analyst for Mizuho. Joining me is COO Richard Simoncic from Microchip, and Sajid Daudi, heads up Investor Relations for Microchip. Thank you for joining us at the Mizuho Technology Conference, gentlemen. Really appreciate it. Just to give a quick background, Microchip obviously makes the tech that works in the background. It's intelligent connectivity, timing management, control, synchronizing, basically connecting and managing an orchestra of industrial comms, space, networks, everything together. It's my pleasure to welcome back Microchip COO, Richard Simoncic. Richard, we had you last year.
Yes, you did.
Let's make this an annual event, Sajid, Microchip's obviously seeing a very strong rebound. Margins are recovering nicely. Revenues are still on a rebound. With that, again, my pleasure welcoming you guys to the Technology Conference here.
Thank you.
Maybe I'll start off with a couple of quick big picture questions on what you guys are seeing. If you can go through some of your end markets. Obviously, 2023, 2024 was a big year. Then 2024 you had some of the inventory normalizing issues, 2024, 2025, we started to see that inventory normalization happen. 2026, we are starting to see some sort of recovery, but maybe you can level set what's happening in the different markets from industrial, aerospace, defense, comms-
Okay
consumer, et cetera, and we can go from there.
All right. Before I begin, I wish to remind you that during this presentation we may be making some projections and other forward-looking statements. These are predictions, and the actual results may vary materially. I refer you to our filings with the SEC regarding some important risk factors about the company.
Okay, Richard, I thought I was the only one.
That's a life-compelled read
I thought I was the only one supposed to read from them.
Yeah. That was, so I always read that just to make sure I get it right.
Yeah.
Not drop a word. Markets that are doing well for us right now are within, obviously, data center. Within data center, we have Ethernet connectivity, USB connectivity. We have atomic clocks that are growing in growth there. Our power controllers are growing within data centers, not to mention more specifically our CXL PCIe switches and retimers are starting to grow within that marketplace. Within the A&D complex, Microchip is the largest supplier of radiation and Rad-tolerant types of devices out there to that complex.
With the missile replenishment and defense replenishment, we've been asked to scale up almost 4x what we're doing in terms of products, in terms to replenish some of the arsenals. We're very involved in trying to ramp up manufacturing on all of our A&D products. Not to mention aerospace. In a typical Boeing or Airbus, we have over $100,000 of content, if not more, that's also ramping up pretty considerably for us.
Got it. Let's start off with the most interesting segment, I guess, which is aerospace defense. I think as you look at the space market, you have a big content in some of the low-energy satellites, radiation-hardened devices. You obviously are one of the biggest suppliers in space, I guess. Maybe you can talk to what you're seeing there. How's the backlog? How's the pipeline? How's that growth? There's a lot of focus on that space given a massive pipeline of satellite launches that are probably, as people look out, so maybe you can talk to what.
Yeah. As you start to build out some of the LEO satellites and longer-lasting satellites, those are all going to be somewhat radiation-hardened product. If you're talking really the low Earth, cheaper satellites that you may have an expected lifetime of two to three years, if that, then those are really COTS or off-the-shelf type products. If you're looking for a high reliability that will last a little bit longer, then you're going to be looking at radiation-tolerant or radiation-hardened products. Data centers, once they start doing those, will be radiation-hardened.
Yeah.
Sajid just let me know that Artemis III was just announced today. That's literally thousands of our products are on Artemis on those launches.
Any way to size the content? Like, is it.
We don't.
10,000 or 20,000?
Yeah, we don't typically break out that content.
Got it.
to that degree.
Yeah, typically these are radar devices which carry a high margin and fairly sticky, I guess.
Yeah.
Yeah.
A typical satellite may have 30+ products from us, right?
Yeah.
Margin, like when an Artemis rocket has thousands of our products on it, literally.
Got it. In the past, I think it was the last year, you guys sized the aerospace defense industry as a $750 billion industry by 2035. Where do you think it is today? What is the growth in that market? Do you expect to grow in line or faster than that market, I guess, if you were to?
Faster right now because of the replenishment of stockpiles.
Yeah.
Right. That is, whether it's in missile defense or missiles themselves, or drone defense systems-
Yeah
Microchip has a lot of content in those particular product areas.
Yeah.
Just because of replenishment alone over the next 3-5 years.
Yeah
We're going to see pretty good growth.
Yeah
For at least the next five years.
Yeah. I think among the semiconductor suppliers, you are globally the biggest defense aerospace supplier. I mean, 16%, 18% of revenues from aerospace defense. I don't think anybody has that high of a proportion. When do you start to see that replenishment cycle start to pick up? It's been fairly flat. Granted, it's lumpy. When do you actually see?
We're actually starting to see those orders roll in now.
Got it.
Right. When you It actually is a fairly long lead time for some of these rad-tolerant or rad-hard products. The typical lead time to make, or cycle time to make them are anywhere from 12-24 months. Right. That will start to layer in over next fiscal year.
Got it. That's really, you are seeing the orders now, but the revenue ramp might be more.
Over, yeah. You don't start to ship those for, like I said, 12-24 months.
Got it. That's got to be pretty margin accretive too, right? Given, you're already seeing a good margin rebound, and we'll talk about that.
Right
This should layer on top of that margin.
Right. Those are.
The product-
Because of the amount of work that goes into them,
Right
The amount of effort, and it's actually quite a bit of testing and scrap associated with it, too,
Yeah
that they typically command much higher margins.
Right. I used to work at an analog company, and we used to do rad-hard, and things didn't change for, like, 30 years, so we used to make like 500% margins on some of this stuff.
Yeah.
Used to make it for satellites.
The amount of the difficulty involved in it is quite high.
Yes. Right
It's not that simple.
Yeah, you don't want satellites falling out of the sky because.
No.
Okay, great. I want to move to the data center side. Obviously, AI data center has been a big focus among analog guys. You guys, it's about $300 million last year. This year, you had talked about it growing to $500 million, almost 65%-70% year-on-year growth. Maybe you can talk to what are the key buckets you supply on the AI data center side.
No, there's actually, I think when we first announced that, there's $303 million of just dedicated PCIe switches, retimers.
Yeah
CXL devices. There's another $484 million of the timers and DSPs and controllers for power supplies.
For the rack.
For the rack, Ethernet, and USB connectivity in the rack.
Got it.
The total last year was almost $800 million.
Got it. Okay.
Right?
Yeah.
What we tried to take out of that was just the PCIe switches and retimers and those CXL devices, that was $303 million last year, and it's expected to exceed $500 million this year.
Got it
coming up.
How about the total pie, the $800 million, what does that do?
That also continues to grow, but not as fast a rate as the.
Got it.
the PCI switches and CXL devices.
Got it.
Vijay, we break down the data center and compute at 18%, we kind of provided a pure play split, there's a compute portion still in that 16%.
Yeah
There's the ancillary products that kind of come with the data center.
Got it.
solutions. The pure play is the PCIe connectivity switches, retimers.
Yeah.
There's the memory controllers, and then the storage part.
Right.
Those are the three kind of pure play.
$800 million last year in total. This year, PCIe is obviously growing a lot faster, but expect this to be 20% of revenue, something like that? You guys are doing, like, five?
We're not projecting that.
Got it. Okay.
It's going to continue to grow.
Got it. On the retimer PCIe side, obviously, you have competition there. You are bringing out PCIe Gen 6. I think you guys have talked about six to eight CSP and ODM wins, right? Maybe you can talk about the competitive landscape, the Broadcom, Astera Labs. How are you competing with them? Are you getting traction on the rack, in scale-up, in copper? Or is it somewhere else where you see-
We're up to about eight major wins on PCIe Gen 6 switch. We're at our first win on our retimer. We've just launched these products, so we still have lots of design activity that's taking place.
Yeah.
We're also seeing some newer wins on some of our older generation products, too. Our Gen 4 products within robotics, Industry 4.0 in automotive, and we're also seeing our Gen 5 products. There's still quite a bit of business in that area, too-
Yeah
where those are being picked up in some new designs, too.
Got it. In terms of data center, where you are seeing the PCIe retimer wins, is the architecture actually specified by the CSP and the ODM just runs with it, or are both wins separate? Like, does the CSP specify the PCIe Gen 6 separately for their scale up and the ODM specifies the?
I've seen a little bit of both.
Got it.
There's no particular side to that.
Got it. Okay. How do you see that, Think you start to gain share versus some of the incumbent PCIe retimer suppliers.
Microchip, we were the predominant supplier on Gen 3, Gen 4, Gen 5.
Yeah.
We missed the window. We rectified that with what we did with coming out with the three nanometer device on Gen 6.
Yeah.
Which actually gives us the ability to stay in the forefront with Gen 6, and then Gen 7 will follow very quickly.
Right
We're already on three nanometer node. It allows us to catch up quickly and then move from there. The tool suite, the Switchtec tool suite, is still considered one of the best in the industry.
Yeah.
We continue to make that Switchtec tool suite even better for our memory CXL devices and for our PCIe.
Yeah
devices. I think customers are happy to see us back there from a support perspective.
Yeah.
Knowing that we've already got PCIe Gen 7 well down that path.
Got it
to getting released.
Just to take a step back, when you look at the space, A&D, aerospace defense, and the data center PCIe retimer, do these run through your own fab? Because you have a substantial foundry component as well. Do these run through the fab? The reason I ask is these are obviously going a lot faster. They're margin accretive, but they also drive better utilization for you, and that should also help on the margin line. Just wondering how?
The aerospace and defense, majority of discretes power, those devices all run through our own fabs.
Got it.
We run those out of multiple wafer fabs.
Yeah
that we have. Then our FPGA rad-hard, rad-tolerant products are done at a foundry partner on U.S. soil.
Yeah.
That's on a 12 nm FinFET technology.
For the rad-hard devices?
For the rad-hard devices.
Right, okay.
We're also doing our high-performance space compute, the Octal RISC-V 64-bit device for JPL and NASA. That is also on that 12 nm FinFET technology on U.S. soil.
Got it.
All right.
If I look at the gross margin trajectory, you guys are already back to like 62%. You talk about 300 basis points of underutilization. Once you get to that, you get to the 65%, which is close to your long-term target, or what you are close to that in the last three years back, I guess. That is just part of the story, right? That's just utilization. When you layer in the higher margin contribution from some of these faster-growing markets, wouldn't that argue for a higher margin, as you start to fill the fabs with these products?
Yes, we will have a richer mix, but we're not trying to really target the overall company to that higher margin device. We don't want to fall into that margin trap of leaving growth on the table.
Yeah.
I think that 65% really is used to send a message to a lot of our internal BUs to not get stuck, that it has to be
Right
at this 70 or 75 or 80, so that we continue to grow.
Yeah. In Aerospace Defense, you guys lead by a mile, so you don't have to be price competitive.
You still have to be reasonable, right?
Right. Yeah.
Not everything runs at, I think everybody thinks just right away that it's outrageous margins.
Yeah
Some of it is not, right?
Right.
It has to be competitive or reasonable, and because of the amount of testing and hardening and just the amount that you throw away.
Yeah
To get to those few that work. Some of them are great, some of them not so great.
Yeah.
I would just add to Richard's point is that think of it as a two-cycle margin, right? Again, we had a very broad band last cycle.
Yeah.
Obviously, we didn't manage that effectively, and that's really where the thinking is, that there'll be moments in time where it might be higher, and then lower. Just narrowing that bandwidth, and then over time, I think you can potentially address it. At this point.
Got it
insight as to the target levels.
Yeah. I want to hit the faster-growing market. We hit AI data center, I guess we hit the aerospace defense, right? Robotics, I think this is another market where there seems to be a lot more investor attention as a longer-term roadmap, either from factory automation, Industry 4.0, or humanoid robots and quadrupeds and all that other stuff. Maybe you can talk to what you are seeing on the robotics side, what your content is. What's your exposure to the big humanoid robot?
Our exposure to the big humanoid robots, and you've probably seen it in some of our reference design partners, whether it's Qualcomm or whether it's NVIDIA or AMD. Within the robotics area, you'll find some of our FPGAs for our sensor fusion. You'll also find our PCIe Gen 4 switches as well. We're the only ones that are producing PCIe Gen 4 switches for industrial applications and automotive applications.
Yeah.
When you're looking at most of the robots that are being built out there, they need PCIe switching in order to deal with the amounts of data that's traversing through that robot. It's the same thing with our Ethernet products as well. Probably the two biggest product lines that we see growth in in robotics are PCIe switches Gen 4.
Yeah
Ethernet. What we're also seeing too is, for the longest time, even within our own factories.
Yeah
We're using more and more robots because the AI-enabled software finally has met up with the hardware to give us a capability that actually gives us a very good ROI. Yeah. We're using material robots in our factories. We're looking at robotic arms at certain stations to take it from robots and then load it in equipment. Yeah. These are older factories that need to be automated. Right. It never really made financial sense. For the first time in a long time, we're doing the analysis, it actually makes sense. Right. I imagine if it's making sense for our factories to do more automation on older types of equipment, it's probably going to be start to make sense for many other companies as well.
Yeah. When you look at the PCIe Gen4 switches and sensors and obviously there's probably a lot of microcontrollers in there too for all the actuators and stuff. What is the dollar content you're looking at on the humanoid robot side? On the factory automation side, what's the opportunity there? How do you size that whole market?
We haven't published a number on that. Yeah. In terms of numbers for robots, because most of the robots that were designed into there, they're claiming that there won't be real production volumes of those until 2029 or 2030.
Got it.
Right.
You haven't spoken to Elon Musk yet.
Yeah. Be that.
I'm just kidding. I'm just joking.
I think they're still out at that time too.
I sat with a team that just trying to get a same answer, right?
Right.
About the content opportunity here and stuff.
Right.
There's just a host of numbers that are out there.
Yes.
Really, I think what we identified, ended up identifying is six critical layers where there'll be just a host of our products, right?
Right.
You mentioned actuation, sensing is another area, connectivity Rich mentioned.
Yeah.
On their compute, which is part of that FPGA style.
Yeah.
Functional safety is another area where we have a lot of post-quantum crypto type of products.
Got it. Yeah.
Those six or seven layers of robotic play.
Yeah,
is really where we think our opportunity lies, and longer term we'll see where it all squares up.
Got it. Obviously, you are working with some of the leading U.S. robotics guys too, human robotics guys too. What's your exposure between China is making a lot of noise with Unitree and a whole bunch of other guys there, running up the robotic side. Do you guys actually?
Our products that are mainly within those robots within that marketplace are our PCIe switches and our Ethernet devices.
Got it. Okay.
Right. Not so much on our microcontrollers or power, that's very much local content on that.
Yeah,
PCIe switches and Ethernet products.
Got it. If you were to look at that bucket of your faster-growing, those three segments, would that come? Obviously, you already sized the data center side about $800 million last year. Put some nominal growth, you can see where that should be. You obviously have the Aerospace Defense side, which is pretty big chunk, 80% that can ramp on that 18-month lead time, and you have robotics. All of these are accretive margins. How do you size those faster growth market as a bucket within your outlook as you look at 2027, 2028?
The problem, especially with robotics.
Yeah,
we can put numbers around AMD.
Right.
We can put numbers around. With robotics, we have a team of people that's working on it.
Yeah.
We collected all of the robots that we're designed into and what devices, and put together a plan in terms of when they may go to production.
Yeah.
It's so new is when does that actually happen?
Got it.
We've not committed anything to the street or investors yet, because we have to start to see some of those roll out. Even some companies that have some of the larger installations of our chips in there, their comments are that they won't start to do production roll-outs until the 2029 timeframe. Even then, you're having to sign up or pre-invest.
Yeah,
in there to be on the list to get an initial robot.
Yeah,
to actually try that within your factories.
Got it.
I think we're still a little bit.
Yeah,
premature in terms of trying to size it.
Yeah. I want to quickly hit the FPGA side. You have the PolarFire FPGA. It's been having some good traction in Aerospace Defense. As you mentioned, robotics and a lot of other markets. You have a peer out there, which is private. As you had mentioned before, they have gone through a significant atrophy because people were not invested, Altera. Is there any part of that that makes sense, or is that a market that's just coming your way?
I think that's a market that's just coming our way. We're finding a lot of customers that they're choosing AMD.
Yeah,
and they're choosing Microchip.
Right. Even on the quantum side, you need FPGA switches, right?
Right. Yeah. We're finding a lot of customers when they want really high security, PQC, and that's when we come into play. When they're looking at needing extremely low power nodes.
Yeah
or medical or. Right, that's when we come into play.
Got it.
We're finding a lot of customers are choosing AMD on the super high end, and then, you know, Microchip for that next mid-tier, low power type applications.
Got it. I want to quickly hit some of the macro trends that you're seeing. Obviously, a lot of your peers have been out there, they have raised pricing a couple of times already, right? There was a price increase in March, there was one in April. There's one supposedly coming up mid-year here. Your input costs have gone up. You have been fairly disciplined. You kind of mended your fences. You had burned some bridges, I guess on the last, you kind of come back, have kind of established some good relationships again. You have to transfer some of those, pass on some of those costs, right? The input costs.
Right
How are you seeing, are you assuming a price increase here? How should we look at that?
Yeah. It's really interesting. We spent a lot of time installing first principles in terms of how to work with customers and make sure we mend all these fences. Then we threw out the old playbook that we had on raising prices from COVID.
Yeah.
It seems that ironically, all of our competitors picked up a playbook that we threw away and have reinstituted it.
Yeah.
I find fascinating. Knowing that we failed with that playbook, we threw it out, and we're doing it differently here. We're looking at where are the price increases, where are the cost increases.
Yeah
We're working with customers to adjust each of the products individually. There's a more truer measure in terms of what that cost impact is to each customer. Customers have responded favorably to that. Okay, you've held off for a long time. Most of our competitors are on a third round of price increases, and we're just entering our first.
Got it, okay.
Right? I think a lot of customers have responded extremely favorable to that, and we're seeing a great deal of design win activity.
Yeah
Hoping to regain any possible share that we may have lost on some of the applications that we had before. I just came back from South Korea and Singapore, going through this with a number of customers and they seem to really like the approach that we're doing. That no one likes price increases.
Right
They like that we're doing it with.
Transparency.
Yeah, transparency and candor.
Got it.
not just slapping 10% on them and saying, "Thank you very much.
Got it. Just for clarity, we did put out a press release regarding this topic earlier, last week, and really our guidance for June is unaffected by it. This would be a September quarter type of event.
Got it.
really October quarter, where it will be fully reflected in that.
Got it. I want to ask one quick question and then maybe open it up for the audience. When you look at inventories, maybe you can help characterize what you're seeing in inventories at the customer level on industrial, automotive it's not for you, but how you're seeing inventory levels there. What are you seeing on the distie side? Where are inventory levels now versus last year? How are lead times looking this time versus last year? Just three.
Do you want.
Quick.
I can start it now. Okay, during COVID, a lot of our industrial customers, and those fit in that small to medium-sized customers.
Yeah.
When they were able to get product, they wound up buying as much as two to three years' worth of product. Right? When you think about it, someone that may buy 100,000 chips.
Yeah
a particular injection mold equipment, they bought 300,000, right? They were behind all of the large companies, to some degree. When they finally got their chance, they said, "You know what? We can't chance it. Let's just buy a bunch." What we found in talking to one of our largest distributors, I just did a review with them, that middle ground of 10,000 to 100,000 people have finally started coming back to buy.
Yeah.
Which means that they finally depleted inventory. In a way, it makes sense because December was three years.
Yeah
since the COVID peak, right?
Right.
It's taken about three years to flush out that inventory. Our customer counts are coming back at an extremely fast rate.
Right.
Thousands, literally coming back and our distributors are saying the same things to us. Where they had written off material that was written off well over a year or two years ago, they're starting to see that come down at a very rapid rate, too.
Got it.
We're seeing our written off material starting to sell through as well.
Got it. How about lead times?
Lead times have been kind of bouncing around the bottom four to eight weeks, then now with some of the tightness that we're seeing, it's starting to inch up here. But still, we think distribution's largely cleared out at 26 days, and in fact, may be running a little lean. On average, distribution days tends to be high 20s, low 30s.
Right
You're now you're starting seeing this replenishment activity starting to layer in, so, yeah. As lead time expand, I think that'll be kind of the next thing, but it's moving up as things are getting better.
Yeah, I think we keep using that word replenishment. It's really not replenishment. It's run rate.
Yeah.
We're starting to see customers just order to run rate. I don't think there's much replenishment.
Right.
It's just that they finally ran out of inventory. They're not racing to put more inventory in place.
Yeah.
They're just placing orders to get back to run rate.
Right. The point of sales is picking up nicely.
Point of sales is picking up. Customers are coming back.
Got it.
That's what we're seeing.
Right. Okay, maybe we can take some questions from the audience, Jordan, go for it. Yeah, go.
There's a lot of excitement that the underlying power architecture of the data centers will change over the next few years. Given your DNA in analog power management, do you have an opportunity there?
We are probably one of the predominant controllers that is used for AC-to-DC conversion and DC-to-DC, we're seeing an awful lot of activity and growth in our dsPIC microcontrollers, right, our dual core DSPs. That's growing extremely well, and we just launched some new for solid state transformers based on silicon carbide to help build out solid state transformers within data centers. That's something that we've been working on for quite some time with several universities within the U.S. Technology's been there for a long time. It's been in development. To try to get municipalities to really launch solid state transformers, they've not really been as motivated. We think that data centers will truly be motivated to move to solid state transformers. Lots of benefits.
It's tremendous hype on the street right now, all the power management, analogs, semiconductors are through the roof. I'm just wondering, given your piece in the industry.
It takes a long time for people to make a move in power, right? It takes years, right? There's going to be a lot of testing, there's a lot of review on it, because the last thing you want to do is take down a data center, right? Because your transformer blew up or power supply chain. When it comes to power, the movement is very slow and conservative. 800 volts is here, right? It's trickled down from automotive to some degree, right? That was the main powertrain for most of the cars that were out there. It was 400 volts, and then it was 800 volts.
A lot of the technology that was developed for the inverters for cars is what's going into data centers, right? There's a definite real merit there, and it's not as new as a solid state transformer, but that type of power conversion is understood and can be done.
Okay.
I have two questions. Hi. Maybe give us a bit of an update on what you're seeing in China, how that's changed since maybe the start of the year, if at all, and anything on your end, in terms of go to market. We hear a lot about that market, but I'm always curious from your product end market standpoint, if there's anything new or exciting there. M&A, it seems like, with all this new opportunity for some of this lagging edge mature technology or semi content, there's a lot of opportunities. Vijay's talked a lot about them. You seem to have a lot of this already in your portfolio, and Steve seems pretty content to de-lever the balance sheet and pay the dividends.
Doesn't seem like M&A is something that you would be thinking about, but I guess never say never. Some of your peers have reacted a little bit more aggressively, I'm just curious, since you've been acquisitive in the past, would there be some technology that would excite?
Good question. We've been still acquisitive, by the way. We bought a driver product line, TF Semiconductor from Diodes Incorporated. We've done some smaller acquisitions. We typically do probably one every two or three months. Nothing, though, on a very large scale. A lot of them have been tuck-in pieces of technology that roll into Microchip Technology, and we're still looking at some today. In terms of large, transformative acquisitions, our preference would be to pay down debt for a period of time, and then continue with these smaller ones for quite some time. We've got large, high growth drivers in front of us, between PCIe Gen 6, retimers, Ethernet, what's happening in post-quantum security with RT FPGAs.
I could go down a whole list of where we're number one and we're growing. When you've got a lot of intrinsic growth built into the organization now, doing a very large acquisition just is defocusing. We've definitely chosen to do these smaller ones that kind of add to our total system solution. It seems more prudent. In China, many years ago probably, I'm going to say almost six, seven years ago, we started focusing on just what was the export out products. We try not to compete with the local brands locally. That just doesn't seem to work.
A lot of what we focus on is what is not made in China and what we can sell there, like PCIe switches, Ethernet, export products being our microcontroller products, or where we have people that have designed in and are building in trade zones within China. That's where our focus has been.
Great. Thank you, everybody. I think we're at the top of the hour.
Okay.
Thank you, Richard and Sajid.
Thanks, Vijay.
for coming. Always a pleasure.
Thank you for having us.