Okay. Good morning, everybody. Thank you for joining us at the Mizuho Tech Conference today. I'm Vijay Rakesh, the semis analyst for Mizuho, and joining me today is Richard Simoncic, Chief Operating Officer at Microchip. Just to give a little background, Richard joined Microchip in 1989, became COO in April of 2024. Mr. Simoncic oversees Microchip's worldwide operations and is leading the company's AI-related efforts. He founded Microchip's internal analog business in 1998, grew it to $2 billion. Obviously, huge legacy at Microchip. Rich's 35-year career at Microchip tells a story of sustained innovation and strategic growth. He has been key to developing the AI strategy at Microchip. As Microchip continues to power everything from automotive systems to IoT devices and mixed-signal analog and flash IP solutions, Richard has been very instrumental to all that. With that, let's welcome Richard Simoncic. Thank you for joining us, Richard.
With that, let's get started. Okay. Richard, I want to go back to the earnings side, obviously. Last quarter, you guys talked about, when you look at the Microchip business, March quarter was the bottom. You talked about the book-to-bill going back above one and looking much stronger. Can you talk to how that's trending? Is that demand continuing to improve across the multiple markets? Is that the trajectory you see into June and into the second half of 2025 in terms of bottoming the March quarter and kind of coming back up?
All right. So, before I start, let me get my legalese out of the way. Thanks, Vijay. I'll start by giving the typical disclaimer that during this course of this discussion, we may be making forward-looking statements about the future financial performance of Microchip. And I'll refer you to our SEC filings, our 10-K, which was filed last week to identify more important risk factors for the company. Now that that's out of the way. Book-to-bill for the first quarter, March quarter, ended about 1.07. Our bookings started to increase. The tail end of December really started to improve going into January, even before the so-called independence or transition day in the February time period. It has continued to improve. In fact, May bookings was the highest that we've seen in several years. Bookings have continued to improve. Book-to-bill has continued to improve as well.
Is that a commentary across all the markets, like automotive, industrial, kind of seeing?
Yeah, we're seeing it come across from all different marketplaces. We're starting to see that a lot of our customers' distribution inventory has really started to bottom out. People are having to replenish now.
Got it. How about when you look at geographically, China versus the U.S., obviously, with the liberation days, there's been a little bit of a flux in some of the issues. How do you see the trends geographically between China and the U.S. in terms of the bottoming, I guess?
Within overall, probably the best way to look at it is distribution inventory. Distribution inventory is sitting at around 33 days now. Our typical range on distribution inventory is 28-35 days. Right now, it seems like it is sitting in a pretty good spot. If distribution inventory is sitting there, then our direct customer inventory, 55%, is probably sitting in the same range where it has come down. We are starting to see that signs of expediting or running out or pull-ins from various customers.
Got it. I want to get to your bread and butter, I guess, manufacturing, the manufacturing footprint. Your 9-point plan actually talks to resizing the manufacturing footprint, understanding that 40% of Microchip is in-house and 60% is foundry. Can you talk to what the footprint looks like now versus, let's say, the peak, like two, three years back when we were in that COVID, kind of in that energy hype phase, I guess? How does the footprint, manufacturing-wise, capacity-wise, et cetera, look now versus that?
The footprint is still relatively the same, about 60-40. That has not changed that much. What has changed is more of assembly and test have come in-house. We have taken control of much more of the assembly of our products. We are probably about 80% is assembled in-house and about the same range, 80%-90%, depending on which SKUs we are running, are now tested in-house as well. A lot of the back-end manufacturing has now moved in-house.
Got it.
Which serves us well as things start to turn around or improve. The first place that usually becomes tight in capacity is usually assembly and test.
Yeah. Typically, the back-end is where there's a lot more cost involved. To bring it in-house is very marginally accretive for you, I would assume, right?
It is accretive, but more importantly, what it does is allows us to control that capacity. That is one less thing that we have to worry about. It is typically, like I said, the first thing that gets affected on turnarounds in the cyclical semiconductor industry.
Got it. I want to go to the product diversification side, which I know you love to talk about because that's your goal now with AI. We'll come back to microcontrollers, I promise.
I've been highly involved in setting the technology strategy for many years at Microchip.
Sure. Talking about product diversification on your FPGA side, the PolarFire, I think you guys talked about June quarter trending higher on the FPGA side. Maybe you can talk to how the bookings look. What are the markets you are going after on the FPGA? You have machine vision and robotics and IoT. Where are you taking that business from, some of those markets, even in defense and military, et cetera, too, right, on the FPGA side? Maybe you can talk to what the roadmap there is.
The roadmap for quite some time, when we first acquired Microsemi, they were largely focused on the aerospace and defense market. There were two key attributes of that product line that were important to medical and industrial markets. One is overall power consumption is about half of any other FPGA manufactured. They also had tremendous security attributes, which also became very important to communication customers. Anyone connecting anything in data centers or high-security areas. We have taken those two attributes, leveraged them in medical accounts, industrial accounts, communication accounts, and more than doubled that revenue since the acquisition of Microsemi way back. That continues to grow. Typically, a customer would look at an account and choose for a very high-end FPGA. They choose an Altera or a Xilinx.
They'll choose Microchip as a secondary for security features or low power in various applications.
When you look at that FPGA market, obviously, there are, like you said, data center. And then there is kind of the other side of the spectrum, which is consumer. Which markets are you kind of trying to go after? Where do you think Microchip has an angle to kind of get in and establish a footprint, I guess?
The biggest thing that we're seeing is industrial control. It's factory inspection, Industry 4.0, optical recognition. That's probably our largest markets for FPGA in terms of some of the fastest growing.
Got it. When you were part of Microsemi, I guess it was small, but you obviously doubled it, as you said. Was it kind of somewhere in that 10% range for you guys? You think that's where a lot of the R&D and efforts will be focused?
Yeah.
It's good margins too, right?
Yeah. It's sitting just a little bit over 10%, very high-margin product line, and is growing very well. We continue to invest in it. We are bringing out even higher-end products on a 12-nanometer FinFET technology and some lower-end products that we are working on too, to have the full range of product technology there.
Got it. Just keeping on that diversification roadmap, one of the products that you guys go after is PCIe retimers. Obviously, all the semi investors are pretty in tune with this retimer and PCIe roadmap because of how much all the AI server boards and data centers are going after, whether it's scale-up or even top of the rack, NIC cards going into PCIe. Can you talk to what the roadmap is there? What do you bring to the table, let's say, on the data center AI side? How do you look to break into that market, I guess?
That's like breaking into that market. When we acquired this from Microsemi, from PMC-Sierra days, they've been in the PCIe switch market for many years through PCIe 4, 5. PCIe 6 is the next one that we'll be sampling here before the end of the year. Where we win on PCIe 6 is we have the lowest power per channel on that, again, focusing on that low power aspect of it. We're working with a number of clients to design in or give them preliminary information to design in our PCIe 6 switches and retimers. We've also taken some of the older versions of that technology, PCIe 3 and 4, and we've reintroduced those products in a much lower lane count to anywhere from 5-12 lanes for automotive and industrial control.
As you start to advance into more AI applications, you need higher speed data. PCIe is lending itself very well to humanoid robotics and robotics. We're on reference designs with NVIDIA and Qualcomm on that industrial and robotics and humanoid robotics side where they're using our, we call it Stanford or PCIe Gen 4 switches, and in industrial control.
Got it. You mentioned low power. Obviously, that's a criteria for both the consumer side, but also when you look at AI racks, they're trying to lower the power.
They're all trying to reduce the power per lane count.
Right. And so, when you look at your PCIe products, do you actually compete with somebody like a Broadcom or Astera, or how do you fit into that?
We compete head-on with Broadcom and Astera on those product lines.
Okay.
Those are the two main, there's three main suppliers within PCIe switches, 5, 4, 6, and that's Astera, Broadcom, or Microchip.
Got it.
Depending on who you're aligned with and who you work with, those are the three companies you buy from.
Interesting. Do you work on the consumer side too? You mentioned some of the PCIe 2 and 3, 3 and 4, where you can reduce the lane count. Is that more of a consumer strategy there?
Yeah. So, that's automotive. Automotive is a large area where when you start to go to software-defined vehicles, the old communication buses that were in cars are not good enough. You have to upgrade to PCIe. ASA is another Ethernet-based protocol, or T1 or T1S is a two-wire-based 48-volt Ethernet protocol.
If you could size that business, when you look at the PCIe side, how big is it for you? How does that split between data center and automotive now, and where do you see that going, I guess?
On the PCIe Gen 4 switches, those are all in design in. So, that's all new revenue streams coming into Microchip.
Interesting.
On the overall switch tech and PCIe switch technology, we don't break that out for vendors, but we're in there with the top three for supplying that marketplace.
Talking about the auto industrial side, I think the other market that you guys have been going after is 10BASE Ethernet. Again, that looks like more like automotive, kind of replacing the CAN, FlexRay, kind of.
It's interesting. It was targeted at first for automotive when we came out with it. The first design wins, though, are actually in industrial control. Two of the largest industrial control manufacturers have designed this into several product lines. That was with the first big design wins for that product line. Now we've got our first automotive companies that will be designing in our product across multiple platforms.
More like a factory automation side, kind of?
On the T1S side, those are more in data center, some of the industrial customers that are building on data centers. On the T1S side, one large automotive manufacturer, we just finished almost a week of training where they had about 125 engineers in the company training with us on using T1S in automotive backbones for cars.
Got it. When I look at all these three markets, whether it's PCIe retimer or the FPGA or even the Ethernet side, they carry probably much better margins, product margins, than traditional analog microcontroller side, I would think. How do you size that business? What's the growth? How do you see that kind of driving the broader COP margins for the company, I guess, if you look out?
We're seeing, when you look at Microchip today, you have different categories. You have microcontrollers, you have analog, you have connectivity and networking, you have compute, which is FPGAs and 32-bit NPUs and 64-bit RISC-V NPUs. The last category is really AIML, right, in terms of how we're enabling, whether it's model zoos or accelerators or we're attaching to other GPUs that are out there as another segment of growth for the company. Right now, probably the biggest discussion with almost every customer out there is that next generation of connectivity. Even though Microchip is known as the microcontroller company, and I met with an investor last night, and we started off the conversation at dinner talking about good old 8-bits again.
We're coming to that.
I was like, I'm sitting there and I'm thinking to myself, we are so much more than 8-bit microcontroller customer supplier. It was fascinating. A lot of customers are dragging us in to talk about connectivity and the future of networking and connectivity. We are attaching, believe it or not, all of our microcontrollers, our analog, our timing products to go with that. That is one of the reasons why we modified our Megatrend approach to really focus on networking and connectivity. There is a lot of investment at Microchip in that area now because it drags everything else along.
Got it. Now, since you told me about your feelings about 8-bit, I'll give you one other question before we get to 8-bit. I think if you look at the AI, ADAS, networking, these are all the new markets that you want to go after from a product diversification standpoint. Seems to be where you want to take Microchip. Can you talk to how you look at the M&A side? I mean, are you guys, do you think you have enough of the IP technology in-house given PMC-Sierra, Microsemi, all the acquisitions that you guys have made over the years? Or do you need more parts within the AI side, within the networking side to kind of be more current with where data center is going today, I guess? Do you see any more M&A tokens?
Yes. There's always insatiable need for more technology. We acquired a small token three weeks ago. We acquired TF Semiconductor, which is a small analog company that did high-voltage drivers up to 600 volts for motor control and power supply applications. Most of our high-voltage drivers for DC to DC converters and AC to DC and motor control only went up to about 100 volts. They filled in the rest of our product roadmap. That was just done about three or four weeks ago. We closed on that. We're continually looking for other technologies or token technologies that improve that total solution. When you looked at our, that's a good example on why would we buy this company, TF Semiconductor.
In high-voltage motor control applications, we had the digital controller, we had the timing devices, we had the temperature sensing devices, but we did not have the drivers. What that did was complete that total system solution. When we walk into those customers, we are now selling five or six devices for that motor control application.
Got it.
It really filled that out.
I think you had a couple of other ones too, Neuronics and Techron, I believe.
Neuronics and Techron, those were in the AI area. That allowed us to do modeling and simulation, and again, focused on low power. It allows us to do some of that modeling and simulation on the edge from a very low power standpoint.
Got it.
That is one of the key themes that we try to have in almost any technology or product we do, to try to lead from a power standpoint.
Got it. If you were to size those key markets, AI, networking, et cetera, Ethernet connectivity, what percent of the mix would it be? How do you see that growing out, let's say?
We have not provided which is more higher growth or faster growth. I think when we look at AI on the edge, we are seeing a great deal of growth on the product side. One of the things that we decided to do at Microchip was not get so caught up in providing an AI device, but providing AI services to make design in simpler. In February, we were the first company. Microchip has been investing in AI for engineering and business and internalizing to the company since 2017, 2018.
In February, the same large language model that we've been providing data to and helping ground it and creating a RAG model on it to allow our engineers to develop code for our microcontrollers and get 20%-40% productivity, we essentially unleashed and let that model be used by all of our customers for free to design in our microcontrollers. What we're also providing to our customers are services, right, using AIML to get them to design in our products easier and faster. It's not just providing AI at the edge, but providing productivity using those tools. We've been doing quite a bit there since, in fact, that tool has consistently gotten better to now where we've essentially vectorized a lot of our YouTube videos and data sheets.
When they're in our development environment, they can say, "Hey, I don't understand how this A to D function works." They hit a button, it'll bring up the YouTube video index to 12:42 to the exact spot that someone's teaching that. They can just continue learning and programming with our parts.
Okay. Great. We did not know you guys use YouTube video to study too. We will come to the most interesting part of the discussion, Microcontrollers. Maybe we can talk to how is 8-bit, how size that business. Obviously, you guys have a roadmap to 64-bit and maybe 128-bit down the road. Maybe we can walk slowly into the RISC side of the business. Obviously, RISC is more considered a sub-ecosystem than commercially outside. Maybe we can talk to 8-bit, what is happening there, what is the competition like, and maybe kind of walk through to the 64-128.
Okay. So, in 8-bit, we are still by far the market leader in terms of overall market share. We're still introducing innovative products on that side. A typical 8-bit product is really more of a mixed signal analog SoC. When you look at that, whether it's an 8-bit core or a 32-bit core on many of our microcontrollers, that's only about five percent of the die area. Then you've got memory and analog. Most of our microcontrollers are made up of predominantly analog, whether it's D to A converters, A to D converters, drivers, charge pumps, DC to DC converters. When you think about it, we talk about it as a microcontroller device, but it really is a mixed signal analog device.
That's where you differentiate whether it might be a China competitor, might be doing just a Pure Microcontroller.
Just a Pure Microcontroller.
Versus you guys may be doing like a full board-level solution with a memory and analog power line.
Memory and analog. That is how we differentiate. When a customer gets into that device, they decide they do not want an 8-bit converter, or they want a D to A, or they want this other peripheral, or they want programmable logic. We now offer essentially FPGA equivalent logic so that people can program their own digital onto our microcontrollers to do what it wants, what you need to do in that particular application. That is something unique to Microchip as well.
You're also pushing the roadmap to 16, 32 and.
Yeah. On most of our 16-bit devices are DSP-related digital signal processors. Those are really finely focused on power conversion. When you want to do the very low power conversion, whether it's in data centers or motor control, that's where that technology is specifically focused. 32-bit is much more of a broad market approach in terms of every little device. We're using the same philosophy that we did on 8-bit with putting as much analog peripherals around that from a support standpoint. On 64-bit, the origin of our 64-bit devices came, I think most people would not even grasp this in the room. The next generation of space computer is coming from Microchip. It's not coming from Intel. It's not coming from AMD.
NASA and JPL contracted with Microchip to produce the next generation space computer, which, believe it or not, they're still using 286-like products. That is an octal 64-bit RISC-V processor. That is on a 12-nanometer radiation hardened FinFET technology that Microchip had co-developed with one of our foundry partners for FPGA. Utilizing a lot of our FPGA technology, we built that into that device in our sampling and working with a number of aerospace companies, in fact, every aerospace and defense company around the world on getting that device or part of that family to them. What we are doing now is we're taking offshoots of that device and technology and offering that for industrial control, heavy machinery, industrial control. That particular device has accelerators built in and high-performance math functions so that you can use it for many solutions from an edge AI standpoint.
Got it. I'll ask one more question. Maybe we'll open up to the audience after that. Obviously, when we talk microcontroller, a big concern is China. I mean, they're coming in, they're trying to take share. How do you look at that? I mean, has Microchip evolved a lot more, even if it's 8-bit MCU to kind of stay above the China supply chain, or do you still see yourself impacted by some of the aggressive price competition that you see there, the supply coming on from there? How do you look at that landscape?
We did something interesting probably about five or six years ago. There are three types of customers in China. You design with a Western company. They then have production in a free trade zone, and then that is shipped out of China with no tariff incumbents. The second tier of customers are Chinese companies that white label products for Western companies, also export market. Then there is the third set of customers in China, which is local companies, local content for local market. That tier of customers probably started six years ago and more than that, almost 10 years ago when China made its intention that for local products, they need to have local semiconductor content. What we have done over the years is we have migrated our focus to the other two types of customers in China. We focus mainly on export from China.
We do not focus on China for China. Now, do we have products that sell in China for China? Yes. Where there are products that they cannot make locally, then they will buy our products for that market. We do not focus on that.
Got it. That's a process that's mostly complete is what you're saying. Your exposure to the domestic market is fair.
Our exposure to the domestic market is very low.
Got it. Okay.
It's just not been a focus for us.
Fantastic.
Because you're not going to fight that nationalistic. We felt there was no way to fight that.
Sure. They get the first opportunity to bid on it.
They get the first option. Anything that you move to build there will essentially be built there by some other company.
Got it.
It is just not our focus.
I got a minute. We'll take some questions from the audience. Hey, go ahead.
Is that in-house SerDes that you're doing for PCIe? [guess]
Some of our Ethernet products use in-house SerDes. We have a very large in-house SerDes development group. Depending on the product, we'll use external SerDes from some of the major suppliers, or we'll do internal SerDes. We do both.
Second question. Defense, historically, you're known for high-end FPGAs. Do you think that attachment comes in as things kind of go towards the growth? You've seen a lot of sequential growth there. I mean, it seems pretty poised for acceleration in the end market. Do you think that could continue as you guys attach to smaller-end stuff? Is that a place that you're excited about? Just kind of what might be your thought?
No, we're still seeing a great deal of design activity in that market. We're also seeing a great deal of design activity in the new defense space as well. There is a great deal of focus from Microchip on taking some of the products that we've done that were radiation hardened and making them radiation tolerant or downgrading, putting them in plastic packages to make those products usable and approachable by these new lower-cost entrants.
Great. Any other questions? Fantastic. We've got 20 seconds to the chat. Maybe I'll ask one last question if there's none other from the audience here. The customer relationships, I know one of the nine-point plan from CEO Sanghi [audio distortion] was kind of strengthening, rebuilding some of those relationships. Any thoughts, updates on how you're seeing that progressing?
That has actually gone extremely well. I think that was really kind of overblown. When we did our survey of the 7,000, we went out there to get real data on this. We had about 24% where it improved, 12% where it degraded, and that center portion where it was about 64% where there was no change. We spent a lot of time on that 12%. Myself personally visited probably almost every one of those customers and really spent a lot of time trying to understand what we could have done differently and work with them. We've improved the relationships in almost 90% of those where we were on design hold. We had also won a lot of customers too. Really fascinating, Sajid and I had met with a new investor that bought several million shares of Microchip.
They said, "You know, we heard about that whole customer thing, but we've been researching your company for two years and have talked to a number of your customers and your distributors before we decided to make the plunge in your stock." We made the plunge because we thought the entire market was overblown because we did not find that. We found a lot of your customers love Microchip. I asked them for the report and the data, but they were not willing to share that with me. I would love to get that. I do not think I have seen investors spend two years investigating us. It was interesting.
That's great to hear. Thank you, Richard. I think we are at the end of our time, I guess. Any other questions here? Okay. I think that brings us to the end.
Thank you.
Thank you.