along to the end market. So I think that's the context for the semiconductor industry for 2022. And I think it's really driven by the macro trends that we're all living in and experiencing today. I mean, look, what's happening in automotive with the electrification of the automobile. That is a large and growing business for us. We'll start it small, but it's getting larger for us and growing. It's going to be a kind of $1 billion-ish franchise out into the longer term for us based on design wins that we have. IoT, I mean, everything connected. In prior years and in prior periods, the most important thing was how fast is my processor? And the conversations that we have today based on everything being connected is not how fast is my processor, it's how good is my connection? How fast is my connection? And then maybe secondary or perhaps primary to that discussion is if you're -- if everything is connected, the majority of those devices are probably untethered to power. And so you've got connectivity and power management that are now kind of on equal footing with respect to characteristics as compute. And so depending on your end market now, that's actually driving what's important. And so I think automotive, I think IoT, everything connected, I think comms, communications and infrastructure, continues to be a growing market and, of course, smart mobile for us or our mobile devices, kind of along that trend with everything connected but more specific to mobility and the actual kind of handset. I think those are all markets in which we play, and it's -- those trends and the growth in those trends will continue.

David, with capacity being top of mind for the entire industry, I was wondering if you could walk us through the current status and future timeline of your CapEx and capacity plans over the next 2 to 3 years at your Singapore, Dresden and Malta fabs.

Sure.

Are there some specific processes that are getting priority? I think it would be helpful for folks to understand that the capacity timeline, which areas you're focusing on, which process node technologies you're currently investing.

Sure. Well, first, let me kind of paint the bookends and then maybe I'll fill the pieces in the middle and if you have any follow-up, I'd be happy to answer them. In 2020, GlobalFoundries shipped about 2 million wafers. And I'm speaking in terms of 300-millimeter equivalent. So anytime I quote wafer quantity, it's always 300-millimeter equivalents, which is where most of our capacity resides. And by the time we finish our capital expansions, we'll be exiting 2023, we're going to be at a run rate of about 3.2 million units, so going from 2 million wafers per year to about 3 million to 3.2 million wafers per year, kind of exiting the run rate. That's not how many we'll ship in 2023, but that's the run rate that we'll be at kind of fully facilitizing and tooling all of our campuses around the world. And so we -- Singapore today is full. There's no room at the end. There's no additional room for equipment in Singapore. And that's why we broke ground on a new facility. That's about 450,000 wafers that starts to come online and really gets to full entitlement by the end of 2023. So that's Singapore investing. That new facility is about $4 billion spread over a couple of years. You've got our Dresden facility there in Germany. In 2020, it produced a little north of 300,000 wafers. That's what it was tooled for. But the physical footprint that exists can actually support up to about 825,000 wafers. We're in the process of tooling that facility in 2021 and in 2022 to get it up to 825,000 annual wafers coming out of that facility on a run rate basis by the end of 2022. That's a couple of billion dollars of total investment to get that facility fully tooled. Again, infrastructure already exists, right? The electrical generation, the gas farm, the actual physical facility. It's adding the tools and some [ in-depth ] process piping to make that happen. And then Malta, our Malta facility shipped about 300,000 wafers again, slightly north of 300,000 wafers in 2020. It has the physical footprint to be able to support about 325,000 -- or 525,000 wafers per year, excuse me. And it will be kind of fully tooled and ramped to that run rate kind of exiting the second half of '23. And so you add all those up, that gets you roughly from 2 million wafers to about 3.2 million wafers. And that has a total CapEx still of roughly $2 billion in 2021, about $4.5 billion, a rough and tough number, I'm rounding a little bit here, in 2022 and then a couple of billion in 2023. And then that fully tools our footprint. And then we start to get some really nice fixed cost absorption out of all those fixed assets.

And then kind of post 2023, the CapEx should start to come down and normalize. And what do you think the more normalized level of CapEx once you're up and running across all of those fabs?

Sure. It's -- our long-term model is 40% gross margin, it's 25% operating income margin. And that includes capital intensity of about 20% of revenue. So you're investing about 20% of revenue in CapEx. And what does that allow you to do? So if you fast-forward to kind of let's start January 1, 2024, if you were to do the math on our current ASPs times that number of wafers, 3 million-plus wafers, you get a company -- and again, I'm liberally rounding here to kind of give you speed and direction, you get to a company that's about $10 billion in revenue. So if you invest 20% of revenue, that's $2 billion a year of CapEx that you can spend. And that's essentially building a new fab, just like Singapore, 450,000 wafers, that's essentially building a new fab every 2 years, which is about how long it takes to then add and onboard that capacity. So it creates this escape velocity for the company that really enables us to grow. And again, given the right demand environment, given the right partnership with our customers and prepayments and long-term customer agreements and given the right relationship with our government partners, it gives us this kind of escape velocity to then be able to invest in a very, very sustainable and accretive way.

Which process nodes or technologies are you currently investing in, in this year? And where do you see that going to the most capacity coming online for you?

Right. Well, we're really investing in all of our key platforms. So we've got kind of six key technology platforms. We've got -- I'm going to lump the SOI together, it's RF SOI and it's FDX technology. That's very -- we have a real franchise there, everything from the majority share in front-end modules and RF connectivity to the growing applications of millimeter wave and automated driver assist as well as in 6G and 5G and beyond. And so we've got a growing franchise in those 2 areas, where we're investing in capacity in both of those areas. We're investing in capacity in our silicon germanium technology. And that's in our Burlington facility. I didn't speak about that one because it's a much more modest investment. Compared to the billions, this tends to be $100 million here or there. But we are growing that capacity based on the demand that we see. So we're investing in our growing SiGe business, which also that's kind of specialty compound semiconductors, also includes GaN technology as well. Our silicon photonics, we're the market leader in that space. And we've got key design wins with the who's who of who's going to win in that market, that end market. And so we're investing in our silicon photonics technology. And then the last two, let's not forget, we've got the feature-rich CMOS piece, where we're not necessarily investing in how do I go to single-digit nanometer technology, we're taking our existing platforms from 12-nanometer all the way really up to about 90-nanometer. But the majority of it stops at 65. Quite frankly, that's where the bulk of it is. And we're adding features to that technology, so those platforms. So we're adding nonvolatile memories. We're adding BiCMOS for power management. We're exploring things like SOIs so that you can have the lowest possible battery usage on all of those platforms. And so we're spending our $600 million in R&D, not in developing a new process node, we're spending it and adding features that the end markets need to be differentiated and competitive. And so we do that both in FinFET as well as our feature-rich CMOS, which the bulk of which sits from 28- to 65-nano.

28 to 65. And a similar question, I think you've answered it a little bit, but I think it speaks broadly to this kind of bifurcation that's existing between kind of single-digit nanometer process node technologies, which TSMC is primarily supporting for data centers and high-speed servers on one end, where smaller process node technology is key, versus more of the mature nodes, where other attributes, as you mentioned, power management, nonvolatile memory, sensors are more important. But there are some concerns -- I would say concern, but there's some discussions around you not pursuing sub-12-nanometer nodes that it could potentially lead to some sort of obsolescence. Can you explain how you're thinking your competitive edge and why your strategy makes sense in this world, where it is being kind of bifurcated between digital and kind of high-performance analog?

Sure. We get this question a lot. So I'm glad to be able to talk about it.

Yes, just to kind of continue to hit that point of -- because...

So if you were to take the foundry market today, the foundry market, about 75% of the foundry market is greater than 12-nanometer and about 25% is less than 12-nanometer, the bulk of which is single digit, right, it's 5-nanometer, 7-nanometer, and of course, 3-nanometer coming up. And historically, the most important thing with respect to the foundry business, again historically speaking, was this progression of Moore's Law. And with Moore's Law, you've got two things. You've got faster compute speeds and you've got cost reductions every time you did a shrink. Well, a curious thing happened. Once you got from about, call it, 28-nanometers, mid-20-nanometer type technology, that type of node level, the cost curve broke. And so you no longer got cost scaling as you continue to migrate down process nodes. And so as you drop below 20-nanometer, now all of a sudden, your cost per transistor for 100 million gates, it actually starts to go the other way and it starts to increase. And then when you drop below 12-nanometer, it's even more exacerbated. And so the cost starts to really increase. And it's both manufacturing cost, it's design cost and it's also, of course, the cost of the equipment and everything else, right? So it's the true manufacturing cost, even fully depreciated, and the design costs. And then -- and so what that led to was that led to people asking the question, "Will I -- am I -- will my end market pay for that incremental cost?" And if you're in a compute yet all-cost market, the answer is yes. So if you're in the data center, right, and you need machine learning and things along those, where you have to grind through billions of computations, you need that process capability. And so the end market will pay for that. And so the most -- if the most important element is compute, then you absolutely move to single-digit nanometer. But if the most important thing for your end market is connectivity, the connection speed is the most important thing to me, you're not talking about single-digit nanometer, I mean, 5G is on 130-nanometer. That's cutting RF -- that's cutting-edge RF at 130-nanometer. And power management, that's not single-digit nanometer, you're at 65 and 90 mostly on power management. Millimeter wave, 22-nanometer technology, there's a little bit in the 45 node, but you're in that space. And so when you look out to the universe today, you have three elements that are really important in design now. It used to only be speed. It used to be compute speed historically, how fast is my processor. Now you've got how fast and how good is my connection. It's my connectivity. It's my power management, and you also need compute. But those three now are kind of on equal footing. If I'm on battery management, if I have an untethered device, battery management is just critical. If I have to be connected, like a blink camera, I've got to have connectivity for security purposes and I have to operate for 2 years on a AA battery and I have to have good enough compute to be able to do some type of recognition with faces and images so that you can then send it up there in the cloud and do the heavy compute there. And so there's more to the world than just compute. And I think that the market reflects that. And that's why 75% of the foundry market is in greater than 12-nanometer.

I think it's a great explanation. If I can just add to that point, the end markets that you are focusing on, and particularly the automotive and kind of industrial areas specifically, the design cycles are also quite long. So these process nodes can exist for quite a number of years without being kind of refreshed or moved down to the smaller node. Just quickly, can you talk a little bit about that -- you talked about power, you talked about RF SOI, CMOS. Can you just quickly talk about the time frame of when those technologies need to move to a process -- a lower process node, a smaller process node vis-à-vis high-speed compute servers or HPC systems, which are -- or application processors for smartphones that are constantly moving to the smallest every 9 months or every 12 months? This is a very different animal.

Yes. Let me talk about it a couple of different ways and maybe sprinkle some proof points in there. Because I think that -- I think it's helpful when you can give like real tangible examples, right? So why don't we start with -- let's start with 28-nanometer and then let's start with image sensor processors. So I mean, I'm sure everybody has a camera, right, or a phone. And now you look at your phone, and I mean, this particular iPhone has three cameras on it, and it used to have one -- and by the way, I'm not including the front-facing camera. So every one of these cameras, right, you have to have the physical pixels, right, the CCD and then you have to have what's called an image sensor processor. The image sensor processor, the most kind of cutting-edge, if you will, is 28-nanometer, but there's also a huge portion of image sensor processor that really sits at 40-nanometer. And you look at, well, what's the limit? What would that end market like look like at way out into the future, right? Because the bulk is still at 40. You've got some at 28. Well, way out in the future, you look at that market and you think to yourself, "Well, I have to marry this image sensor processor up to my CCD and pixels, right?" Well, once you shrink below 28-nanometer, your chip is actually smaller than the CCD and the pixels. The pixels can't get any smaller. The image sensor processor could get smaller, but it doesn't help you anymore, right? And so now you have to get some other benefit. So you had to get some other benefit, like does it help my power management, right? So maybe it makes sense to switch from a CMOS bulk epi wafer to an SOI wafer. That may make sense because that may help your battery management. But the shrink itself will no longer help you in that application. And so that's an application that probably stops somewhere in that 20-ish nanometer, whether you're 28 or 22, but somewhere in that range, because you're limited by the physical size of the actual pixels themselves. And so that would be an example of once you go fully from 40 to the mid-20s, you'd probably work on some other elements at that point. You get more value engineering or cost per effort or benefit per effort out of doing something different other than focusing on that anymore, right? And so that's a real example. I think you look at automotive. Autos today are not shipping because of semiconductors because you're in the single-digit nanometer. You're not shipping autos today primarily because of products that sit at the 40 to 65 and the 90-nanometer nodes and even a few at 130- and 180-nanometer nodes. And so automotive, those design wins are typically about 7 years. And that's 7 years of production. So it takes about 2 to 3 years to take a design win to kind of first production. And that tends to be smaller volumes. And then to kind of get to full production from design win to full production, you're probably in the 5- to 6-year range. And then to get to kind of real true peak, you're probably in a 6- to 7-year range. And then you've got another 4 to 5 years of tail after that because autos stay in the market for a long time, secondary repair market and so forth. And so that's a design cycle that, cradle to grave, probably is closer to 10 to 15 years than maybe the -- a mobile device that may be closer to 18 months to 30 months. And so that's a very long-lived cycle. And the bulk of that technology tends to sit, let's just round it and call it 65-ish nanometer, tends to sit in that type of node. And it's going to be a long, long time for that -- I mean, I'm talking decades for that to migrate to anything even close to talking about single-digit nanometer now. Now you could say the car of the future though, maybe it needs some type of zonal architecture, where it needs some brains that potentially could be more single-digit nanometer that would sit in some of those areas of the car. And I think you don't see that today, but it's a valid point, but that would still be such a small, small fraction of the total BOM associated to semiconductors that it would -- well, I think that it would be drowned out by the things that -- like battery management, for example, and semiconductors associated with battery management. And I could go on and on, but...

Yes, I think that's a very good -- I think those are good examples. Because there are questions online also to that point basically. The questions are 3 to 5 years from these products you speak to may require smaller geometries. You have the road map to retain these products at GF. So there are questions around what the road map is going to look like. But you talked about that a lot of these technologies and end markets are over multiple years. Now does that also coincide with the fact that these customers are entering into kind of multiyear, long-term supply agreements in which the duration of these contracts are 4 or 5, 6 years? So there obviously should be a parallel connection between two, the technologies that you're focusing on, the end markets you're focusing on and then hence, the agreements that you're entering into that require more of a multiyear kind of support. Can you talk a little bit about that kind of dynamic? And the long-term agreements are gaining more popularity across the semiconductor industry. Are there downsides to these type of agreements as well? I understand the positive side. But are there also potential downsides of being almost completely booked for several years out?

Right. Well, let's talk about kind of the genesis of a long-term agreement and how that starts and what the elements of it. I think after our pivot in 2018, GlobalFoundries really became, I would say, a market-focused company. So we didn't pursue a strategy of, well, if you build it, they will come. In other words, if I add a new technology node -- if I just add that new technology node and throw it out to the market and say, "Hey, market, I've got a new node, 12-nanometer, come design on my 12-nanometer." We actually took the opposite approach. We took a very product-centric approach. And we said, "Let's purpose-build a technology to win in an end market." And we've got about 27 kind of unique, we call them battlegrounds that fit within kind of those end markets that we tend to talk about and aggregate into those end markets. But let's take something like automated driver assist, ADAS, what technology do you need for ADAS? Well, first you have to figure out where do I want to play in automated driver assist? Do I want to play on the image side, more on the camera side? Or do I want to play on the radar side? And if you do want to play on the radar side, do you want to play in LiDAR? Do you want to play in millimeter wave? Where can you be differentiated? So we looked at our technology portfolio, and we said, "We like automated driver assist." We believe in the market of the electrification of the automobile and we believe millimeter wave is really what's going to help, amongst many other things that fit into automated driver assist, but we think that's a market where we can be very, very differentiated. And you saw a design win with Bosch exactly to that point on our 22FDX platform. So then we said, "Well, what are the elements that we need?" We took those characteristics and we designed a platform, 22FDX, that has great millimeter wave technology, has very good processing power in terms of digital compute with that millimeter wave technology. And it also has phenomenal battery management because we believe that, in fact, it has a lower battery -- lowest power battery management that can apply not only to automotive but also IoT. And if you believe every car eventually becomes electrified, battery management becomes critical. So you need design technology platforms that will be able to intersect that electrification and criticality of battery management. And so that's an example of like how do we approach the market. Well, what does that approach translate into? Well, it translates into the majority of our business today is single-sourced. In fact, about 80% of our design wins are single-sourced and about 60-ish percent-plus of our revenue is single-sourced. And so what does that mean? That means that a customer engaged with GlobalFoundries, and only GlobalFoundries, specifically for the technology and the differentiation that we offer. Okay. So what does that translate into? So now that translates into they don't have anywhere else to go. We're adding capacity for them and they have bet their in-market future on us delivering. Well, guess what, they want security of supply. And guess what we want, we want security of demand because we want to invest this CapEx that we just spoke about. And that's what led to the LTA. So the LTAs are fixed price, fixed volume, fixed duration. We have more than $20 billion of long-term agreements. We have secured more than $3 billion of customer funding commitments in the form of prepayments and access fees. And that partnership, long-term demand visibility plus customer funding has enabled us to then make the commitment to go off and grow our capacity from 2 million wafers a year to more than 3 million wafers a year between 2020 and exiting '23.

That's very helpful. Just shifting gears to gross margins, your gross margins are rising. You hit 18% last quarter. What are the near-term drivers of this? What could -- or where can we expect to see you in terms of gross margins by the end of 2022?

Sure. So let me talk kind of like directionally year-by-year. And we'll even talk directionally out past 2022. So our long-term gross margin target is 40% gross margin. And long term for us is kind of '20-ish, '24, that type of time frame for us. And so our long-term model is 40% revenue growth, call it, 10-ish percent. 8% to 12%, I think, is the actual number that you'll see. Gross margin, 40% operating income margin, 25%, and then capital intensity 20%. So speaking to gross margin, 2021 was really a year of fixed cost absorption. So when you see our gross margin that's expanded pretty significantly in 2021, the majority of that was on the back of normalization of depreciation and fixed cost absorption. So in other words, take Dresden as an example, they shipped 300,000 wafers on its fixed cost footprint in 2020. It's going to ship a couple of hundred thousand higher than that in 2021. And it's essentially the same fixed cost, only instead of being spread over 300,000 units, it's over 500,000-plus units, right? And so that fixed cost absorption falls straight through in a very accretive way. So 2021 is the story of normalization of depreciation and fixed cost absorption. Well, guess what, we're continuing to fill out the footprint in 2022. So you're going to get continued fixed cost absorption. It won't quite be as great as what you saw in 2021, but you're going to get some continued fixed cost absorption in 2022. But you're also in 2022, you're going to get more than 10% ASP increase year-over-year. It doesn't all happen on January 1. That's not the way the LTAs ramp. They ramp throughout the year and as the capacity comes online pretty methodically throughout the year. But the net effect of ramping those long-term agreements in 2022 is a 10% ASP increase year-over-year, 2022 over 2021. And so 2022, the gross margin story in 2022 is increase of ASPs as well as continued fixed cost absorption. 2023, you get the full year impact of the ASP increases. Remember, they don't all kick in January 1, they kick in throughout the year in 2022, so you get the full year impact of those in 2023. You have a little bit of fixed cost absorption still to come. And now we're mixing up the business in 2023. Because when we strategically pivoted the company in 2018, we were getting out of some businesses or minimizing our exposure to some businesses, and we were targeting specific businesses, like I mentioned in those battlegrounds. And those businesses that those end markets that we targeted were accretive to us. And so you're starting to see the mix impact start to kick in, in 2023. And then by the time you're in 2024, I think you've got a little ASP, you've got a little fixed cost absorption, but 2024 is more a story about the mix of the business and the mixing-up and the accretive value of the mixup in 2024 and beyond.

Okay. That's really helpful to understand the margin. I just want to hit a couple of questions from investors that are online. Is the additional demand, is it going to consumption or your customers building inventory buffered due to the many factors such as geopolitics and pandemic effects? So is demand for consumption or is it for inventory buffer?

Consumption.

Consumption.

Yes. Based on the seat I sit in today, I mean, we're getting multiple calls per week for expedites, for increased allocations. It's going to be a dynamic environment throughout 2022, just like it was in 2021. And there's going to be a lot of balancing in 2022 as customers advocate for their allocation versus others. And so it's -- from the conversations I have, if it's inventory build, they deserve Oscar-winning performances because it's really hard. It's really hard to tell a customer that they can't get the full amount that they need and they tell you what they're losing in sales.

So if they want it, they really need it. They really need it.

If they don't really need it there, they're Oscar-winning performers.

They're good actors, good actresses. Okay. Interesting. We have about 5 minutes left. The -- just a quick discussion on the end markets. Which of your end markets or process technologies you believe are the most underestimated in terms of their growth potential?

Well, I think automotive, we've talked about it a lot. We've got some really critical and key design wins in the automotive space. And as we execute on those design wins, I think you're going to see that business grow into a really nice franchise for us. And there's a lot of things to like about the automotive business, the longevity of the design win and the life cycle of those products. The macro trend around the electrification of automobiles and the amount of content that, that drives on a semiconductor basis, that's a market that is really exciting to us. And we think we've got some unique technologies in that space. And so we're quite encouraged by our progress there. And we're quite excited on what that business can be for us. I think the one thing that I think a lot of investors don't realize is what a franchise we have on connectivity. And that could be connectivity on the WiFi space. It can also be kind of connectivity on the front-end modules in mobile. But if you look at who's gotten majority share in the front-end module space, I mean, that is an area, the RF SOI space is an area that we are just incredibly competitive in. We supply all the major suppliers to that kind of end market. We sell to all of them. And it's just a space where we've really carved out some real differentiation that's unique to GF. And in that franchise, it's half of our total revenue today. It's still growing at a very nice clip for us. And it's an area where we have some real differentiation and we have a road map to keep our competitive advantage there. And so I think if I had to highlight kind of two areas, it would be the end market associated with automotive and the design wins that we've won there and continue to win there, including partnerships that we've announced. And then the other one would be the RF SOI. And then FDX obviously is kind of almost like a follow-on of some of that RF SOI technology because that's a real franchise for us.

And just to follow up on the RF SOI and the mobile phone market, obviously the amount of RF content has increased significantly for 5G phones, number of frequency bands have increased, the bandwidth within those bands have increased, the number of LNAs and antenna tuners have increased. So we saw a pretty dramatic increase in RF content in the last couple of years. How are you seeing 5G smartphones? Do you continue to anticipate that RF content will increase? And you have a very strong position at all the top RF customers there. But I'm just curious about how you're thinking about RF SOI as we get into a higher penetration of 5G smartphones. We have about a minute left.

Sure. I'll try to be brief. I think when you look at the total handset units, right, handset units is relatively flat. I mean, plus or minus a little bit here and there. But it's a big market. It's 1 billion, 1.3 billion, 1.4 billion, I forget what the actual numbers in projections were for 2022, but north of 1 billion units. And I think if you looked at 2021, I think only a couple of hundred million units, it varies by which report you read, but only a couple of hundred million units out of that roughly 1.3 billion units a year were 5G, right? And so the 5G rollout with respect to handset has a meaningful way to go. And as that happens, then you continue to get more and more RF content as you move from 4G to 5G. And look, ultimately, we'll move from 5G to 6G or equivalent. And so I think this progression and this insatiable hunger for faster and faster connectivity across all things, not just mobile but all things connected, all things needing connectivity speed, security of connection as well as power management, those are macro trends that really play right into our wheelhouse in areas where we invest a lot of R&D into staying differentiated in those areas.

All right. Very good. We'll leave it at that. Thank you so much, David. Sukhi, I appreciate it. Best of luck for the rest of the year. Thank you, everyone, also for joining.

Thank you, Raji.

Thank you so much.

Right. Talk to you soon.

Bye.