SolarEdge Technologies, Inc. ($SEDG)

Welcome, and thank you for standing by. I would like to inform all participants that this conference call as well as any Q&A may be recorded and made available to clients of JPMorgan. Where a company is presenting, any recording may also be posted on their website. Views and opinions expressed by any external speakers on this call are those of the speakers and not of JPMorgan. Parts of this conference call may also be reproduced in JPMorgan Research. Participants are prohibited from posting, sharing or distributing any part of this call or its content on social media platforms or any public forums without prior written consent from JPMorgan. If you have any objections, you may disconnect at this time. This call is intended for JPMorgan clients only. Press participants are not permitted on this call should disconnect now. Unless otherwise permitted by internal JPMorgan policy, members of JPMorgan Investment and Corporate Banking are not permitted on this call and should disconnect now. I would now like to turn the call over to your host, Mark Strouse.

Great. Good morning, good afternoon. Thank you, everybody, for joining. My name is Mark Strouse. I cover clean energy and power infrastructure here at JPMorgan. We are very happy to have SolarEdge with us this morning to talk about their new solution for kind of the AI revolution here. So very exciting stuff, very early stage. So I think this will be helpful for everybody. There is a white paper that's available. Hopefully, the link that was in our website worked for everybody. But if you have any issues with that, let me know, and I can forward you the white paper. We are going to start with a roughly 20 minutes or so kind of presentation here. There will be plenty of time for Q&A at the end. So I think what we're going to do is the raise hand function on Zoom here. If you would prefer to just ping me on Bloomberg, I can ask the question for you. So with that, let's get going. Shuki, I'll hand it over to you. Thank you very much.

Thank you, Mark. Thank you for having us, and thank you, everybody, for joining us this afternoon or this morning. We are having with me Asaf Alperovitz, who is our CFO, and Meir Adest, who is our Co-Founder. He will take you through the presentation that will be focused on the product, the technology and where the market is heading with regard to our solid-state transformer, with the revolution that is coming to support AI factories with DC. And after that, we'll open the session for questions. With that, Meir.

Thank you. Good morning, everyone, and thank you for joining us today. So as Shuki said, my name is Meir Adest. I'm one of the founders of SolarEdge. And I'm here to talk about what we believe is one of the most consequential infrastructure challenges of our time and how SolarEdge is uniquely positioned to solve it. The AI revolution is no longer a future promise. It's happening right now. Billions of dollars are flowing into the data center build-outs. But here's what most people miss. The long-term bottleneck isn't compute, it isn't chip, it's power. And that is exactly where SolarEdge comes in. In today's presentation, I'll review a recent white paper, hopefully the first in a series, laying out a maturity framework for data center product -- for data center power architecture and show you why SolarEdge has a credible differentiated path to lead this transition. Let me walk you through it. Before we begin, I'll direct your attention to the safe harbor statement on screen. This presentation obviously contains forward-looking statements. With that noted, let's get into the substance. So let's set the stage. According to McKinsey, over 100-gigawatts of AI data center capacity is expected to be built between 2026 and 2030. Just look at this buildup curve on the right side of the screen. It's accelerating year-over-year, reaching 31-gigawatts annually by 2030. But here's the problem. The grid is the gating factor. Generation and transmission constraints are limiting new data center build-outs with substantial power shortfalls expected. Grid connection capacity is increasingly the ceiling for expansion, and this is where the current architecture completely breaks down. Look at the diagram on the bottom of the slide. In a traditional AC data center -- in the traditional AC data center, electricity passes through 5 to 7 conversion stages from grid to chip, transformer, uninterruptible power supply, UPS, PDU, power distribution unit, ramp power supply, each one takes its cut. The total system loss within the data center itself is 9% to 16%. We call it the conversion tax. Now NVIDIA actually puts the conversion tax at an even higher number, up to 30% because they factor-in the transmission losses from the power plant all the way to the data center. That's an interesting number because it also makes a compelling case for on-site generation, cogeneration, photovoltaics and similar solutions that bypass transmission losses entirely. But that's a discussion for another day. Today, we're focused on what happens once the power reaches the data center fence and even there, the waste is staggering, and it gets worse. Consider the density trajectory. The H-100 racks needed about 40-kilowatts each. NVIDIA announced Kyber racks for Rubin Ultra chips, which are approaching 1-megawatt of power per rack. That's a 25x density leap in just a few product generations. You cannot push a megawatt through a legacy AC chain. The physics don't work, the thermals don't work, the economics don't work. Now in a power-constrained world, the most direct lever you have is efficiency, and let me be direct about what that means economically. When your data center is capped by its grid interconnection and most new builds will be, every watt you save from conversion losses is a watt that goes into compute, more megaflops, more tokens generated, more revenue. The relationship is linear. Cut your losses in half and you can serve meaningfully more AI workload from the exact same grid connection. It's not an engineering detail. It's a revenue multiplier. So the question isn't whether the industry transitions to DC. The question is how fast and who leads? To bring clarity to this transition, our white paper introduces a five-stage maturity framework for data center power architecture. I think this is a generally useful framework for the industry, so let me walk you through it. Before we go into the stages -- through the stages, let me orient you on this slide. What you're looking at is the power passed through a data center, reading from left to right. On the far left is a 34.5-kilovolt medium-voltage grid connection. And I want to stress this point. Any AI data center will connect at medium voltage. And even though edge data centers, the smaller ones might connect at the 12-kilovolt or so, the large-scale data centers require a 34.5-kilovolt connection. That's the starting point. From there, power flows through the gray space, that's where the power conversion and distribution equipment lives and finally goes into the white space, which is where the servers and the GPU rack is. So that's the structure left to right. Now as we move down the road from stage 0 to Stage 4, watch how the gray space simplifies, fewer boxes, fewer conversion steps, fewer losses. Stage 0 is where most data centers are today, pure AC architecture going all the way to the rack. So it's about 84% to 91% system efficiency capped about 200-kilowatts of power per rack. This is the legacy world. Stage 1, the white space retrofit. This introduces 800-volts DC racks, but bolts an AC to DC sidecar converter into the white space next to each rack. We get DC to the rack, but you're still running everything through the older AC backbone. You've added complexity, consumed floor space and are still at low efficiency. This is a bandaid, it's not a solution. Stage 2, hybrid power distribution. This starts to get smart. Here, your white space is 800 volts DC. You introduce DC UPS connected in parallel. So during normal operation, power flows directly from the source of the load without passing through energy storage. So this is better, but you still have the complexity and footprint of legacy AC components. Stage 3, basic solid-state transformer. This is where solid-state transformers enter the picture. And SST replaces the AC-to-DC conversion stage. This is a meaningful step forward. But at this stage, you might still need a step-down transformer to bridge higher grid voltages as some of the solid-state transformers could only accept 10- or 12-kilovolt input. And a separate to low-voltage distribution units are also typically needed. SST efficiency typically peaks at around 98.5%. So this is good, but it's not the end state. And finally, Stage 4. This is where it becomes truly DC-native. This is the destination. A high-efficiency SST above 99% connects directly to 34.5-kilovolt medium voltage grid. DC distribution is built into the SST and DC UPS capabilities are integrated into a single platform. No intermediate transformers, no separate distribution units. Look at the gray space in this row. It's almost empty, one streamlined path from grid to GPU. Now here's the key takeaway. Most of the market today is stuck in Stages 0 and 1 with early emergence of Stage 2. That's essentially bandaids on top of AC infrastructure. SolarEdge is targeting Stage 4 directly. We're not implementing we're leapfrogging. So what does our Stage 4 solution actually look like? It starts with a direct connection to the 34.5-kilovolt medium voltage AC. That's the standard for large industrial loads, including modern data centers. No step-down transformer is needed. A solid-state transformer is being designed to deliver above 99% conversion efficiency, taking that medium voltage AC and delivering clean 18-volt DC directly to the racks. Below the SST sits an integrated DC UPS, an uninterruptible power supply, connected in parallel to the DC bus. During normal operation, it's not in the power path, so there's no AC-to-DC-to-AC round-trip penalties. But it's there when you need it for backup and critically for active peak shaving. And we've designed in per channel monitoring with fast isolation -- with fast fault isolation. This isn't about -- just about efficiency, it's about intelligence and safety at every point in the power chain. And the result, up to 98% total system efficiency from grid to compute. Think about what that would mean in a power-constrained world where legacy systems weigh 10% to 16%, we're targeting a loss of just 2% to 4%. Every point of efficiency recovered is a point of revenue gain, more racks, more compute, more AI workload from the same grid connection. Let me go a little deeper into the heart of the system, our solid-state transformer. This is the product you're looking at on the right of the screen, built into a 20-foot container and designed for the industrial demands of modern AI data centers. Four things to take away. First, higher data center revenue. Above 99% operating efficiency means power that was previously lost to conversion and cooling is now freed up for additional racks. Efficiency translates directly to revenue for operators. Second, maximum uptime. The architecture is modular and redundant, delivered in 2- to 5-megawatt building blocks. Cell level redundancy within those blocks means there is no single point of failure. If a module needs service, the system keeps on running. Third, comprehensive grid support. Our SST performs active harmonic filtering and delivers ultrafast transient response. And this is an area where our heritage really pays off. SolarEdge has spent 20 years navigating grid support standards across dozens of countries through our PV inverter business. We know grid codes inside and out, reactive power, voltage ride-through, frequency response, and we're bringing all of that know-how into the data center market. In an era where grid operators are increasingly concerned about power quality from large data center loads, this is a major differentiator. We don't just take power from the grid. We assure the grid maintains clean power. And fourth, future-proof by design. The platform supports 800-volt to 1,500-volt DC outputs, OCP plus/minus 400-volt configurations and medium voltage input connections up to 34.5-kilovolts. Whatever the rack standard looks like in 3 years, our platform is ready for it today. Now a fair question is why SolarEdge? Why should you believe we can deliver on this? And I have two words for you, expertise and scale. On the expertise side, SolarEdge has 20 years of leadership in DC power electronics. This includes 6 years of dedicated medium voltage SST development so far. We are the market leader in DC-coupled solutions. The technologies we've perfected for distributed energy, granular power tracking, advanced string management, rapid arc fault detection. These all translate directly into the rack-level power control and safety demanded by high-density AI workloads. On the scale side, we have an operational track record exceeding 60-gigawatts. We have over 140 million units deployed worldwide. We have manufacturing and supply chain readiness at gigawatt scale. This isn't a start-up prototype. This is a proven industrial company extending proven technology into an adjacent market with enormous demand. Let me put this in an industry timeline context. So 2026, that's now. We're in the proof-of-concept phase. NVIDIA's Vera Rubin is in early production. The industry is beginning to validate 800-volt DC architecture. 2027, initial deployment at scale. Most of it is still -- will still be sidecar-based Stage 1, with early SST validation happening. Rubin Ultra is introduced with around 600 kilowatts per rack. And then later 2028 and further, this is when it gets real. SST-based solutions, Stages 3 and 4 become the standard. NVIDIA's Feynman architecture arrives likely heading to a power consumption of about 1-megawatt per rack. At those densities, Stage 4 native infrastructure isn't a nice to have. It's a necessity. SolarEdge intends to be there when the industry needs it with Stage 4 solution ready for the next-generation compute racks. Before we open for questions, let me just leave you with this. The AI revolution runs on power, not just any power, efficient, reliable, intelligent DC power at unprecedented scale. The legacy AC infrastructure that serves us for decades has hit its physical and economic limits. SolarEdge is building the power platform for the next generation of AI factories. We bring 20 years of DC expertise, 60-gigawatts of operational track record and Stage 4 architecture that we believe is the most integrated, most efficient solution in the industry. The conversion camps is over. The DC era is arriving and SolarEdge intends to lead it. We're happy to take any questions.

Great. Okay. That was very helpful. Thank you. All right. [Operator Instructions]. So maybe on that last point around kind of the timing kind of being more mass deployments in '28 and '30. That's a question I get quite a bit from investors. So understand kind of '28 to '30 mass deployments. Can you talk about the lead times that you see in order activity? So if that is your expectation for customers ramping, when would be a reasonable time that we start seeing kind of real purchase order progress with customers here?

So first of all, thank you, Meir for taking us through a deeper dive into the technology. I understand that this audience is mostly about POs and revenue. But it was important for us that everybody understands where the differentiation is and where the market is heading according to our understanding. And if you look at the market evolution, one would see that the data center -- if you are building a new data center today, most probably you will start with a sidecar solution in order to support your 800-volt DC installation. But at the same time, you'll start POC-ing, piloting, testing, everybody is using a different term in order to see how -- which vendor of SST you're going to rely on, whether it fits into your overall infrastructure and architecture and then start "playing with it or testing it" in order to see that it works in your data center. And we believe that what we will see in 2026 and 2027 are essentially these things, right? POCs, maybe small orders in order to support some small tests in existing data centers or in labs of that sort. As we said earlier, we expect revenue not before 2027 because we are looking into the road map that NVIDIA have published, which is what Meir is taking us through. So POs in a massive scale are going to be for 2028 and beyond.

Okay. On that point, on NVIDIA, I think that's another question that we get. Just kind of the go-to-market, can you talk kind of talk about your relationship with NVIDIA? Are you on their list of partners for this 800-volt DC ecosystem? And then are you working with anybody else with -- that has kind of heavy data center experience, just off the top of my head, a company like Vertiv, for example, anybody else?

No, it's a great question. So as Meir indicated, we are taking a technology that has been -- the building blocks have been designed and worked on for the last 20 years, and we are implementing them in an adjacent industry. So the go-to-market aspect is something that is new for SolarEdge in this particular industry. And what we -- as everybody knows, right, it's NVIDIA offsetting the tone as an industry horizontal maker, if you will. And then you have between hyperscalers, Neoclouds and players like Vertiv and others who have been in the power electronics business for data center for years. It's a small set of potential customers or prospects. In the last several months, we've started conversations with different players in the ecosystem. We -- most of the conversations have been around technical aspects of the solution. So people get to appreciate what we're bringing to the table and the expertise that we are bringing. So at this stage, we are talking to, I would say, the entire ecosystem, not the entire, but we are open to talk to different types of players in the ecosystem, and we are talking to some of the players there. The feedback that we've received so far is twofolds. Number one is our understanding about where the market is heading and what Meir shared with you is actually supported by what these people are saying. And the second thing is there is a tremendous appreciation to where we are, both in terms of expertise, but also in terms of where we are in terms of the development of the prototypes. So we feel that at this stage, our engagement with the ecosystem is good, and we are open to engage with different players in the ecosystem. We've not said no to anywhere.

Okay. Great. Getting some questions coming in from investors. Kind of on that point, or -- at least initially, are you focused more on kind of hyperscaler customers? Or are you looking at the co-locators? I mean who is kind of the target customer base? At least initially out of the gate here.

So I'd like to emphasize here that this solution is mainly -- is targeting the AI factories, if you will, the new AI data centers. So co-los are getting there definitely. The hyperscalers are obviously there, it's a major part of their business. And the Neoclouds are -- this is what they've been doing. So it mainly depends on each of these particular players where they are in their evolution towards switching their data centers into AI factories that are based on the DC architecture. As you may imagine, the Neoclouds are probably more advanced in that direction because that's not the entirety of their business, but it's a major part of their business. But all three segments that you mentioned are interesting.

Yes. Okay. So I think you used the phrase leapfrogging in the presentation. This is completely new stuff. But I mean, who would you view as your competition? I mean there's other solid-state transformer companies that are out there, Vernova, Delta, et cetera. I don't know. I mean just maybe -- I don't know, maybe can you just kind of talk about the competitive environment? Or the -- are those solutions that are out there kind of to the extent that it's known, can you talk about where they are going with their product and how that compares to where this solution ultimately fits-in?

Well, I'd say there three types of competitors that we're seeing without getting into specifics of what each of them is doing. One is, let's call it, power conversion companies who have been supplying to data centers for years and want to add this offering, this type of offering for the next generation of data centers, large companies. Second is start-ups that have come up in recent years trying to make products. And as the market is heating up, we're hearing of all kinds of other companies from other segments of the market of power electronics trying to grab into this market as well. I should say though, moving to medium voltage is a whole different ball game. There are a lot of challenges that don't exist in, for example, in the solar, low-voltage DC. That's the -- that's kind of the easy side of the known engineering. The medium voltage side is a lot more complicated, especially because transformer and solid-state transformer, they both have transformer in their name, but they're totally different. Transformers is a lot of copper and usually a lot of oil. Solid-state transformer is a power electronics device, and the know-how of how to make power electronics devices that work at these kilovolts is challenging. We were actually -- I didn't write explicitly, but it does appear in the white paper. Basically, what we're doing here is we're using the development effort that we've put in for about 6 years for a direct utility scale inverter which connects directly to the medium voltage, and we're now using it with power-point in the opposite direction as a load for the data center. So we have a lot of know-how that came from years of work in this area and just now is becoming relevant for the data centers that would be challenging for companies that don't have this experience. And again, this is without going into the specific because I don't know what exactly every competitor does.

Yes. Okay. Another question from investor. I believe you guys are based on silicon carbide. Can you talk about your views on gallium nitride? I believe Enphase is talking about a Gallium Nitride Solution. So would you view them as a potential competitor?

Yes. So I don't know what exactly Enphase is planning to do. But in general, gallium nitride is -- it's a good device. We actually have products that we're also developing with gallium nitride, but mostly lower-voltage switching devices. So if you're doing a micro-inverter, which is all low voltage, 480 volts AC and 400 or 800-volts DC, that's fine. When you go to higher voltage levels, usually, you need higher voltage switching devices. Now there have recently been announcements of GaN devices with higher voltages. Just as a general remark, if you look at what happened in the silicon carbide industry, it could take a few years between when a new switching element is announced and when it's actually reliable enough over time with a high-yield production line and something that's stable enough that you would feel comfortable using in the field. So personally, I would be a little hesitant, but -- and that is something that silicon carbide is a much more mature technology for these higher voltages. Again, not going into specifics of what competitors are doing, just a general observation about the mature level of these different technologies.

Okay. That's great. Another question from investor, and I think this is -- I've received something similar from several people now. So you mentioned in the prepared remarks, roughly 6 years of experience in solid-state transformers. Just to kind of level set, do you have an SST product right now in the market? And then can you talk about some of the milestones that we should be looking for as far as kind of when you would have this working product, what needs to happen for that? Is there a technical breakthrough, manufacturing breakthrough? What are some of the things that we should be evaluating as investors?

So we don't have such product in the market today. I mentioned that we had 6 years of experience. Basically, we were developing a utility scale inverter with direct connection to the medium voltage. And we kind of reached the point where we had a prototype, we proved that the technology was working. And from there, it was a scaling issue. But then we decided, we had a business decision just a couple of years ago that at least at the time, the utility scale market was not something -- the PV utility scale market was not something that we wanted to pursue. So we completed the development and then froze the project. And recently, as this 800-volts DC and solid state -- this need for solid-state transformers came from the market, we had the realization that basically this is almost identical hardware to what we were developing with obviously different software to have the power flow in the opposite direction from the AC to the DC instead of vice versa. So it's not a product that we were selling, but I think the technical challenges, definitely the major ones are behind us and now it's a scheduling [indiscernible].

Okay. Yes, that makes sense. Another question on the Infineon partnership, which I think you announced a couple of quarters ago now. Can you just talk about who's responsible for what is Infineon bringing to the table in this partnership? Can you, at least at a high level, kind of talk about how the economics of that partnership work? Is it exclusive? Just kind of revenue or market split in that relationship?

So basically, Infineon is a supplier for the SiC that we're going to be using in the device. There's -- obviously, we're going to pay them for the devices, but there's no revenue share. The IP is SolarEdge. The design is all SolarEdge. There are no complicated commitments here. They're a supplier, and we're working together jointly in application engineering and making sure that the devices fit exactly the use case that we need for the product.

No, similar to the way that we work with them. We've been a supplier customer partnership for many, many years. So it's the same type of relationship. They are providing a very important component in our product. And we have to work closely with the engineering teams have to work closely with each other, but that's the nature of the relationship of the partnership.

And then a lot of questions coming in on how you're thinking about the addressable market, which maybe it's a bit early in that process. But I think somebody is pointing out that Infineon is pegging the market at over $1 billion in 2030. I'm not sure if that's an overall market or if that's their kind of -- their portion of the market. But just how would you frame that conversation for the context of SolarEdge?

I believe we did -- we did mention in our Q4 earnings call a couple of weeks ago that we believe that the SST product and the data center opportunity for us as the SolarEdge can be in the billions of dollars. This would be the SAM, the serviceable addressable market for us, specifically for the SST product. This is, of course, based on our view that as Meir presented, 100-gigawatt or more of the AI data center capacity will be built in the U.S. between now and roughly 2030, as we've shown in the webinar. As it relates to Infineon, we've listened to the call. So I think they related to their specific opportunity, and they provide components for overall solution. So it's only a portion. And as you noted, I don't think you can extract from that number how many billions related to us, but it's really a portion.

Okay. Another question on kind of front-of-the-meter versus behind-the-meter. So obviously, this is this is saving some of that conversion tax of the grid. If the market ultimately gravitates towards behind-the-meter, does that matter? Do you have a play in either of those markets?

I don't think it should matter from a technical perspective, the product could work on either side of the meter. If there's more local generation, there might be other opportunities, which could be relevant. But as I said, I think that's probably the kind of discussion for another day.

Okay. Let's see. I think you addressed some of this in the prepared remarks, but a question, kind of the specific products that you're offering to customers, is it just SST? Are there other components that we should be aware of? Where do you have an advantage in know-how, et cetera, that gives you an edge?

So as I said, the edge is the experience we have in power electronics and the know-how of DC, both from the power conversion side and from the storage side. So as I said, there's the SST product itself, which also has power distribution units, which in other solutions sometimes are external, but they're going to be part of the solid-state transformer and also the DC uninterruptible power supply, which allows both uninterruptible power supply in case there's an outage, but also peak shaving and similar to what we're seeing in energy management in the solar market. So again, this is knowledge that transfers over to this new market.

Okay. Sorry, hopping a bit around the place here. But going back to the product and the technology, are there other adjacent products or adjacent solutions that you're looking to expand into? I mean, obviously, you don't need to say it by name, if there is something. But I mean, should we think about what you're working on now is kind of the solution that you're looking to deliver a few years from now? Are there additional things that you're potentially layering on top of?

So obviously, the opportunity here in the AI data centers is huge, and our focus is on this particular market. As we are going to engage with customers and to Meir's point about the DC UPS, for example, or other added value that we can bring to the table, we'll obviously address that and see where we can add value to the customers. But the market segment that we are addressing -- that we plan on addressing right now is the data center that we talked about, the AI data center.

Its somewhat related, you have a battery business? Are you going to try and leverage some of these data center relationships to sell your Battery Solution?

So we haven't gotten to this point yet. And like I said, if it adds value to our customers, if we feel that we can add value to them, as you know, these customers are very large customers. They have the ability to source their own battery solutions if they're looking into terawatts of battery solutions. But if there is some area in which we can add value to the picture, we'll definitely look into that.

Okay. Maybe switching to the manufacturing here. Would that be something that you would look to take in-house? Would you use outsourced partners? Would you look to build in the U.S.? And are there incentives under the IRA or One Big Beautiful Bill for building in the U.S.?

Yes. So the first thing that is important to note is what Meir had mentioned in his presentation. SolarEdge has already manufactured more than 60 gigawatts of inverters and power electronics that are installed worldwide, sitting out in the wind, in the snow, in the rain, whatever. So we are very confident in our ability to actually scale up the business. Initially, call it, in the first 12 months, we are going to rely on our own almost R&D manufacturing, if you will, in order to produce the first units that are going to be used for testing, for pilots, for POCs. Later on, and we've done that in other products as well. We partnered with two of the largest contract manufacturers in the world. These manufacturers, they have -- we've had excellent relationships with them. They know their job very, very well. And they have facilities globally, whether it's in the U.S. or elsewhere. As for the specific location or the specific contract manufacturer, it's early to tell. They are -- they will have -- we will have to show them that we have business that justifies -- that is justified. And at the same time, they will have to show us which facility can actually support us in the most cost-effective way that is also going to be beneficial for our customers. The U.S. is definitely one of the options. But if our customers are going to have other preferences, we'll have to go there.

Okay. Another question. Can your solid-state transformer be used with Google or Amazon chip ecosystems? Or is it just NVIDIA?

Yes. So that's what I mentioned, I think, in one of the slides, the OCP. So 800-volts is an NVIDIA standard. Google and Meta are part of OCP, the Open Compute Project. They are -- it's not final yet, but it looks like their standard is going to be very similar. It's going to be plus/minus 400-volts with a center cap, which is grounded, and we're designing the SST, so it will fit for that. So that shouldn't be a problem. I'll say more than that. Everybody is starting with 800-volts, but basically, the initial motivation was to reduce, among other things, reduce production losses in the copper going into the racks. And conduction losses are a function of current and the higher the voltage, you could have lower current and therefore, lower losses. So they started with 800-volt. Some of the customers we're talking to are very interested in going also higher. So the low voltage directive is anything up to 1,500-volts DC. And it's possible that in the next generation, they might want to go above 1,000 volts, maybe 1,200 or 1,500 volts. And we're built since this SST, what we've worked on for years, it started as something which could accept up 1,500-volts. We have all of that built-in so that whatever flavor a customer asks for, we're confident that we'll be able to serve it with the SST.

Okay. Some more questions from investors. These are great, by the way. So please keep sending them in, everybody. Okay. What is SolarEdge's 800-volt opportunity in the sidecar approach versus in the full SST approach?

As I said, we're leapfrogging it. I'm sure there's an opportunity in the sidecar approach. It's more expensive. There's more losses. It doesn't make sense long term. Even NVIDIA is saying the sidecar approach is relevant for the Vera Rubin. But when you go above 300-kilowatt or so per rack, it just doesn't fit. So there might be an opportunity there for this year and maybe next year. We don't think that's worth defocusing. We're focused on the long-term solution, which is pure 800-volt going in all the way from the gray area, all the way to the GPUs.

Got it. Okay. And then just a clarifying question. This investor is asking, are you formally partnering with NVIDIA? Maybe it's semantics, but...

So you have to define formally. But we -- obviously, NVIDIA, they've established this standard. They're pushing it very well. We're having conversations with NVIDIA, but we haven't announced any formal partnership or anything of that sort with NVIDIA.

Okay. Another one here. Let's see. Can you talk about which parts in the AI data center space that you are initially focused on getting traction with customers? And have you already sent -- I think you talked earlier about kind of the pilot activity and everything. Is that already happening? Have you already sent samples of these products for customers to test out?

No, we haven't yet. We haven't sent yet. We've started -- as I said earlier, we've been engaging conversations with different players in the ecosystem. They share the same perspective that we are sharing about timeline and the way that it's going to progress. And these are where the discussions are. Once they get comfortable with our solution and where we are heading, we'll start talking about shipping samples and making progress in parallel to the evolution of the market.

Okay. Somewhat related to this next question. Can you talk about the main hurdles that you're trying to solve now? Is it really just kind of a proof-of-concept with the customers, getting them comfortable with your technology? Or are there further R&D steps that you're looking to do to kind of technically improve your efficiency or anything like that?

Yes. It's mostly having these conversations with our customers, making them feel comfortable with the technology. And also different customers have different demands, indoor versus outdoor, what level of power fluctuation are they interested in, what input voltage do their data centers need, all kind of -- a whole list of things, which are -- so marketed there's -- it's not a huge number of customers. It's very different from the residential solar business in that respect. But each of them is very technically savvy and has their own concerns and their own requirements. So having these conversations.

Okay. Another one from investor. Why is this not something that other inverter companies could do? Obviously, you've got -- you talked about with the SST, you've been looking at it for the last 6 years or so. Aside from maybe that first-mover advantage, are there patents? Any other thing that you would call out that would make you a more kind of sustainable competitive advantage here?

Obviously, there are many patents that we put into this over the years. And I won't be surprised if we'll see competitors migrate to this market as well.

Yes.okay. A financial question. What is the incremental CapEx that will be needed in '26 and '27 for the SST product? And then what is the mix of CapEx between your base Solar business versus the SST going forward?

So for this year, for 2026, we noted that CapEx will range anywhere between $60 million and $80 million. We did note the majority of which will be invested in expanding our production capacity, mainly in the U.S. for inverters, optimizer and batteries. As it relates for this year, specifically for the incremental required CapEx to support SST, we expect it to be in the, I would say, high-single-digit, not anything more than that. And by the way in terms of OpEx, something in that range, maybe more towards the mid-single-digit. From managing our investment and cash perspective, we believe this spend keeps us on track on the development timelines and efforts and focus while, of course, maintaining a disciplined approach towards the resource allocation. As it relates for next year, we are planning, as you may know, to have an Analyst Day in September. So that time frame, we may be providing some more color for 2027 and beyond.

Okay. All right. Another one. So the investor asks, so they understand the initial focus is on the data center market, but do you see other use cases for other end markets or commercial customers over time?

I'm not sure what the official answer is. But from a technical standpoint, I believe this is a product that could be relevant and interesting in other markets as well. But as Shuki said, our focus is on the data center.

Okay. Let's see. Some of these -- I'm sorry, everybody. I'm just kind of going through some of these that are coming in. I think we've gotten to a lot of these. Okay. Yes, I think that we are at a stopping point. That was great. Actually, sorry. Okay. All right. A clarifying question from an investor. When you say customers, do you mean hyperscalers? Or are you talking about some of the other parts of the ecosystem like Vertiv or an Eaton?

So first of all, we refer to prospects at this stage. You asked whether we got POs so we haven't received POs yet. So these are prospects. And as we said earlier, we are -- at this stage, we are openly discussing our offering and our advantages, both to hyperscalers, to Neoclouds and to power electronics into the data center-type players. There is interest, I would say, from all 3 segments. And as we make more progress, we'll determine whether we proceed with all of them, with some of them. It's -- as Meir said, this is not a market with dozens of prospects. At the end of the day, it's a small set of prospects, and we are engaging with most of them.

Okay. One more question here. On the first part, I'm not sure which you're going to be able to say, but any framework around the dollars-per-megawatt? And then the second part of the question is, how are you thinking about that dollar-per-megawatt relative to the value that you're adding by replacing all of those prior conversion steps? Obviously, the products might be -- maybe it's a premium to other technology that's out there. But if you're really eliminating that whatever it was, 10% to 15% type conversion tax, I mean, how valuable can that ultimately be?

So it's a great question. And I think that what we're trying to emphasize here is as opposed to maybe other markets in which the conversion tax is translated into savings, in the utility bill.And here, it's a different discussion. Here, the value is significantly, significantly higher than that because the conversion tax basically reduces the potential revenue of the data center. If you can have -- instead of having X number of GPUs, you can have X plus 10%, then it's 10% additional incremental revenue coming from the data center, it's lower cost per token. It's all the good things that make the ROI on the data center significantly, significantly better. And because of that, if you look at the value only, then you're talking many, many, many dollars per watt. Obviously, the customers are going to come back to us and have another -- some other arguments. But if you look at the value only, the value is very, very significant because it helps them maximize their return on investment -- on investments that are in the billions. And this is how we look at it.

Yes, makes sense. Okay. All right. With that, I think we are at a stopping point. So thank you so much. Thanks to everybody for sending in the questions. Those are great. SolarEdge, this was amazing. Thank you so much. We look forward to tracking your progress here.

Thank you very much.

Take care, everybody. Bye.