All right. Good morning. Welcome to the second day of our 54th annual TMT conference. Really excited to be joined by Steve Sanghi and Eric Bjornholt from Microchip. Gentlemen, thank you so much for being here with us today.
Thank you.
Steve, we had a really interesting discussion at dinner last night, I think, about how unique this past cycle was. It feels like we're coming out of it, and we can finally maybe stop talking about it. As we reflect on, again, coming out of the cycle and things sort of normalizing, what were things that you think that's different about this cycle and perhaps are going to change for the industry going forward as we think about future balance between secular and cyclical growth?
I think the post-COVID cycle was an extremely unique cycle. Prior cycles were driven by, like we had a PC cycle years ago, we had a cell phone cycle, we had an internet cycle. There've been various different cycles and semiconductor inventory cycles. The last cycle just was so extreme, where artificial demand became so high, and it turned out at the end that a lot of the demand that the customers asked for were not real. When the customer's business came down, it created a huge vacuum, almost an air pocket for almost every company. It was almost the worst for Microchip because we had a program which was very inflexible, which took customers' orders for a year, and it wouldn't take cancellations or push out, and that prolonged the downturn for Microchip. That was the unique in that cycle.
Where we are today is, for the past two years, the shipments to customers have been lower than what they have been using. Our shipments to distributors have been lower than what they have been selling out. That gap has been closing, and we believe that last quarter, the shipments for distributors and their sales out were fairly close. There's only a few million dollar gap. A long tail of hundreds of thousands of customers is much harder to tell. By just customer count, our customer count is increasing by thousands of customers, meaning that they had inventory, they were not buying product, and they're starting to buy product again. We don't know every customer's inventory, the customer count is increasing, indicating that a lot of the customers are coming back. I think in another quarter or so, that part of the cycle normalizes.
Customers and distributors are buying what they're using, and that part of the business is normal. What you have is the innovation-driven cycle that is creating the growth. During post-COVID, every design engineer was working on substituting parts, whatever they could find, and for a couple of years, there was not so much new design activity. When the parts became readily available in a down cycle, engineers went back to work to design new products and went back to innovation. You're seeing now the front end of all those designs going to production. In industrial, in automation, in aerospace and defense, in automotive and data center. We're seeing a huge amount of innovation-driven growth, and I would say in four main markets, end markets. Clearly, data center, we're seeing it in industrial, we're seeing in automotive, we're seeing it in aerospace and defense.
I want to get to those markets, first I want to basically get through the cyclical discussion. If we think about inventory levels, we like to think about the three levels of on-books, channel, and downstream. It sounds like you're very comfortable with inventory on the balance sheet, at least in dollars, not necessarily days. Channel is lean and end customer inventory levels, you're cautiously optimistic that's 90%-95% done. Is that the right way to think about it, and that in a quarter or two, we'll basically be shipping to consumption?
Yes.
Okay. Got it. Eric, is that correct? That you guys, again, the days are a little high right now, but you want to get back into that 130 to 150-day target on the balance sheet just by basically revenue growing into the inventory in dollars?
Yeah, our inventory days peaked at 266 days a year ago March. We've got that down to 185 days as of the end of this last March. That includes about 15 days of what we'd call last-time buy inventory from the foundries that have end-of-life product that we bought product for the next 10 years. If you exclude that 15 days, it's a 170. We're growing at the midpoint of guidance this quarter, 11%. Days are dropping very rapidly to get in that 130 to 150-day range that we've set.
Okay. Last topic on the cycle, I think. A lot of your peers have talked about raising prices more recently in light of rising input costs, I think also just because the demand has been so strong. I think given what you mentioned regarding the PSP program, you guys seem a little bit more reluctant to raise pricing opportunistically, certainly. Can you speak to how you're strategically thinking about pricing right now?
There are two kinds of price increases. One is increasing the price opportunistically because you can-
when the supply is tight. Historically, in the industry, that has happened in memory and other commodity products. Design-in products, microcontrollers, FPGAs, processors, and others have really, in the industry, not followed gouging the customer on a proprietary product. That is a price increase we are less interested in because our margins are good, margins are still rising. We're going to get to a model fairly soon as the underutilization goes away. The opportunistic price increase to gouge the customer to pad your margin, I think that's probably not a good thing for us to do. The other piece is, if our costs go up by our foundries and our assembly and test suppliers, and we're paying higher wafer cost and this and that and that, then passing those to the customer is something we're more comfortable with.
In the history, semiconductor prices have never gone up on that. Post-COVID period was the first time where inflation just went so high and everybody landed up raising prices multiple times in those years. This is the second time it's happening. Largely, we have been fending the price increases and succeeding so far from our partners. If we are not able to do that, then we'll be willing to raise the prices.
Okay.
As per now.
Are you guys thinking about this as an opportunity to lean in and gain share as your competitors raise prices and what's the customer risk?
Because of the first factor.
If the others are raising prices opportunistically, we're not, we're having lots of their customers come engage with us.
Yeah. I guess on that note, one of your key priorities with your Nine-Point Plan when you came back was repairing customer relationships that had been impaired, partly, I think, because of the PSP program. Where are we, I guess, in that process and is this an opportunity to, I guess, accelerate that?
That process is largely complete. We had a massive program where we visited thousands of customers around the world and that part of the process has gone to normal. If you were to do an unexpected price increase to gauge the customers, we'll be hurting and reversing that process. While our competitors are doing that, we're taking advantage of it. The repairing part is kind of complete. You have to always work with your customers and visit them and talk to them, but there's not a massive repair program that we put behind us.
Okay. Let's shift to some of the secular pieces. I think the one that's certainly on investors' mind the most for the industry, but also increasingly you guys, is the data center piece. Can you walk through, you've got products specifically for data center, but then also, your broader product portfolio that is applicable to data center. Can you walk through the key exposure and what secular opportunities you're most excited about in data center?
We have two kinds of products going into data center. The first kind are the ones which are the products we design for data center and sell only in data center. That number is readily available, and it's clean. That comes from our data center business units that build three different product lines. They build storage controllers, they build memory controllers, and they build PCI Express. The business in them is about a third, a third, a third. The fourth is a retimer, but it goes together with the PCI Express, I put them in the same product category from a product line standpoint. All three parts of those businesses are growing significantly from calendar 2025 to calendar 2026. Those are pure data center products. They only go in data centers, and that part of the business is doing very well.
What complicates the data center exposure for us is all of our other catalog products coming from microcontroller business units, analog business units, security business units, power management business units, that also go into data center, but they also go into the other market. Take a digital power supply product that we could ship it to a customer that builds power supplies. They build power supplies for PCs, they build it for servers, they build it for telecom equipment racks. They don't break it out for us how much goes where, which is what makes this breakdown so complicated. Similarly, we make security chips that secure access to PCs, access to cell phones, access to other communication gear, and access to servers, which again, is multi-market product. I can take a very simple product like a temperature sensor.
A temperature sensor can go into a thermostat, which is a consumer product. It can go into a PC to control the fan. It can go to a server also to control the fan, or it can go to control an air conditioner. You got data center, PC, industrial, home. This is what makes the breakdown very complicated. We have a substantial exposure to data center, but not yet comfortable in breaking the numbers in a way. We could make some wild guesses and we're trying to get some more data from our customers to help us break it out, but our customer base is also very broad and long tail. Any analysis would have some level of plus or minus around it.
Okay.
We're working on it, and we want to share more information, but we don't have it today.
Well, maybe we could double-click on some of the more data center specific products then. You're making a bit of a comeback with your PCIe Gen 6 switch. Can you talk about your legacy and heritage in the PCIe switching market, how you got that IP, why you lost some share in Gen 5, and then why you're optimistic that you're going to regain share in Gen 6?
Our data center business unit was acquired as part of the Microsemi acquisition in 2018. They had done Gen 2, Gen 3 before, and Gen 4 happened on our clock after 2018. Gen 5 happened during the post-COVID years. Microchip misexecuted on the Gen 5. Our part was nearly two years late to the market. Therefore, all the designs went to the competitors. When we eventually came with the Gen 5 part, we got some of the designs, but we only got a second source position. We're shipping today, Gen 5 in volume to a lot of customers, but much smaller volume than we were shipping Gen 4 and Gen 3 because we didn't have a primary position, we had a secondary position.
We made a very strong commitment to Gen 6 and Gen 7, and we leapfrogged on Gen 6, where our competitors' parts are on 5nm technology, ours are on 3 nm . We have 160-lane device, competitors have 120-lane device. We have 30%- 40% lower power, significantly higher speed, and more lane count. We have a leadership part that is currently being looked at every hyperscaler, every enterprise, every server manufacturers, and we have disclosed six design wins, not by name, but there are many other who are looking at it. Our knowledge of PCI Express and our knowledge of these customers and their knowledge about us is not new. We just misexecuted on one generation. As we have a leadership part, it's very well-recognized. It's been written about. The companies who have design win with us, they have talked about it.
They've given us testimonials, which we have shared with other customers. I think our recovery and comeback on Gen 6 after the Gen 5 foul would be very strong.
I guess, can you talk about the competitive backdrop? I think the market has certainly gotten a lot bigger as data center infrastructure needs have grown. I guess, one, are you primarily competing with general purpose servers or accelerated server racks? How has the competitive environment changed since you were more dominant in Gen 3 and Gen 4 versus now?
PCI Express used in service and data access, GPU to CPU connection or CPU to other things connection. As the data center market is exploding, the PCI Express demand with that is exploding. It can be in a scale up and scale out. You're building new servers that will need PCI Express. If you're replacing racks and putting high performance things, which is going to scale up, then you will need the high performance PCI Express in that also. The limitation of memory to CPU, memory to GPU doesn't become a constraint. We're really participating in both scale up and scale out.
Okay. You've talked about the six design wins. I think you've said that one of them alone is over $100 million. What's a timeline in which we can expect that to sort of meaningfully start to layer into the model?
The device goes into production at the end of this quarter, which is another month, and then we start making small amount of shipments, but the major ramp is next year.
Okay.
Yeah. The $100 million design we talked about will be production next year.
You also recently announced your entry into the PCIe retimer market.
Yeah.
Can you talk about where that IP came from? Why the concerted effort to move into that market, and how synergistic is it and important with your PCIe switching business as well?
We have some lower performance retimers in other markets, communication, et cetera. We didn't have a data server class retimer that will go with PCI Express in a very high performance, rebuilding the signal when it deteriorates. This is our first retimer product for the data center market, and it works in conjunction with Gen 6 because as the signal weakens, retimer rebuilds a signal and provides a strong signal to go forward. We basically did it ourself. It's the same resources, data center business units, the design and application engineers that build the part. The part has been evaluated by customers. We have one design win on it. Part only came out of fab about six weeks ago, and it works beautiful. I've seen it and it works great, and I think we'll be very successful co-selling it along with our PCI Express part.
Okay. One more on data center. Maybe it's actually not data center. Earlier this week, you announced a 3.3 kV silicon carbide module. I hadn't historically thought of Microchip as playing into that high voltage power discretes and power semis market. Can you provide some background of what this part is targeting, and maybe your overall exposure to high voltage power?
We started a silicon carbide business quite a while ago, we had grand aspirations in the silicon carbide. When every EV was going to use it, there was a mandate nationally that by 2030, all the cars will be EV. The total size of the silicon carbide market was expected to be just very large. That's when we started the effort in the silicon carbide market. We made some FETs, we made some lower voltage, 700 V, 1,700 V devices. This is the highest one. Two things happened in the last five years. One, I believe it was Tesla that first came out and said that 80% of what we want to do in electric vehicle with silicon carbide, we can do it with silicon. It's not as good as silicon carbide, but it is good.
That brought the market down significantly and crashed the whole silicon carbide. Second thing that happened is, the EV mandate went away and consumer largely did not adopt EVs. They more adopted hybrids and range extenders and other things, and the rate at which the EV market is growing significantly slowed down. That was the second blow to the silicon carbide market. The third was the Chinese overbuilt the capacity in silicon carbide. Therefore, when I came back as CEO a year and a half ago, in my deep dive analysis, we largely took the emphasis on silicon carbide away because of those reasons and combined it with our silicon power module business. We're doing some silicon carbide, but the business unit resources are doing a lot of other things. This is one of the device that came from them.
It will go into EV charging infrastructure. It could go into a data center. It could go into a power management where you want to do really high power switching without loss of power, without heat generation.
How should we think about, I guess, the timeline for this materiality compared to the PCIe business, I guess?
This is not going to be a huge business.
Okay. Got it. Shifting gears, Aerospace and Defense is, I think, an area that investors appreciate your position and high exposure, both from legacy Microchip, but also a lot of the Microsemi assets you acquired. Can you walk through some of the key products and exposures you have within Aerospace and Defense, and what are you seeing in the demand backdrop there?
Certainly. In aerospace and defense, we have three businesses. We have business in Aviation, we have business in war machines, offensive and defensive weapons, and then we have business in space. They happen to be about a third, a third, a third. Aviation business is doing very good. Boeing wasn't building planes for a while when the MAX planes were down, and now they're rebuilding. They got a 10-year backlog, and those planes are loaded with our products and the demand is very strong. That part of the business is doing very well. The offensive and defensive war machine, they had built this ammunition arsenal over 20 years, and they used it up in two months, half of it. Now they're trying to rebuild it. President Trump is asking prime suppliers that build these to ramp the production by 4X +.
I think the number I heard is 4X- 8X, and nobody can get to 4X, but they don't even know how to get to 4X. For that now, I'm talking about primes like Raytheon and United Technologies and Honeywell and others. They're all ramping production. They're asking us. I've spoken to CEOs of Primes. They're asking our capability to ramp- up. We are in every single offensive and defensive weapon. We're in every missile, every plane, every radar installation, every drone, every interceptor. We're the largest supplier to that market, and we have broad penetration with all sorts of devices, from diodes to controllers, to voltage management, to power management, to RF, to everything. We're just huge. That business should do very well. Aerospace and Defense business, it doesn't rise at the rate of the data centers, but it prolongs.
It will happen over multiple years. Some of these high-reliability parts have a long life cycle to build. It takes nine months to build the part sometimes. We're getting orders. Backlog is building. This should be very good business in the coming year, and it will last multiple years. The third piece of that is space. Now, we're the largest supplier of rad-hard, radiation-hardened parts to space. Space has three levels itself. There is a low orbit, then there's a medium orbit, and there's deep space. There's a different level of radiation hardness to go to those three levels of space. We make all of them. World has a new fascination with space.
There was a huge fascination during Apollo times and all that, and then it kind of went away, and we went to Moon 40- 50 years ago, and then we didn't even go close. Now we want to land back on the Moon in the next two or three years. Recent trip we had around the Moon had millions of dollars of Microchip product on it. As we are preparing to go to Moon in 2028 and land on the Moon, I think all the missions leading up to it will have large amount of product in it. Plus, we want to go to Mars. We want to land on Mars. Somebody wants to build a data center on Moon. God love him, we'll have a lot of product in it.
Not very close, but closer to home on the LEO side. There's a lot of positivity now because of the SpaceX IPO. Is that something we should think about you guys participating and benefiting from, given your heritage in radiation-hardened products, or is it a different set of components than what goes further to the Moon?
Some and some. Elon doesn't do anything in the normal way. It's an incremental opportunity, but the difference is the regular space satellites that are launched by others historically are looking at longevity in decades. It should last 20- 25 years. Someday, slowly, it will come back in and burn off, but it will last a long time. These constellation makers are looking at launching these things in volume with a high amount of redundancy, and they don't care if it burns up and comes back in in five years. They'll just launch thousands more. They want to do it cheaply, but they want to do it in volume. They're not using as much radiation-hardened devices. They're using many of the industrial and automotive-grade parts, but using a large amount of redundancy to achieve the performance.
Mm-hmm. Sure.
Therefore, the price levels are a lot lower for the products. The business model is totally different. Business model looks a lot like the automotive industrial business model than the space business model. The part ship there, if we know they're being bought by those companies, they would be an A&D business, and it's an incremental opportunity, but they also buy a lot of those parts from distributors.
Lower ASP, higher replacement cycle.
Yeah. If they buy those parts from distributors, then we sometimes don't know who's buying it.
Okay
or what application they're going into.
One of the markets we didn't talk about yet is automotive. I think you guys are optimistic that your portfolio of microcontrollers, Ethernet connectivity, 10BASE-T1S, are going to benefit from the shift to software-defined vehicle. How much can that growth accelerate versus what it's done the last few years for your auto business specifically?
We're seeing a lot of auto growth right now, but they're not coming from T1S or ASA, because those two protocols are still in design. Our growth is coming from automotive inventory running out, lots of new designs we have won in the regular USB, Ethernet, MOST bus, and others. What's layering on that is we're leaders in T1S and we're leaders in ASA, and essentially what car manufacturers are trying to do is replace all the old protocols of CAN bus, LIN bus, MOST bus, Ethernet, and everything else, RS-232, into a single T1S standard, which runs on a single twisted pair Ethernet wire. See, the past, all these modules were not connected. All these things didn't talk inside the car to each other.
The protocols could be different. When the car is getting all connected, software-defined, then they all have to talk to the central computer and talk to each other. If they're not on the same protocol, then you have to build bridges, a USB to Ethernet bridge, USB to MOST bridge, CAN bus to Ethernet bridge, and these bridges are expensive and heavy and all that. Therefore, there is a drive to take away all of those protocols and consolidate into a single standard, and we're leading that standard. We got leadership devices in there. We have design wins with every major manufacturer, and these are 2028- 2029 production. It will largely begin in 2028. Some could be late 2027.
They will add longevity to the cycle as the inventory-driven growth normalizes, and we're getting growth in lots of other segments, and this will be just another layer. When that begins, it will be a 10-year cycle where these protocols will merge over a long period of time.
Okay. Well, we're running out of time, I'm going to ask Eric two questions at once. Eric, I think you're getting back up close to your 65% gross margin target. Inventory write-offs are pretty much done, underloading charges are going to be here for a bit. Can you talk about the levers to get back to that 65% level? Can you speak to uses of cash right now? I know de-levering is a priority, how much longer you think it'll be until we can start to see some repurchases again also?
The gross margin story is pretty simple at this point in time, right? If you take our gross margins from last quarter and add back $46.6 million of underutilization charges, we're at our long-term target. We're ramping fast. I don't expect those underutilization charges to completely disappear by the end of the year, they're going to be reduced significantly. That progress is just blocking and tackling and growing back into our capacity. From a use of cash, we still are relatively highly levered. We do expect that our net leverage will fall below three this quarter, which is a good milestone for us, there's still work to go.
We want to continue to reduce debt, make sure that our balance sheet is strong and can withstand the next cycle, which there will be another cycle in semiconductors. We're going to continue to generate cash. Earnings are going to improve. EBITDA is growing, the cash generation is going to go to support the current level of dividend and pay down debt.
Okay. All right. Well, gentlemen, we're out of time. Really appreciate you joining us. Thank you, Steve. Thank you, Eric.
All right. Thank you.