Good afternoon. Hi, I'm Tim Arcuri. I'm the Semiconductor and Semi-Equipment Analyst here at UBS, and very pleased to have Microchip with us. We have Steve Sanghi, who's the CEO of Microchip. We have Eric Bjornholt, who's the CFO. And we have -
Brian.
Brian McCarson.
Brian McCarson, who runs data centers, so thanks to all of you.
Thank you.
So let me start off. Steve, you made an announcement yesterday, took the fourth quarter, took the December quarter to the high end of the range. Can you just speak to what the drivers are of that and what you're seeing in the business?
Certainly. Before I begin, I wish to remind you that during this presentation, we'll be making some projections and other forward-looking statements regarding the future financial performance of Microchip. These always involve predictions, and the actual results may vary materially. I refer you to Microchip's filings with the SEC regarding some important risk factors about the company.
Having said that, Tim, when we made our announcement in November and had a conference call on November 6, we were really worried about the impact on December quarter from the holidays. Customers, December quarter is historically our weakest quarter of the year, and many year-end customers often want to dress up their balance sheet and not hold a lot of inventory. We were just concerned that the turns required may not materialize, even though lead times were short.
Usually, when the backlog is low and the lead times are low, turns fill up. But I think we were spooked by the holidays. And so seeing what we saw, taking the risk for the holidays into account, we guided the December quarter to be down -1 sequentially. It's usually down -3 to -5 , so it was better than seasonal, but probably not good enough.
As the month has gone by, the results have been very good. November was a phenomenal month. We had very strong bookings, very good turns. I n addition to that, a lot of the customers who had placed their backlog for January decided to pull it in.
We 're getting a large amount of pull-ins, and some of them we're having trouble satisfying because a lot of our inventory is in die, and we still have to assemble and test it. I think with the strength we saw in November, I would say that the holidays now have been de-risked. So therefore, I had to pull the guidance up, and we feel very strong about it.
Now, when we told back in November that our March quarter was shaping up to be very good, a number of investors and analysts asked the question, "What gave you the confidence that the March quarter would be strong? People can always push their orders out." Well, it was a good question, but the reality has been different.
In fact, they have pulled it in, and they have placed a lot more orders where the March quarter has even strengthened further beyond what it was on November 6. G ood this quarter, very strong March quarter, and the June quarter also looks very strong. So I think maybe the well-expected inflection point in the business is here.
Is there any particular end market that's been stronger, or has it been broad-based?
W e're seeing recovery really in all the end markets. We've been talking for a while that our data center and aerospace and defense business looks good. The other businesses, industrial, automotive, communication, they seem to have joined, and now I will put them in the plus category.
The consumer business, which is kind of one of our smaller businesses, was hurting somewhat from the uncertainty from tariffs, but people have finally resigned to the fact that there is no immediate settlement coming on tariffs, and they are really just switching to running their business a normal way. So I would say that business is doing well also.
I think you talked about customers pulling even from June into March. Is that what's making March better, or are these new bookings that are shipping in March, or is it orders that were in backlog that were supposed to ship in June that are actually pulling into March?
They're not pulling in from June to March. They're pulling in from January into December and placing a lot more backlog in March. Despite some backlog pulling in from March quarter to December quarter, March quarter is still stronger. June is just new bookings aging into December, into June.
I think you said November was the best bookings month in three years. Is that what you said?
Yeah. I've said that probably four or five times. September was the best booking month in 3+ years. I think so was June, and so was July. November beats them all by a mile.
Oh, wow. Okay. And is there any sense, Steve? So typically, March is. I think March is usually up a little bit seasonally, 1%-2% is what our math is. Y ou had already said that you expected to be better than that. T his just sits on top of that. So you should be quite a bit better than what?
We should be significantly better than the numbers you just said.
Great. Okay. Let's just talk about the sort of long-term growth rate of the business and sort of what the dynamics are. What do you think changed for customers? Do you think that there was some of this backlog aging that you were talking about coming out of last call, and it was like it was almost as if customers were placing orders on a prospective basis, and they weren't sure what was going to happen because of tariff uncertainty or whatever? T hey were putting orders in the backlog that was sort of longer dated. W hat sort of broke in the month of November, do you think?
T wo things are happening. Number one, the inventory has been coming down. I would say the distributor inventory correction is in the ninth inning, and a number of distributors are really kind of fully corrected, but a few of them have a little bit left. So that's in the ninth inning. So therefore, distributors are starting to buy a lot of product because their inventory is fully corrected.
The OEM customer inventory is probably in the eighth inning, so they maybe got another quarter to go. But these customers, on all their products, it's not, inventory is not same on every product. On many products, inventory is low, so they're buying those. S ome of the strength is coming from people starting to take their purchase equal to their consumption. So that's where some of the strength is coming from.
The second area it's coming from is in 2023 and 2022, most of our customers' design activity basically stopped. Parts were so hard to acquire. They were spending all the time finding whatever parts they could and then substituting the designs in the board with the new products so that they could continue to run their manufacturing. So there was a significant demand destruction over those couple of years because of lack of NPI, new product introductions.
That changed in 2024 and 2025. Our customers have been designing their new products, and many of them are now starting to turn to production, and we'll see more and more of that as 2026 gets underway. T hose two factors are really driving the growth. T hat could continue for several quarters.
E ventually, sometime in 2027, let's say after four, five, six, seven quarters, the growth kind of goes back to what the microcontroller and analog and mixed-signal market growth is someday. W hat we are trying to do is building a new Microchip, which we have tried to do in the last year, in which we're adding three additional pillars to our business other than just the microcontroller and analog and mixed-signal.
And those three pillars are: one is network and connectivity. The second one is high-performance compute. And third one is AI on the edge. And all these three pillars have products and markets which are growing substantially faster than the mature microcontroller and analog and mixed-signal markets are growing.
Our goal is to basically link up where these new accelerated growths begin when our generic microcontroller and analog business is becoming an industry growth so that we can have a larger-than-industry growth for years to come. Pr obably the best, most lucrative part of that business seems like the opportunity in the data center business.
So I brought my data center head, Brian McCarson, with me today, and I'd like to have him tell you a little bit about our data center opportunity. I can come back, and I can talk about the rest of network and connectivity and compute, if you so like. So, Brian?
Yes. So thank you, everyone. Thanks for having me. I'm Brian McCarson. I lead our data center solutions business unit within Microchip. O ur business is broken up into three basic categories. We support memory control infrastructure. We support storage control infrastructure. And our fastest-growing business is around PCIe-based connectivity. T hese are switches and retimers.
Now, if a GPU is going to communicate to a CPU, that language that an Intel CPU or an AMD CPU communicates with is PCIe-based. And so we are one of two major players in the PCIe switch market. W e have now a 4th-generation product. So we have PCIe generation three, four, and five. S teve just recently announced our latest innovation, which is PCIe Gen 6, which is the world's first three-nanometer PCIe-based switch. W e're particularly excited about this product launch because for a few reasons.
One, it shows Microchip at the bleeding edge of data center infrastructure technology with, again, that world's first switch on three-nanometer. T his product also has a lot of differentiating features which make it really interesting in the growing data center market. O ne of those is it has the best power efficiency of any other switch that's out on the market to the extent it could have meaningful savings for an entire data center from an energy infrastructure perspective versus our competitors.
T he second is it's the only product on the market with fully vetted and developed multicast technology. W hat multicast means is your GPU needs to send a signal to 20 different storage banks because it's doing retrieval augmentation for training. It has to send that signal 20 times, one to each of those banks.
O ur technology allows that GPU to communicate to our switch, and then we take care of that one-to-twenty communication. So we're offloading some of the burden that the GPU has to carry, and that's adding a lot of additional incremental value to choosing our product. T his represents, as Steve stated, a shift in Microchip's position in the market where we are also participating in the bleeding edge of the innovation cycle with AI data centers.
W e have additional products which are coming, which will be supporting both enterprise server markets as well. So it's an exciting time for the data center solutions business, and we're seeing some pretty significant interest on these new products.
Brian, can you just talk? Can you give us a sense of how big do you think this business can be over time?
I won't speculate on how big this individual business will be, but what from a number of different analysts have externally predicted the PCIe switch and retimer market to be upwards of $12 billion a year by 2030. We intend to be a meaningful participant in that market with industry-leading products. Of that, maybe 15%-20% is speculated to be retimers, and the rest goes towards those PCIe-based switches.
H ow do you win in this market? This is not a market you participated in much in the past, and you're competing with companies who are fairly strong incumbents. What do you bring?
We did compete in this market. We're, in fact, the only company today that supplies Gen 3, Gen 4, Gen 5, and now Gen 6. Originally, the business came with the Microsemi acquisition, which was only Gen 3 and Gen 4 at that time. So the company and its prior Microsemi, we've been competing in it for years. We didn't do well on the Gen 5 because the Gen 5 product was about a year and a half late to market.
So when we came back, usually the hyperscalers adopt any future generation much faster than the enterprise customers do. Enterprise customers come in about 18 months, two years late. W e had a lot of designs with Gen 5 now, and the enterprise customers came later, but we missed the earlier hyperscalers.
In Gen 6, we are the first to market with the most advanced product, and we're talking to pretty much every hyperscaler except one for some unknown reason. I think many of these will be turning to production in about a year or so time frame. There'll be a significant growth in that market. A fter that, about a year, year and a half later, a lot of the enterprise customers start designing. They usually follow the hyperscalers by about a year to 18 months.
So the inflection of that business is more of a 2027 event versus?
No, it's the second half of 2026.
Okay. Great. Maybe, Eric, I can ask you about I think the goal you sold in about $50 million below in DiSTI sell-in was about $50 million below sell-through last quarter. I think the plan was that it would gradually get better, but it wouldn't get to parity until maybe June or later in the year, actually. Sounds like maybe distribution gets to parity faster now. Can you just kind of talk about that?
Yeah. We think we're getting close. We think it's within the next two quarters that it gets there. I think there'll still be a difference this quarter. D istribution inventory is getting pretty low, and it's not across the board at every part as Steve kind of talked about, but it's getting in a good spot where they're going to have to start buying in line with what end consumption is. I think we're within two quarters of that happening.
Can you talk about factory loadings? I know, I think you had $51 million in underutilization charges last quarter, and you were planning to start to bring the factories back on. Certainly back end, for sure. Can you talk about sort of the trajectory? It sounds like you're going to turn the factories back on a little more, a little faster now, actually.
W e look at capacity really closely within the company on a regular basis. We are producing more out of our wafer Fabs this quarter than we did last quarter. And I expect that progression to continue in 2026. T his $50 million or so of underutilization charges are going to take us some time to work through. There's two major pieces that are impacting our gross margins negatively today.
One of those is our inventory write-offs, our inventory reserve charges, and the other is capacity utilization. The inventory write-offs, which was almost $72 million last quarter, that will get back to normal quicker than the capacity underutilization. It depends on the slope of the revenue curve, obviously. But we added a lot of capacity during the last upcycle, so it's going to take us some time to work through that.
Steve, you talk about product gross margin. I think it was 67.4% last quarter, I want to say. Should we expect it to stay in the 66%-68% range? And does the data center business help maybe bring the product gross margins up?
There are a lot of puts and takes. To do the math on it, you're correct that our product gross margin was about 67.4%. From that, we took two charges: inventory write-off and underutilization. Those charges had a - 10.8% impact on the gross margin. 67.4% - 10.8% gave you the gross margin that we announced, which I think was 56.7%. This quarter, we're guided up to 58 point something.
We're fairly confident that we'll put a six handle on it in the March quarter. From there to 65%, I think it's probably a slingshot. I don't know how long it takes, but it looks very good. Our product mix is very rich. The strength is in data centers, FPGA, communication business unit, aerospace and defense, some industrial.
The mix is very rich, and the strength is in all those products where the gross margins are high. T hat's why you're seeing that basic product gross margin being 67.4%. Also included in that is a licensing business, which is 100% gross margin, and that's very strong. As the foundries are ramping, every single foundry is ramping right now with AI and other things. And if they license on our technology, we're getting royalties on it.
T hat's adding to a very rich gross margin. But what you said is 67.4%, and as the data center ramps, does it go above that? Well, I think our long-term margin target is still 65%. And if we longer term do higher than that, let's get closer before we take a look at it. There are a lot of moving parts in the business. There are parts of our business which are also very competitive. But I think we feel pretty good about gross margin.
Great. Thanks. And Steve, so if you go back and you look at what you think the business grows at, what the long-term categories of the business, and maybe layering in these new data center opportunities, what do you think? And it depends on the base because we don't really know what the right base here is because it's shot around so much. But what do you think the long-term categories of your business?
I think we're not prepared yet. I think we want to probably wait another some time where our business gets back to the baseline so we understand where we are, and then layer in the normal growth in our legacy businesses, microcontroller, analog, and some connectivity, and then link up to the higher growth rate from data center, FPGA, high-performance compute, network, and connectivity, which will kind of have a much higher growth rate, and then blend it all together and see where we are before we can talk to you. That exercise is underway, but not ready to roll out yet.
Great, and then with respect to the Fab shutdown, back to you, Eric, so how much does this play just in terms of where your utilization is going to be, and do you think now, given that the business is actually getting better? I mean, there was some thinking maybe there might be some more consolidation, but now that the business is actually getting better, how do you think about the Fab shutdowns and how it helps margins going forward? Have we seen the impact from it yet?
I want to make sure I understand the question on Fab shutdowns. So we shut down Fab 2, which was the Arizona Fab. That's what you're referring to, correct?
Yeah.
Yeah. So I think we're in a good spot today. I mean, we expanded the cleanroom space significantly, both in our other two larger Fabs in Colorado and Oregon during the upcycle. We've got a lot of equipment that we purchased that is not placed in service yet that we can grow into, and we can take some of the Fab 2, the Arizona Fab tools, and transfer them as needed to these other facilities.
So I think from a capacity standpoint, we are in a really good position. We can't just turn a Fab on overnight, though, right? And so it will be a gradual ramping back into our capacity over time. And that's why I talked about the underutilization charges being kind of a gradual decline as we move through the upcoming couple of years.
Great. Steve, I want to ask you about China. And you have this JV in China to look more like a Chinese company in China, I think you've said. So can you just talk about that? Can you talk about how that's going? And I think you've taken your exposure and you often whittle it down to be, you say, well, by the time that you whittle it down, it's a low to mid-single digit number that's actually at risk of domestic displacement.
We do not have a JV in China. We had talked about a strategy a year ago where we would build a partnership with a company who will buy our die and then assemble and test it in China and then ship into China as the local product. That was when the definition of made in China was assembled in China. Through these various rounds of trade negotiations, China has changed the definition of made in China.
The definition today is Fabbed in China, not assembled in China. We don't do very much Fab in China. We abandoned that strategy quite a while ago, and I think we have spoken to investors about having abandoned that strategy. Since then, we have instituted some other strategies that we are not at liberty to say because of competitive reasons.
There are a lot of ways to have intermediary involved who buy your product, program them, write some software, put it in the boards, and then sell it to the customers as local product and things like that. We're doing various different things to be able to have the customer think customers just want to check a box made in China for Chinese government. They don't want to buy the local product. They want to buy our product, and we are basically helping them do it.
Our business in China is not suffering. We've been getting the same question in China for more than a decade. There's not a meeting where the China question doesn't come up. Our China business is solid. It's growing. It's doing very well year after year every year. Our percentage of China business has not dwindled. Chinese threat is largely overplayed. I think Chinese competition is in a very low-end commodity area. They don't largely not make the kind of products we make.
And how much do you think, Steve? How much do you think the microcontroller world is still benefiting from Moore's Law? And does that have some effect on limiting the TAM at all? And can you talk about just how much do you think that the MCU business is actually growing?
M icrocontroller business is really not on Moore's Law trend, really, and really hasn't been. There's no microcontroller, generic general purpose microcontroller, which is at three nanometers. I think the center of gravity, or rather than the center of gravity, let's say the most advanced general purpose microcontroller is probably being built on 22 nanometers, which is many generations behind the three nanometers. It is on a treadmill where parts go on more advanced technology every couple of years, but it's not competing with our data centers and GPUs and CPUs.
Right. But I mean, just that treadmill, do you think that that's in some way capping the growth of the market at all?
I don't think so.
You don't think so, and maybe you can talk also about your competitive position in that market. I get questions. Some investors perceive you as overexposed to 8-bit. This is not my words. I'm sure that you get the same questions as you overexposed to 8-bit and underexposed to 32-bit, so when you get these questions, how do you respond to them?
Well, I mean, I think we've been answering it for a long time, but anybody can believe what they want to believe. 32-bit is our largest business, and 8-bit is the second. As part of the restructuring after I came back, we combined 8-bit and 32-bit together. N ow it's really one single business unit.
We go to customers with this barrage of products, and they can design whatever they like. They can design our 8-bit. They can design our 32-bit. They can switch back and forth. But I think what it gives us is there are competitors that only have 32-bit. So if you're a hammer, everything is a nail.
C ustomers have low-end applications where they want a very small footprint, near zero sleep power, very small pin count, and just a tiny bit few instructions to be able to just wiggle something or move a fan or do something a little bit. T hey don't need the complexity of 32-bit. They don't need to connect to internet or Wi-Fi or Bluetooth or anything else.
M any of our competitors don't have a solution for them. We have a solution for a toaster, a blender, an iron, a garage door opener, as well as servers and automobiles and high-end robotics and everything else. W e are full-purpose microcontroller supplier. I think we're proud of what we do, and we have succeeded for 35 years and continue to do well.
Do you think as you push more aggressively into these new opportunities, such as data center, does that increase the amount of OpEx and the amount of R&D that you're going to have to spend?
No, it doesn't. I think per product, per invention, they also have much larger revenue. So look at, for example, when you design an analog product, the revenue per product is quite small, but the investment to make that product is quite small.
In microcontroller, the revenue per product is much larger than analog, but the investment is much larger because you need software, development tools, and all that. When you get to a product like Brian's, the revenue per product is huge, and the investment is huge too. So I think in terms of percentage of OpEx from the revenue, I think it scales appropriately.
Then, Steve, last question, and I asked you this on the call last time. You still have these LTSAs, and your point was these are still around because it helps you stay connected to the customer. Can you give any tangible examples of maybe a situation where you have a customer sitting in LTSA that you're able to then go and say, "Well, I can sell you something else because we have this kind of capability"?
I think we did these LTSAs, which stands for long-term supply agreement, at the height of the market when the parts were very hard to get, and what happened is customers gave us a significant amount of money and made commitment to buy a certain number of unit volume or revenue dollars per quarter for five years.
W hen they buy the quarter's quantity, then they get one-twentieth of their money back every quarter, and in five years, the money would be returned. Number one, two or three years have gone by, and most customers have gotten half of their money back or more already, but the second thing that happened is pretty much no customer met their target of what they were going to buy.
And what we were doing prior to I came back is that we were forcing customers to buy what they signed up for. And that's what resulted into a large amount of inventory. We were forcing people to buy what they didn't need and distributors to buy what they didn't need. So we got rid of all those rules. So today, we still have some of customers' money, but we give that money every quarter, almost irrespective of what they buy.
And we're letting them buy what they need. So we're very flexible. But we still keep a little pressure, "Hey, I'm helping you. You made this commitment, and I'm relieving you of that commitment, giving you still the money back. But you should give me a preference for design." And we're getting that. We're getting that.
I mean, just recently, just in the last two weeks, there is a very, very major design on a T1S networking product at a major customer in Europe, and it's a whopper of a design we were heading towards, and we would not have gotten it if we didn't have the money.
Wow. That's great. Well, we're out of time. Thank you to all of you.
Thank you very much.
Appreciate it.