Everyone, welcome to the last day of Citi’s Global TMT Conference. My name is Elizabeth Sun, and I am in the Citi Semi and Semiconductor Research Team. Today with me is Advanced Energy. Joining me on the stage, we have Steve Kelley, President and CEO, Paul Oldham, Executive VP and CFO, and in the audience we have Edwin Mok, Senior VP of IR. Before we begin, I believe Paul has a few comments regarding Regulation FD.
Great. Thanks, Elizabeth, and thanks everyone for joining us. Just a reminder today, any statements we make are subject to a number of risks. You can find those risks articulated in our recent SEC filings. Also, we reported our Q2 earnings on August 3rd, of this year. Today we will not be providing any updates to the guidance provided at that time.
All right. Steve, let's start with some questions on your semi business. After delivering a meaningful semi upside in Q2, you raised your H2 guidance for semi to grow almost 50% year-over-year. I think earlier this year there was a lot of concerns around you all being very conservative on guiding the semi's growth. What changed over the course of the past six months, and what gave you the confidence to revise up the semi's growth this year?
Yeah, thanks, Elizabeth. I think what we saw in the H1 was an increase in our customer confidence over the next two years, and also an increase in their forecast. We saw a number of new orders that came in that gave us a lot of confidence in the H2, as well as in 2027. One of the actions we took in response to those increased forecasts was to accelerate the opening of our new factory in Thailand. In Thailand, we are just opening a 500,000 sq ft facility, and we will be launching that facility with our largest customers, semiconductor, as well as our largest customers in data center. That is a very good sign for future demand.
Good. We'll get to the facility part of it later. But staying in the semis business, if we look beyond 2026, you just touched a bit on 2027 looks good as well. For components guys like you typically outperform in the early part of the WFE up cycle. If we look beyond 2026, looking into 2027, what will drive this kind of momentum for you basically to outperform WFE?
Yeah, so what we're being told by our customers is they have good visibility for the next two years. That's a lot of visibility. They're also telling us they're constrained by available fab basis. As their customers are able to accelerate the build-out schedule for new fabs for advanced memory, advanced logic, that's an upside for our customers, upside for us.
Normally, your customer typically build inventory at the early part of an up cycle. But I think you talk about you are largely shipping to customer demand. I was just wondering why this time is different, why the customers are not really building inventories.
Yeah. I think it's interesting phenomena. We think we're shipping to demand and not into inventory. I also look at how we manage our supply base. We learned a lot during COVID about managing our supply base. We divided the suppliers into two categories, the reliable ones and the not so reliable suppliers. If you're a not reliable supplier, we'll typically put in place more inventory to give us some insurance against execution challenges. I think the same phenomenon happens with our customers. I think Advanced Energy is perceived as a reliable supplier. We've got very strong factory infrastructure. We're successful in managing our supply chain. So they have to put in place less inventory as insurance for shortfalls from Advanced Energy. So I think we're in the good supplier category, and therefore our customers don't need to put in place a lot of inventory.
What is your lead time?
The lead times?
Yeah, within semi.
Yeah, so in our semi business, we typically operate on a just-in-time model with our largest customers. There is no lead time per se, but we know what they need and when they need it, and we are able to keep up with that.
Good to hear. I want to touch base on the new products that we've talked about for the past two, three years, eVoS, eVerest, NavX. I think good feedbacks continue to be good. Can you remind us where are we on the ramp? Are you expecting them to start contributing more on the growth or share gains next year?
Yeah. We're very enthusiastic about the customer acceptance of our new technologies, namely eVoS, eVerest, and NavX. Where those technologies are used is leading edge. Today we're being designed into sub-two nanometer applications logic, for instance. We expect those products to become more meaningful from a revenue standpoint next year. We're already seeing shipments this year, but compared to the overall shipments we make, they're not that meaningful. We're very encouraged at the adoption of these technologies, leading-edge logic and leading-edge memory, because they're necessary to maintain throughput and yield at these very advanced geometries.
I'm curious within that ramp next year, how much of that is driven by continuous slot wins and how much of that is driven by new sockets?
Yeah. As you see the ramp this year, that's almost all older technologies. They're existing wins from many years ago where our customers are ramping the volume. I think as you move into 2027 and 2028, you'll see a more significant percentage of the buy coming from new products, as those new processes ramp to volume.
Got it. In a longer term question, as semi process become more complex, you talk about sub-two nanometer processes. You need more etch and dep steps, have higher aspect ratios, and you need more maybe demanding process control. Overall, how do you think does that translate into the power demand and specifically for AE's product?
Yeah. Obviously, higher etch and dep intensity is great news for us. It's great news for our customers, too. The reason why there's higher etch and dep intensity is there's more process steps. So the end customers, the fab operators, are trying to build very complex structures, very fine geometries. This requires a lot of steps, and that drives the etch and dep intensity. In addition to that, the technical challenges of dealing with these structures require higher power from our products. Sometimes multiple frequencies and also more advanced metrology. All these capabilities are built into our boxes and they drive higher prices.
Just final question on the semi side. I think you talk about you're starting to gain some traction on the testing side. Can you talk about what do you provide there? How is that different from the front-end guys? What's the expectation of the ramp from here?
Yeah. It's interesting. Our heritage has been in plasma power. Very strong technical challenge, helping our customers plasma lit inside plasma chambers. Over the past few years, we've put more emphasis on what we call our system power business. This is basically the box between the wall and the demands have become increasingly complex, particularly as we move into the AI testing challenges. So, we've had significant success there, getting our products designed into those testers, burn-in ovens, other applications where our customers need reliable system power.
Should we expect this demand to grow with this testing, or is it more about share gains?
We haven't modeled exactly how it's going to grow, but what we know is we continue to expand capacity every quarter for these test applications, so it's a very good business for us. It drives margins similar to what we have in the plasma power business.
Switching to data center side, apparently you guys are a data center company now.
Yeah.
Data center revenue more than doubled in 2025, and you're talking about you're guiding more than 50% growth in 2026. For 2026, I'm curious, is the growth moved from continued build out in your top customer in one top customer in data center, or is it also a broadening of your positions in the other hyperscalers?
Yeah. In 2026, what we've seen is a ramp based on our hyperscaler customer only. We've talked about a second wave customer set, but we don't see meaningful revenue from those customers until 2027 and 2028. What happened in 2026 was a number of our wins from 2025, the design wins, are ramping to volume. These wins typically carry higher ASPs because they're driving higher power requirements. It's basically our ability to execute with a chosen subset of hyperscale customers that's driving our revenue growth.
And just to that higher ASPs, can you help us understand directionally, how does that pricing increase generation to generation or year by year? Give us a range of this.
Yeah
Pricing offset.
Yeah. The cadence right now is every year a new solution's required. That's loosely tied to new introductions from Nvidia and other suppliers. The power requirements go up every year, but the space that we have to operate in doesn't change. Basically we have to deliver higher power, same size box, and that's called higher power density. The challenge is how do you do that, increase the power density, you maintain high efficiency, very high reliability. That's the technical challenge, and that's where Advanced Energy shines, essentially. That's why customers come back to us because we deliver a very reliable, efficient, and high power density solution.
Got it. You have one customer that takes maybe more than half of your data center revenue. Just wondering, are you continuing to gain new contents, new sockets at that top customer?
We focus on a subset of customers and we are gaining share at those customers. Each of those customers has a wide variety of applications that we are designing. Even though we are highly customer concentrated, within those customers, we have some diversification across different applications.
You talk about a second wave of customers earlier. How different is that opportunity from your existing hyperscaler business in terms of product requirements, customer concentration, and content for AE?
When we deal with hyperscale customers, they require a high degree of customization. It is very engineering-intensive effort. That is why we have basically constrained ourselves to a subset of hyperscale customers, because we do not want to get overextended there. I think that has been a good strategy for the company. As we look at the second wave customers, these customers will typically take standard products from us with some minor modifications. The engineering intensity is much less for our second wave customers. This allows us to engage with them without overstretching our engineering.
Does that give you a higher margin, in essence?
What we see is the margin is very similar to what we have with our hyperscale customers. But obviously, since the engineering intensity is less, I think the operating margin is probably a bit better with our second wave customer.
How big is that opportunity, the second wave of customers? I think you talk about it could be a similar size to one big hyperscale customer, but when do you expect to reach that level?
Yeah. We have said that in aggregate, we think the second wave customers could be as big as one of our hyperscale customers. We expect to reach that probably sometime in the '28, '29 timeframe. Hard to gauge exactly. But we're very enthusiastic about the second wave customers as a way to diversify our customer risk.
Good to hear. I do want to ask on 800V. There are a lot of questions around this, I am sure you are receiving a lot of questions on this as well. The first is, as power architecture moves from 800V, how should we think about the change in Advanced Energy's dollar content, both in the per rack level and also per megawatt level?
Yeah. We are heavily engaged with customers on 800V solutions, and the challenge is very similar to the challenge we currently have with 48 or 50V solutions. It is all about power density, efficiency, reliability. It is just the next step for us. Generally speaking, every time an architecture changes, it is good for Advanced Energy because we are a technology leader. We can bring technologies to bear to maximize the efficiency, power density, without sacrificing reliability. That is our value proposition. The way we approached 800V is through a modular approach to our products. We have different modules we have designed, then we can combine these modules in different ways to meet the needs of different customers. That is working pretty well for us, actually. It is a mix and match strategy, because the expectations and standards are changing. Pretty dynamic.
At the end of the day, we do not expect 800V to be significant from a revenue standpoint until 2028 at the earliest. We think 2027 is a design win year for us, and 2028 and 2029 start to see revenue on 800V product.
Any early thoughts on your content?
Yeah, we look at the content, and we think the content goes up like it has in every other generation. Because the power is going up, the challenge is getting larger from a technology standpoint. We think we are in pretty good shape to grow overall content, maintain and improve our gross margins.
Content, that means content per rack?
Yeah. It would be content per rack where we participate. I think it is important in data center to remember that our strategy is not to grow market share necessarily. It is to operate those applications that will drive higher margins for the company.
Something that investors are trying to figure out is when AC/DC moves from the rack to sidecar or a power rack, people are trying to figure out if you guys have the solution for that and you guys are already engaging with customers on that and expecting some.
Yeah. We do have solutions, and we're engaging with customers on the sidecar as well as more conventional architectures. I think that's one of the advantages of a modular strategy. It gives us a lot of flexibility on how we deal with changing customer requirements. I think we're in pretty good shape, no matter which architecture a customer chooses.
And what will be the biggest content opportunities for you guys in this kind of transition? What needs to happen in order for AE to grow your content? As you said, grow your content meaningfully in this one.
Yeah. I think the two key parts of this market are, one, during the design phase. It's having the right engineering team in place, having the right IP blocks in place so you can move quickly. The second is your ability to scale manufacturing quickly and reliably. I think we've got a really good engineering team and we need to add to that team, but it's difficult to find really good engineers. That's the constraint. It's not our willingness to spend money on more engineers, just our ability to find them. But we are expanding. As far as new factory space, I think we've done an excellent job in the Philippines and in Mexico expanding capacity to meet the hyperscaler requirements. We've also done a great job upscaling our development sites because it costs more and more to develop these products with the higher power requirements.
But looking to the future, we have this large factory in Thailand, 500,000 sq ft factory, and that will be the future for our next wave of expansion data center.
Just to double-click into the talk about the kind of bottleneck will be the engineering innovations. What are the most difficult things to do there? Is that efficiency, power density, like thermal measurement, reliability, or is really all of them?
It is all the above. That is the key. How can you do this? How can you juggle these different requirements and come out with a solution that is manufacturable, reliable, and scalable? That is what we have been doing for the past few years. We have built a fair amount of confidence with our customers that we can get the job done. One of the things we try to do from generation to generation is reuse as much as possible, so we are not inventing something new every year, or at least we limit the invention. I think that is the key is focus on a subset of customers and execute 100% on their needs.
Got it. In the longer term, there is also the 800V directly to step down to 12V, rather than using an intermediate 48V and 50V bus. Does that increase or decrease AE's content opportunity? On the one hand side, it is just one big step down, and the other hand side it is a big step down.
Yeah. I think it depends on which customer you talk to, what they want to do. That also drove our decision to go to a modular approach with 800V. Way to think about this is we have what we call a PDU, a power distribution unit. Think of it as a motherboard, essentially, where we could plug in these different 800V modules and you could do whatever you want with it, 800V to 6V or to some other intermediate voltage. It is up to the customer how they want to deploy this PDU. That flexibility is really important at this stage of the game because I do not think anybody knows exactly how the architecture is going to look a year or two from now.
But if you have flexible modules, you are able to experiment and figure out what makes most sense for you.
That is your strength, like customization.
Yes. We are all about customization.
I will take a pause here to see if any questions. Okay. Let us move on to industrial and medical. I think it finally looks on the way to some recovery and some growth there. How do you see the rest of the year shake up and then look forward to next year?
Yeah. The industrial medical market has recovered. There is significant increases going on. I think we were up 11% quarter-on-quarter, 16% year-on-year in Q2. As we look forward, we see each quarter getting stronger, 2027. We are seeing anything AI related, let's say physical AI is driving a lot of the high-end industrial medical business. We operate in the high end of that business. So it is factory automation, robotics, test apparatus, those types of application. That part of the market is definitely strong. We are basically reaping the benefits of our design wins over the past 2- 3 years in industrial medical. A lot of those are starting to ramp the volume, so we are basically expanding capacity rapidly in the Philippines and in Mexico where most of these products are built.
I think it was in Q1, you had to make, for lack of a better word, sacrifices to the industrial medical because other products are ramping a lot faster or demands come in, like got pulled in a lot faster. At this stage right now, do you have enough capacity to support the ramp across your sector?
Yeah. We had a hiccup in Q1. It was really our own problem. We had an execution problem in the company, which we have rectified. We have enough capacity industrial medical. We have basically corrected for our execution problems. So I think we will be caught up with our delinquent backlog by the end of this year in industrial medical. There are no limits right now as far as what we can invest in industrial medical capacity.
Good. I feel like this quote, Telecom and Networking was never really asked a lot. How strategic is this T&N segment market to your long-term target?
We're starting to see more opportunities that are AI related Telecom and Networking. I say that definitely works into our strategy, but as a market, I wouldn't say it's strategic. It's more opportunistic for us. Where we can reuse our existing technology to satisfy the needs of Telecom and Networking customers, we will, but we're not going to develop ground up products for that segment.
What kind of AI opportunities, or maybe just overall, what do you provide there for?
It's basically very similar products to what we provide to the data center customers. You see a convergence there with different companies that need the same basic thing, which is reliable power, highly efficient power, high power density operating.
Good. Some questions for Paul.
Sure.
On the capacity side, maybe can you explain your capacity expansion strategies? You talk about Thailand or expanding a lot of capacity right now, and your customers are also aggressively adding products capacity as well. We heard a lot of doubles over the past couple of days during our conference. Will you be able to keep up with demand if the demand grows more than expected? On the other hand, do you have any concerns for putting too much investment in capacities?
Yeah, it's a good question. We've clearly biased ourselves to make sure we have enough capacity, both in terms of physical capacity and in materials to capture upside. I think we've said that for a while. We do see a ramp going on both data center and the semiconductor market, and we're committed to make sure we can capture those opportunities. That means we've carried more inventory, and we're investing quite a lot more in capital expenditures. Good news is we started this a while ago. We saw this trend coming, and we were able to get in front of it. As Steve mentioned, we started our Thailand facility build-out back in 2023. So that factory is now coming on stream right when we need it, with an ability to scale it very quickly.
Today we have over $3 billion of capacity in our existing facilities, with the opportunity to potentially expand that more through adjacent space, through how we configure things, what we put in those manufacturing line. But with Thailand coming on, that would give us over $5 billion of capacity as soon as 2028. So we feel like we're in good shape from that perspective. In terms of what's the risk of over-investing, I think there's a few things to think about. First of all, the payback on this investment is very fast. Secondly, a lot of these investments are very fungible because we're in multiple markets and we're basically doing final assembly and test. We can move this equipment around depending on what's the actual need at the time.
The third thing is our investments, while they're significant for us, we're not building a fab or a back-end testing facility or something like that. The relative capital investment is fairly small. So all in, we believe that these investments are already built into our business model. In fact, if anything, all we've done is we pulled forward what investments that were already in our 2030 model, because frankly, the demand is running ahead of what we expected.
Yeah, got it. With that $5 billion capacity with Thailand ramp up, I think maybe a quarter ago or four months ago, the number was kind of $3.5 billion- $4 billion. What drove that upset? Where does it come from?
I think it comes from a number of things. I think first of all, we have sharpened our pencil a bit. I think capacity is never one number. It is sort of a range. I think we have tried to look at that more on that context. Part of it is accelerating investments. We have increased our CapEx to capture, make sure we have got equipment and space needed to meet the growth ramp.
While Thailand is ramping, is there a period I would think about there is a kind of underutilization cost? Ultimately, I believe the Thailand facility is margin positive.
Yeah, it is a good question. The investment in Thailand and the other investments capacity were always in our 2030 business model. It is not new or it is not incremental. What has happened is we have pulled that in because we are much further ahead. We are almost getting to the revenue levels in 2027 that we had in our model we put out just years ago, which was at 2030 target. So we have pulled that in. It is contemplated in the model. In fact, this last quarter we said we expect gross margins to exit this current year around 42% and next year around 43%. That also contemplates the investment in bringing Thailand and the other manufacturing investments on stream.
Yeah. Just on the gross margin, you are very close to your target model right now, but gross margin is just a bit below. I think there is tariffs impact in there. There is the ramping cost you talk about, but is there anything else that we should think about that is impacting the gross margin not reaching the target below that?
You hit the big ones, right? Tariffs were not embedded in our original model. So that's a headwind. Also, I think the mix of being more data center heavy is a little bit of a headwind, but frankly, margins are improving across all of our markets. We wouldn't be able to continue to improve margins if data center margins weren't also improving. When we look forward, I think the important thing is, yeah, we're approaching our model from a revenue perspective. We're also approaching it from a gross margin perspective. We believe that the underlying factors that are improving gross margin continue. We still expect the majority of the benefit from mix of new products is ahead of us. So that's a positive. When you ramp this quickly, it's not always as efficient as you'd like it to be, to your last question.
There's still opportunity to wring out some efficiency in manufacturing. We should continue to reap the benefits of volume. While we put out a model that said 43.5%, we think we can certainly go beyond that as the business continues to grow and our new products actually become a bigger proportion of the total revenue.
Just to that point, your analyst day was like late 2024?
Yeah, just two years ago.
Yeah.
Not even two years, yeah.
Are you expecting to update your target model anytime soon?
It's a good question. Ultimately, we will do an update to it, but I think the beauty of the model is it's, when we put it in place, it's a scalable model, which says that there's an opportunity to actually continue beyond kind of what would be the bookends of that model. Clearly, revenue's tracking ahead. Data center's been much stronger than expected. Semi's in a fast ramp, and I&M is finally recovering. I think those things bode well for continued growth. We think the elements of gross margin outside of tariffs and a little bit of mix are structurally intact, which can also go beyond the 43.5%. Frankly, we're quite a bit ahead on operating leverage, because we've grown faster. We think while we haven't updated the model yet, the underlying tenets of the model still hold.
And just know you are doing so well in semis and data center, and industrial and medical is recovering. But when you think about investment across the segments, how do you balance this kind of making investment in semis, in data center, or doing some M&As in the industrial and medical?
Yeah. So the way we think about it is, we've really doubled down our investments on semiconductor and data. Those are largely internal investments. We're invested significantly in inventory to put ourselves in a position to capture upside and be ahead of the ramp. We're investing in CapEx. We're spending roughly 2x the CapEx we would normally spend. So while that's running high today, as we get through these ramps, that can move back to a more normalized level of 3%-4%, but we're certainly making those investments today. We think from a capital allocation perspective, M&A certainly is an important part of our strategy. We think that that's most tuned to the industrial and medical market, where it's a much broader based set of customers and set of competitors. And M&A is a way to build a bigger beachhead in that market.
Because ultimately, we'd like to have three very strong markets that we address going forward.
And just to wrap up, is there anything that you think investors are not understanding Advanced Energy story very well, or anything important that we didn't really touch base, Paul?
Yeah. I think the key thing is that we have multiple legs of growth. If you look at our customers and the market dynamics, that growth is projected to continue. It may not always be linear every quarter, but it's projected to continue over the next couple of years, and we're tied right into those leading-edge technology that support it.
That's it.
No, thank you very much.
All right. We're out of time. So thank you, everyone, for coming, and thanks, Steve and Paul, for joining us today.
Thanks, Elizabeth. Thanks, everyone.