Micron Technology, Inc. (MU) Earnings Call Transcript & Summary

Got it. Got it. Now this is a very helpful perspective, because I have not always been a believer in how critical DRAM is getting to be as we go into higher bandwidth and the compute as it continues to go up. So that's awesome. Let me switch to a topic which is very close to my heart, and I'm sure to many, many investors is your dividend announcement. You have addressed it. Investors have been asking you this. So just maybe just help us understand the evolution in the thinking. One of the reasons that you highlighted during that call was you also mentioned your confidence around the supply-demand dynamics, which I think we are worried about near term as we should or we shouldn't. But just thinking out, just help us understand thinking that went behind it, the confidence in the business model, and then the factors that led to this decision. Because I'm sure this was not something that you just said, okay, we're going to do it. You've been thinking about it. It's a very deliberate decision. So I would love to get more color on that.

Absolutely, Ambrish. That's very right. You said it well. We have thought about this very deeply and for a good length of time because we don't take these decisions lightly. These are very long-term decisions. They have long-term implications for our shareholders and for us in terms of how we manage the business, how we think about the business. So I would say the most important thing is as we assess the dividend and really our capital allocation program when it comes to returning cash to shareholders, we thought of that in 2 ways: One is the opportunistic buyback and the other is the dividend. And as we thought about both of those aspects, we consider the fact that Micron has never been stronger in our history from a strategic, operational and financial perspective. So when we look at those 3 pillars, strategically, operationally and financially, we have never been in a stronger position. And if I just double-click on each of these, certainly the industry demand is now being driven by a multitude of end markets. Interestingly, automotive and industrial are the fastest-growing end market that we have over the next decade. The cloud and data center market overall is very strong growing as well. And these are pretty important secular trends. We haven't even started seeing the killer apps that ultimately will come up in 5G. So if you look at all of the demand environment, very, very strong and robust. The DRAM industry, as you know very well, is pretty consolidated with 3 players accounting for 95%, 96% of all of the big output. And it is a tremendously profitable business, very high free cash flow, very high margins through the cycle. And even at the low points of the cycle, it is an intensely profitable and high free cash flow business. Even at the low point of the cycle, even at the high point of the cycle, of course, dramatically so in terms of profitability and free cash flow. So when you look at through-cycle dynamics of DRAM, it is an extraordinarily strong place. And DRAM, as you know, is 70% of our revenue, give or take, much more of a percent of our overall gross profit. And so that is in a really strong place. And of course, those things are not possible if we don't have disciplined supply growth. And we believe that there has been disciplined supply growth in the past, which allowed us to post these kind of numbers. And of course, we'll talk about the operational aspects, which contributed to these profitability numbers as well. But we also look forward and see that the supply growth is going to continue to be constrained. And it goes beyond the industry discipline that we have observed. But more importantly, I would say that as Moore's Law slows down, as scaling slows down, less and less of the annual bit growth of supply comes from tech transitions. More and more has to come from additional wafer supply. And wafer supply is: a, difficult to create; b, is also more expensive. And so there is a natural desire on part of any producer to ensure that there is very high ROI on those extra dollars to put in new capacity from new wafers. And that has a natural limit on the -- creates a natural limiter on the amount of supply growth that happens. And I think we have strong confidence in that view that the supply growth will remain disciplined, and it will remain a very strong business. Operationally, we have done -- and again, I'll just go through this very fast because we have mentioned this in the past. But we are now leaders in the world when it comes to first to 1-alpha node in DRAM, first to 176-layer NAND. So we have technology leadership, world's fastest graphics memory, highest percentage of QLC NAND, which is the most cost-effective NAND you can have, highest percentage in the industry on a broad-based basis across all the segments. So a lot of data points that point to the Micron broad product portfolio as well as technology portfolio coming together in a way that positions us very, very strongly for the future. And we've also made tremendous advancements in our operational capabilities, both on the manufacturing side, supply chain side. All of that is back end, front end. It's working dramatically better. And more customers rank us #1 in quality than all of our competitors on the DRAM side put together. So these are some really strong foundation for us to build on. And then as we look at the balance sheet, our balance sheet is super strong, significant levels of liquidity. The dividend that we initiated is a very small percentage of our EBITDA. We expect that EBITDA to grow over time. Our operating margins and free cash flow through the cycle are very strong. And one important point I want to emphasize is that as we look ahead to the next decade, we think that the trends will continue for memory and storage to grow faster than the rest of the semiconductor industry. Memory and storage, meaning DRAM and NAND combined, was just around 10% of the industry in the year 2000. It's approaching that 30% mark in the current time horizon. And as we look ahead, we expect that it will keep increasing as a percent of the semiconductor industry. So today, almost 30% of the industry is DRAM and NAND. So we believe that memory and storage will keep growing faster than the overall industry. And as we think about the overall industry being an attractive industry because it outgrows world GDP, you can think of memory and storage being the fastest-growing major segment of the semiconductor industry. So it's a very good place to be that tells me that between all the demand underpinnings and the constrained supply, we should have strong cross-cycle profitability, cross -- strong cross-cycle, free cash flow, we should be able to continue to, over time, grow the dividend. And very importantly, I wanted to mention that even though we have created the dividend, bulk of our return of cash to investors will come in the form of buybacks. And the reason we are going to allocate the bulk of our capital to buybacks is because we believe that the stock is extremely undervalued, and we continue to be strong purchasers of the stock through every cycle. Of course, we are going to make it more opportunistic, as we said, because sometimes that discount to intrinsic value becomes even bigger. And we are going to take advantage of that, even in our current fiscal quarter, we are strong purchasers of the stock. So that's how we think about this over the long term.

Great. Great. That was going to be my next question on overall capital allocation but you answered that. Yes, the structural profitability of the company versus before you guys got there, it's not debatable, at least in my mind, and what you guys have done through on the free cash flow front. So that's -- thanks for sharing more details on that. I wanted to transition to the cost side. And there was some disappointment. So higher level, I want to go into technology transition, both for DRAM and for NAND, but maybe start off with the nearer-term question is there was some disappointment and you're not able to get the cost down. Just help us understand -- my understanding is it was -- is, that it was mostly on the back end. So kind of how should we think about cost down medium term? And longer term, what should be the cadence both for DRAM and for NAND?

Yes. I think that's a good question, and that's an area where we could have, perhaps, communicated better in terms of the cost dynamics that we are seeing. I think I should, for the purpose of simplicity, break up the cost into 2 separate buckets. One would be sort of a like-for-like cost decline. And then the other is the changes that happened to the cost structure based on portfolio level changes, right? And when we think about transitioning the portfolio to higher-value solutions, whether it is selling more SSDs instead of selling components in NAND, or to selling more higher-value DRAM, whether it's graphics, DRAM or other types of DRAM instead of selling simply component DDR4. Those types of portfolio level transitions come with higher costs because, as you can expect, and SSD is higher cost to build than just simply shipping a NAND component. But the reason I want to separate that kind of discussion from the overall cost picture is because those things that are higher cost because they are higher value solutions also allow us the opportunity to create higher gross profit per bit that we sell. And so that higher cost is more than offset by the higher ASP. And we end up with a higher gross margin percentage as well as higher gross margin aggregate dollars by making that shift. So those are positive kind of shifts that we make in our portfolio. And those are things that we not just embrace, but we drive actively to move our business into areas where we have higher gross profit potential, and we can take unusually high share of the profit pool of the industry. And so we don't think of that impact on cost as a negative at all, even though it has a negative impact on cost because it has a positive impact on gross profit and gross profit margins. So separate from that is the like-for-like cost reductions, which is driven by our tech transitions and things like that. And there, we have very strong capability with our industry-leading 1-alpha, industry-leading 176-layer NAND. Our 1-alpha and 1Z is going to be the majority of our production by the end of this calendar year. Our 176-layer NAND is going to be the majority of our production ending this calendar year. So I think these things are going to be strongly underpinning our like-for-like cost reductions, which are going to be really -- very, very competitive, and we feel very good about our cost reduction capability. Now when we put those things together, of course, the cost -- the net cost reduction becomes shallower than what it would have been on the like-for-like part because these higher-value solutions in both DRAM and NAND have higher cost structures. But like I said, because they have higher profitability, they have a positive change in the portfolio that we are actively driving. So I think there was a little bit of -- we could have communicated that aspect better. I think the other aspect that is impacting cost, but it is not a Micron issue, it is really an issue for the whole industry, which is all of these shortages in components on the semiconductor side, logic ICs and PCBs and substrates and all kinds of things that are in the industry are certainly preventing all of the semiconductor companies that build products from getting the cost reductions year-on-year that they otherwise would have expected to drive every year in all of those BOM cost components. And so this is not something that is a Micron issue, this is an industry-wide issue. It is impacting the cost structure of not just Micron, but every company out there that is using these parts. And so certainly, that is an aspect that is impacting costs but not unique to us. So I think with that, hopefully, that clarifies that our cost outlook is in much, much better shape than what probably was a takeaway from the last discussion.

Okay. And longer term, what is the cadence that we should be expecting for both for NAND and on the DRAM side?

I think on the DRAM side, we're still looking at high single-digit cost decline going forward as a like-for-like sort of basis on the front end. And I think in terms of NAND, it ought to be in that low to mid-double-digit range in terms of cost decline. Some years, it could be mid-teens, some years slightly below. But it depends on how the scaling of NAND works out in the future because scaling is a challenge, again, all across the industry, not a Micron statement just for everyone.

Yes. Got it. So topic that I wanted to go in a little bit more detail is the EUV insertion. So A, can you just help us understand the timing of the update that you provided us? And before that, the multi-patterning was the path that you were going on. But I'm sure you guys were concurrently working on EUV as well, which may have been misunderstood by some folks. So A, help us understand your thinking on EUV. And there's an investor concern, how do you address that, Sumit, is that given that Samsung is ahead of you, SK hynix theoretically as well. So how do you address the concern that by the time you come up, you've done a great job in really narrowing the cost over the years that you typically have had, cost to set vantage, so you closed that gap. So how do you address the concern that, what, now you're going to be behind in EUV and that gap could expand again?

Yes. In terms of -- it's a good question. Like you said, we have been working with EUV tools for many years. This is not something that is new to us. And as you can expect, we work on future-looking technology-related aspects, whether it is new tools like EUV or new materials that we introduced into our process technology road map. We work on like a 10-year road map for these things, long in advance of actual insertion of these capabilities. And so -- it is something that we have been working on for a long time. We have a good understanding of the cost structure for EUV. And we have also perhaps the most advanced multi-patterning capability in the industry in DRAM. And so what we have tried to do is to create a road map that inserts EUV in the road map at the point in time when it is most cost effective to do so because our multi-patterning approaches give us very strong capabilities to continue to drive competitive cost reductions in DRAM. And EUV still has more to mature. It is still not exactly where we would like it to be from a productivity perspective, but it is now at a point in its maturity curve where it is appropriate for us to drive that into our active manufacturing plans. And so we do have a plan to be involved in production with EUV in 2024. And Scott has spoken about that, our EVP of Technology and Products, Scott DeBoer, has spoken about this that in our -- in terms of the considerations, the time line. So we will be doing that in the 1-delta node in 2024, we'll be able to use it in fewer layers in the 1-gamma node before that. So it's something that we feel very good about in terms of our road map, in terms of our EUV usage, in terms of the ability to continue to drive down costs on a very competitive level, both in the interim as well as beyond that point. So we are very confident about that. And that's something we'll continue to focus on executing well going forward.

Got it. A couple of questions. Could you comment on the NAND side, what you're seeing on the supply-demand side? And over the last 2, 3 quarters, you guys have been signaling -- not signaling, you have been talking about what has transpired. DRAM was very early in the financial discipline in executing on that, and NAND was a little bit behind. So just -- what is the current state of affairs on the NAND Flash market?

Yes. On the NAND Flash side, like you said, I mean, the DRAM market was the first one to inflect higher in terms of pricing and so on. And so NAND Flash has improved over the last several quarters. Last couple of quarters, we have posted improved gross margin, improved ASP trends. But as we have said in the past, NAND has more suppliers than DRAM does obviously. So DRAM has 3 accounting for 95%, 96% of the output, but NAND has 6, with Intel's exit. And once Hynix consummates that transaction, it should go down to 5. But it is still a little bit trickier on the sustaining of balance there. And we have mentioned in the past that there is more supply discipline needed to bring that business to a point where it is sustainably robustly profitable. It's not there yet from an industry perspective. And from a Micron perspective, what we have been very focused on is leading the industry with technology, which gives us the lowest cost structure, which is what we are doing with 176-layer NAND. We're also leading with QLC technology, which is the lowest-cost NAND in the industry. And we reached a very important milestone of more than half of our client bits shipped last fiscal quarter were with QLC technology. So these are very important drivers of our overall cost structure and profitability. Our profitability in QLC is much higher than that on TLC. And so these are very important changes for us. And this is how we drive to our portfolio to become a bigger part of the profit pool, even if the profit pool is challenged or impacted by the industry framework. Now over time, obviously, we do expect that the NAND business will keep growing. It is a growth business. And hopefully, with disciplined supply and the industry environment will be more robust from a profitability perspective.

Great. We are about done with time. Sumit, any closing thoughts you'd like to leave us with?

No, I think, certainly, we spend a lot of time talking about some of the short-term dynamics and there'll always be short-term dynamics in the market where they're up or down. I think the more important thing from the way we look at the business that I can emphasize is we feel like the cross-cycle capability of our company to deliver robust operational margins as well as free cash flow is on a very good trajectory and the industry is continuing to show great promise to grow faster than the rest of the semiconductor industry. The memory and storage part is in great shape. And if you are a believer in the data economy, AI, machine learning, 5G, all of these things, then certainly, we find ourselves at the crossroads of all of those big growth trends. So that's what I'll leave you with.

Great. Great. Thank you. Good long-term thought to leave us with. I appreciate your joining us and look forward to catching up with you again. Take care. Thank you, everybody.

Thank you, Ambrish.