Microchip Technology Incorporated (MCHP) Earnings Call Transcript & Summary

With that, good morning, everybody. Welcome to the Goldman Sachs Communacopia Technology Conference. I'm Jim Schneider, the semiconductor analyst here at Goldman Sachs. My pleasure to welcome Microchip with us today. We're happy to have CEO, Rich Simoncic, and Head of Investor Relations, Sajid Daudi with us today. Welcome, guys. The quarterly we reported back in early August, I thought your commentary was fairly constructive. You noted increased bookings, rising backlog levels in some pockets, lead times expanding and reduced inventory on your balance sheet. And I think you also talked about minimal Poland activity. So I'll ask you a few questions about the core and the trends you're seeing right now to lead off. So your June quarter backlog I believe was the head of the prior quarter. July bookings sort of the highest at the time. How do you continue to see trends in your backlog and bookings since then? And how is linearity tracked since your last earnings call?

Okay. So just before I answer that, I may say some forward-looking statements. Please refer to our online documents for appropriate risk factor based on what I'm saying. So before I get going here, we definitely saw a much better bookings last quarter. July was probably the highest bookings that we've seen in probably 2 years. August came in, although not as bad as it could have been, it was -- it looked more reasonable than we've seen in past August where typically August or some of the worst bookings month. And so no commentary on September as of yet, but we have seen much better bookings in the July time frame. Much better.

Okay. Very good. And then in terms of the end markets, which ones for fiscal '26, do you see rebounding most quickly? Is that auto, industrial, defense, cloud, data center, et cetera. And which do you think could be more sluggish for some period of time?

So right now, probably our top markets are AI, data center infrastructure. We're seeing some of our PCI switches pickup or Gen 5 PCI switches pick up. We're working on releasing our Gen 6 PCI switch and re-timer as well. We're also seeing connectivity and networking, our growth in T1 connectivity has really garnered a great deal of design activity. So when it comes to factory 4.0 automation, automotive, robotics and humanoid all of those need much higher connectivity speeds and that's where Ethernet T1 comes into play. And interestingly enough, we used to always look at attach on a microcontroller as the main stay, but we're seeing significant TSS or attach associated with our connectivity and networking products and that will generate quite a bit of growth going forward.

The other area of strength that we're seeing in the Aerospace and Defense segment as well, which is about 18% for last fiscal year. And again, beyond just the spend that we're seeing from the U.S. side, even our NATO allies and NATO customer or NATO -- them spending their defense budgets is also helping, so that's another area. And then FPGAs play -- those products play into that segment pretty nicely and all carry a much kind of richer margin profile as well.

Yes. I think some many people do realize that Microchip supplies about 20% of semiconductors for all aerospace, defense and space programs, right? So for instance, Mars Rover has about 130 devices from Microchip. The James Webb telescope. The other reason it's able to beam those brilliant pictures back to earth, the way it does is because we develop the driver technology that goes in there. And so when it comes to anything that leaves our earth atmosphere or defense, that's where Microchip comes in. Our -- the latest generation of high-performance space computer is Octal risk 5 architecture. Most people would never think that the next-generation space computer would come from Microchip. That should come from A&D that come from Intel that should come from. But no, that's coming from Microchip, and that's being designed into multiple platforms by a number of defense and aerospace customers around the world.

I think one exception to that broad end market improvement, you think you noted on the call was automotive. That's been a controversial end market for a lot of companies. Just curious whether you're seeing any changes there or any kind of improvement or not?

So we don't play very much in China EV vehicles, where we're focused on is a lot of connectivity where in a lot of microcontrollers that are on the edge in those automotive automobiles. HMID. And then when you look in a car and you see those beautiful touch panels that are across the dashboard, 90% of those touch panels all are using Microchip's HMID products. So when you look in that Mercedes and you've got that full wrap around, that's Microchip technology that enables all of that. And so we're seeing great growth in those particular areas. We're seeing a lot of design activity in the next generation of can and control every network and LIN are just not capable of the speeds and the amount of sensor interfaces that you have in those cars, and that's where the next generation of Ethernet comes in. We're seeing a lot of design activity on those platforms, too.

Very good. Just 1 point on inventory. I think last quarter, you talked about your inventory on balance sheet going down by a little over $120 million sequentially. Maybe give us a sense like how much of that was sort of distributor destocking internal improvement? Like maybe give a sense of what you're seeing from an inventory perspective, both on your balance sheet and your customers.

So we are aggressively cutting overall inventory. So we're manufacturing at about 50% of demand to bring that inventory demand down to bring that overall inventory down significantly. And that's what caused us to go from that 261 peak number by the end of September in the low 190s range. And then we'll start to slowly ramp up the factories again in the December quarter to match where actual demand is. So we're bringing that inventory down very aggressively. Just no value to having that high inventory there. For the same reason, we thought to have no visibility, right? When you have huge inventories or huge stock positive inventory most people in their general mind think that, that's a good thing, right? But what it does is it forces essentially customers to give you no backlog and no visibility as to needs actually are. And that actually creates a bigger issue with semiconductors. So you have to have the right balance of inventory and 261 days it's just way too much.

Yes. Makes sense.

And the September quarter guidance close for 195 to 200 days.

Great. Maybe one more short-term question before I move to some longer-term questions. in terms of your growth outlook and what you're seeing from China specifically, you just mentioned not so much on the EV side, but how do you think about the potential impact to the business based on sort of the macro challenges that sort of tariff regimes and everything else that we're seeing right now. And maybe talk about how your overall China business is tracking?

So it's interesting, about 18% of our revenue is China. About 9% of half of that are shipped into trade zones and then ship right back out. So there's really no tariff influence. And then we have about 4% or 5% that is more proprietary in nature, AB converters, higher-end micro connectivity devices, and that's not subject to getting a lot of pressure from -- within China semiconductor manufacturers. That leaves about 4% to 5% of product that's at risk that's from build outside that's brought into China that's been consumed. And we're working on different models to protect that overall revenue. But China is probably the first country that corrected the inventory early on. And so China continues to improve in some ways. Again, we don't play in China EV, but there are many other markets that are still growing within China.

Yes. Very good. Pivoting to some more strategic issues. Since his return to the company's CEO, Steve Sanghi, has kind of had a 9-point strategic plan for the company. You made a lot of progress on those fronts. What do you see as the most important remaining steps for the overall company still to be completed? And then maybe, Rich, where are you focused on most in terms of operational objectives?

We're still working to get the inventory back down to that 130 to 150 days we have. What I'm mostly focused on is we're developing our product strategy within microcontrollers we've reorganized significantly the microcontroller groups in terms of bringing together in one organization. making sure that the connectivity and networking and data center businesses are properly funded to grow. Also have created a new AIML BU or business unit within the company just to really drive growth. We've got a number of business units that all need some type of accelerator or network process are built into those products. And rather than having each group independently do their own. We're working on a unified network process or an accelerator that can be used by 5 to 8 BUs within the company. And so quite a bit of work is going on in that area within Microchip.

Okay. Great. And then maybe within the strategic review, the financial targets you sort of laid out, you talked about operating expense goal of 25% of revenue. Right now, you're just above 30%. Where is the path to 25% look like for Microchip, how fast can you get there? And sort of keeping in mind the fact that the Microchip historically has this kind of philosophy or culture of shared sacrifice and downturns and kind of how you expect that to reverse or not in the upturn?

So we've already announced that we've gone too long without bonuses and we went very long on shared sacrifice. This reminds we have the 2001 inventory correction. Many of the people in this room were not even here then. But when you look at the last time the semiconductor industry got far ahead of its skis, the way it did here was 2001. And it took almost 3 years for the industry to find its bearings again. I know a lot of people are say, well, that's a long time. But when an industry overinvests and gets too far ahead, it takes that much time to get back to where it needs to be some degree. And as we're getting to 25%, the difference between back then and today is there's probably a lot more productivity gains from AI implementation and operations Microchip, we started our AI business operations implementation back in 2017, whether it's AI, ML, deep learning. So probably 87% of all orders placed on Microchip are actually auto scheduled using deep learning product recommendations for customers, the reason why we get so much TSS is that's all AI generated, and that's been being done for probably about 5 years. We're now working and moving into a lot of productivity improvements in software generation and design generation. So that's -- we've been doing optimizing overall expenses for quite some time in investing in this area. So we feel pretty confident about getting back to that 25% Model. And I would just maybe add to that as thinking about modeling. So use the March number, is that really factored in all the layoffs and everything. So that's your baseline number. And the variable piece of that OpEx line is the variable comp which we have a lot of control. So that will vary and be pretty disciplined.

Okay. Very good. Maybe switching to some of the end market trends that you're seeing right now. You touched on it before, Rich, about data center. That exposure has grown over time. It seems to be a good business for you. We've also heard from others that the crowd of the space. So maybe talk about Microchip and what applications do you see? And do you address most commonly. What's your competitive advantage there?

So we address Ethernet, USB hubs, our timing products and our PCIe switches are the predominant products that we're pushing. We try not to chase even though we are an analog power supplier, we try not to chase the power side of that business because to be a jump ball from each generation to the next in terms of winners and losers. And so we focus on the connectivity portion of that. Connectivity timing and our microcontrollers are used for essentially power management within those. And then security, we are the security route of trust for all of the GPUs that are produced. And so our security devices are sold into all of those manufacturers today.

Interesting. And the emergence of Generative AI, has that really changed your focus in the market at all? Or is it just made the market bigger with the same products you had before?

I think it's just made the market bigger. It's interesting with AI has been around a long time and companies that didn't get on the AI bandwagon, back on in 2017, 2018, really are way behind because AI technology has been around for -- you should have been using it to do, to help improve productivity for a long time. GenAI just make it more familiar to everyone. But we brought GenAI on as soon as we could behind our firewall, and it created the brag models for multiple functions within the company. And it's really driving a lot of productivity. So I think as more companies figure out the benefits to AI and that benefits the GenAI that, that market is just at its infancy. I think there's going to be tremendous growth as people start to really figure that out. I talked to a lot of friends of mine that are at different companies and a part of the venture networks and so many companies are just so still behind the learning curve on Gen AI and what it's good for and overall AI in particular. I think you'll probably find some of the people in this room still are not really using it effectively to do what they need to do on a daily basis.

Guilty as charged. But maybe talk a little bit about the Aerospace and Defense. You called that one out tendentially before. It's been a good market for you. Maybe talk about what you think about that market and your exposure to it is underappreciated by investors. Why is the intricate opportunity for Microchip specifically? Your competition?

So you know what, we -- Microchip specializes in long-lived devices, right? That's where we focus. For instance, you mentioned before, power, why are we going into doing something within power in AI and data center because that -- I don't consider that long lived. That becomes jump all of these generations. So we are focused on the security aspect, the timing aspect, which is has much longer life. So within A&D, we have a lot of technology or knowledge in the area of radiation hardening and packaging that and security that defense and aerospace customers really desire. And we've got a 60-year heritage that we inherited from some of the acquisitions and relationships that are in thousands of programs that have been there for quite some time. So we've got people that have, I don't know, a strong association with these customers and an extremely strong association with this marketplace in terms of what it needs. And so they are very devoted when is just at the Paris as show meeting with all of our clients. And it was fascinating to see every product that we worked up to had Microchip in it to some degree everything. And then we're looking forward to our next generations of products and what HPSC could do for them. And most people don't even probably realize this Microchip is but doing silicon carbide for probably 25 years in aerospace and space applications. So we didn't use silicon carbide to go after cars and inverters. We use it to go after power cores and power supplies that are sitting in space for 25 years. And that's been our focus, those long-lived applications, and we have those relationships with those customers. They look to us for that knowledge and dependability.

And within that or maybe on top of that, FPGA, that's been a pretty strong brewer for you. Maybe talk a little bit about sort of where your where you're playing where some of your competitors may not be playing and why that's kind of a drill opportunity for Microchip?

So the original FPGAs were designed for extremely low power applications. you have high compute with very low power. Typically, Microchip FDA for a 50% to 60% less power than all other FPGAs made. And I was specifically designed that way or architected that way for space or defense application radios. We've -- since we've acquired Microsemi, we've now positioned that entire product line in the industrial control into medical customers. So one particular application is think of a probe in a medical application where you're doing an ultrasound on someone. That probably has an FPGA and they are probably 164 lanes of high voltage that Pro gets very hot with a technician using that particular ultrasound probe. If you replace that pro with a Microchip FPGA, you cut the power and have -- and now that operator or a technician can use that probe all day long without fatigue or issues. And so what we've done now is we've found markets that we serve Microchip serves and have slowly been working that technology into all of our industrial customers, our medical customers and finding significant growth areas across the board. And so if we look at the FPGA business, it's probably more than doubled since we bought Microsemi, and we continue to see more traction. So now what we're doing is where we're moving up in the product range with our PolarFire 2 products, and we're also moving down in the product range, offering some lower cost alternatives that we just announced and then offering some alternatives for -- even on a very low end, where customers can do more ASIC type of device.

Great. And then I mean you touched on industrial and obviously, it's a very broad market across a bunch of submarkets. Maybe talk about for Microchip's business where you see the most traction and the most growth right now over the next maybe couple of quarters or so.

I'd probably see probably the most design traction that we're seeing is in our connectivity and networking because all markets because of the nature of whether it's humanoid, whether it's Industry 4.0, whether it's automotive, all of the interface and connectivity standards that have been out there for so long, can land RS-232, 45 or all obsolete now. You cannot you cannot move the amount of data around on an application with those old technology standards. And so probably the -- when we go to -- when I'm going to visit customers next week, I'll be through all Europe. What customers want to talk to us about the most is what are we doing in AI? And the second one is what are we doing in connectivity. That is customers' biggest issues that they're seeing right now.

Pivot to operations for a second. Near your heart, I'm sure. Just maybe talk about the TSS, total system solution approach that you've been taking for several years now. How has the strategy sort of unfolded for you? How does it differentiate you from your competitors? And maybe talk about if you can quantify market share gains or any kind of other quantitative metrics you can put to that in terms of the success of DSS?

So what we've done over time is within the product definition itself when we call them anchors. When we come out with an anchor product, the first thing that we do, we have groups of people that look at, okay, what are the surrounding devices that can add to that overall solution. So for instance, for FPGA, we just came out with a very sophisticated product family of power management ICs or PMICs that are specifically tailored for that product. So rather than having someone else create a PMIC or a power management IC this device is made for that product line to optimize its overall performance. And so we have design where we design specific products to optimize the performance. And that what we also have is we have a recommender tool that we -- an AI tool that we put online probably about 6 years ago that essentially when a salesperson enters account, they pick up an anchor device, let's say, it's a smoke detector or a door lock -- it immediately goes through our system and then with a 90% correlation factor presents to that salesperson, all of the devices that should go with it. And then all of the associated documentation and then the highest probability of win ratio in terms of tracking it. So when you have tens of thousands of products, you can't have each salesperson is not going to know what to sell in a particular account. And so what makes it all work is that recommend their tool that recommends the appropriate device to sell. So when we came out with, for instance, our T1 product line or T1s, we program or help that recommended tool or grounded in information that will recommend the associated devices and -- and so in our design in activity on our T1 products, we are seeing more revenue in our attached products than we actually are seeing in our T1 products, leave it on that. And so that's how that whole system works for us.

Interesting. And then relative to AI, you've talked about an AI coding assistant, which I don't think most people would think about in the microcontroller context. But how long have you had this product internally when you've be really good to customers? And what has the feedback been?

So it's kind of an interesting story. It started probably about 2.5 years ago. I hate writing software personally, I'm a hardware person. And so we had our development tool group, and I went to the -- when these GenAI tools came out, I wanted to have him start working on a cogenerator to assist our customers. And so I went and I wrote some programs. And I hadn't checked them out and they were all perfect. And I could -- before that, I couldn't get them to do it. But since I was this hardware guy and I wind up writing a bunch of software. He -- the Vice President of the group bought into it, and then he started using it. And at that time, we're using the public GenAI software and it hallucinates, it was okay, but it wasn't great. And so we did is we brought it behind our firewall, and we started creating RAG models. And grounding it to eliminate hallucination. And so we grounded it, eliminated hallucination and then we started using it internally for 1,000 or so software developers. And so we -- once we did that, the amount of grounding that took place, the hallucinations essentially were nonexistent. And then we then vectorized all of our data sheets and knowledge bases so that it was easy for the RAG model to consume that information and data. And then what we found was our internal software engineers, we're seeing about a 40 -- 20% to 40% productivity improvement. So then in February, we decided to release it to our customers. Our customers started seeing the benefit. And what we did is we vectorized all of our YouTube videos, all of our training videos. And so now a customer is writing code can bring up the information, it will actually auto fill out the code, see what they're writing or to fill it out. If they are not familiar with how a microcontroller or something works, the code will actually recommend the YouTube video and will actually index to the exact 2-minute, 16 seconds point that they mentioned, what they want to know about in that video. So all in one environment, they get the data sheets, help video context, auto generation of code quality checks, testing all in one application. And now we've the most recent release that we did this week or this month is going on is we've added AI agents to it to actually take a lot of the functions to work off of the programmer itself. And so we were the first to come out with it. And I think it has to do because we've been working with AI since 2017, 2018, we're not unfamiliar. We were very early with GenAI, putting it behind our firewall, grounding it, creating sandboxes and then creating rag models for people to use. So we're pretty sophisticated in this area and our customers are really starting to see the benefit of -- and I know internally, we have absolutely seen the benefit of this.

Interesting. Interesting. Then just maybe switching topics a little bit to pricing. I mean, I realize you're not in the commodity business, you're focused on high-margin long-lived products. But we have some interesting gyrations during COVID, big changes in pricing and competitors have added a lot of capacity operationally over the last several years. So maybe talk about how you think about the analog side of the portfolio in terms of forward pricing trends you're expecting over the next couple of years more longer term?

So obviously, we reduced pricing on our products. But pricing doesn't change that much that quickly. So pricing is typically on the next design that takes place. But you've got designs that were put in production 20 years ago, 10 years ago, 15 years ago. the ability for those customers to move or redesign those products is somewhat limited. And so really, the battle is always on the newer designs. And so what happens is as you get the newer designs, they're much more competitive, but you've got this whole long tail of older designs that are at much higher prices. And so it becomes a blend over time. So -- and that's where we put that forecast of single-digit ASP reduction over time. It gets back to a more normal design approach as we go forward.

So that's more of like a go-forward perspective on new designs, but -- but it's a mix effect in terms of the fact.

There are some price reductions on old designs that's not that much because most customers, let's say someone comes in with this much cheaper device they go and they said, okay, I'm going to redesign this and they find the engineers that worked on it 10 years ago, just don't exist anymore. The design files don't exist anymore. It may have been designed on an older EDA platform. And to resurrect that they have to actually convert that to the newer EDA platform for them to actually redesign that particular product. You could never really -- one of the things I've learned over time because it's very difficult to buy already existing sockets in the design. Just in fact, whenever I develop new product lines at Microchip over the years, and they've developed a lot of new product lines, our target is not to go after what's already designed. In fact, until salespeople don't waste our time with already designed in. Let's just focus on new design wins, new activities and here is why. That's where the money is, rather than wasting time trying to do go back and try to convince a customer to redesign something. It just doesn't work.

Yes. Makes sense. Maybe probably time for one more question, but I wanted to sort of ask you with respect to manufacturing footprint and your strategy, I mean you have production capacity in the United States, you also have outsourced capacity, which is relies mainly in Asia. I think some of your competitors have been very vocal about doing a lot more onshore U.S. capacity. And now we have a tariff regime in place where that incentivizes U.S. production. So maybe talk about your manufacturing strategy long term vis-a-vis your competitors and then your understanding about Microchip's applicability to the new tariff regime that was recently announced?

So about 18% of our revenue, probably look at it this way, about 18% of our revenue is A&D. That's almost all done on onshore, right? Assembly, testing, manufacturing, we've got a number of fabs that do that. And then we've got quite a bit of analog and microcontrollers that have done, I'm sure. And then our foundry partners have built out, whether it's TSMC or GlobalFoundries are in the middle of building or have built fabs onshore. And so we fit that because we do a lot of our production already onshore at Microchip. I think we've -- we've said before that about 45% of our revenue in terms of wafers is 40% to 45% is not sure. And that number is actually a little better when we start to look at assembly and test and we take Wafer Foundries that build onshore.

Very good. I think with that, we're pretty much out of time. But thank you very much, Rich, and Sajid being here. Appreciate it.

Thank you.