Okay, we can get started. Okay. Okay. Great. Welcome to the Navitas Fireside Chat. My name is Mark Lipacis. I'm the Senior Semiconductor Analyst at Evercore ISI. We're very excited to have Chris Allexandre, and I'm sure I didn't get the pronunciation- Almost. Almost. Okay. President and CEO of Navitas. He joined in September of 2025 and brings 20 years of semiconductor leadership experience, including senior roles at Renesas, NXP, onsemi, Fairchild, and Texas Instruments. With that, we're very excited to talk about his priorities for Navitas. With that, Chris, welcome, and thank you for joining. Thank you. Welcome to be here. Excellent. You joined Navitas recently. It seems to be an important transition point for the company. You came in, you reviewed the business. Where do you see the opportunities? Where are you going to take the company? Where is the trajectory? Where is it going? Okay. Thank you, everybody. Thank you for your question, Mark. I'll start by going back to where we were and where we're coming from. The company started with GaN. Okay. We're part of the few companies that kind of started GaN back more than 10 years ago. We pioneered GaN to scale, right? The first market that took GaN to scale was the mobile market with the vision that you need to put more power and more power density in a small footprint, right? We started that. We acquired, in 2021, a company called GeneSiC that came from the high reliability, high density, high power, high voltage SiC technology, right? We've combined those to become a pure-play, wide bandgap company. What's happening today is those technologies are becoming mainstream in what I call the high power markets, okay? The thesis of more power, more density, more reliability is happening across the board of the high-power market, which I call data centers. We'll talk about it, okay, with the 800 V HVDC. The grid, because to power data centers, you need to change, upgrade, and kind of change the architecture of the grid. High-performance compute, which is something that people don't pay attention to, but you saw last week at MGX and Computex that Jensen talked about was new, reinventing the PC. I'm talking about super high-end computing, GPU-enabled, that needs very high-power chargers. As well as everything needs to be more efficient and high-power, which industrial. That's kind of where the company's going, and we've done a big pivot using those two technologies, mainstream for those markets. Got you. Let's just talk about, maybe if you could just drill down on that. It seems to be advantageous to have NVIDIA mention you as a partner these days, and so you talk about the MGX ecosystem for 800 V DC AI infrastructure. Maybe could you just go one more layer deep? What is that collaboration? Okay. NVIDIA is driving for the entire industry, for them, but for the entire hyperscalers, a big change of architecture to essentially get to high power, high density racks. Again, I use the example of the mobile chargers, which was moving from 20 W to 100 W in the same footprint. It's exactly what we are doing on the hyperscaler side and data center, is how do we get to much higher power, much higher density of power within the same rack. This 800 V is the transition that will enable that. Okay. Our partnership with NVIDIA and the other hyperscalers come from the fact that to achieve that, you need to move to this 800 V busbar, replacing 50 V, which is commonly used today. As you move to higher power, if you want to keep the current to a lower level, you just have to raise the voltage. This is a complete change of architecture. As you move to embed this DC/DC conversion inside the rack, you have to move from silicon to GaN. That's where GaN became mainstream. I think what NVIDIA and the others are doing is an acceleration of this transition, silicon to GaN, to achieve that density. Got you. You mentioned you have GaN, you have silicon carbide. Yes. What is the overlap? I remember during the IPO, at that point, you were only GaN. You were focused on the charging cubes for cell phones. You also had in development, I remember seeing the data center power supplies, but it was a little bit of a ways off. You acquired silicon carbide, which I believe a lot of people just view as, well, that's the much higher voltage, maybe lower frequency. Can you just help us understand, when do you use GaN, when do you use silicon carbide? Are both of these opportunities that go into the data center for you? The way I view it is, to drive power efficiency, you have to reduce number of conversions from all the way from the grid down to the GPU. Okay? More conversions mean more power loss. If you want to increase the efficiency to drive density, you have to reduce the number of conversions, right? SiC and GaN are very complementary technologies. I'll walk you through the, let's say, the grid all the way down to the rack and to the GPU, right? I think that would be very helpful. The GeneSiC technology, as I said, which we acquired, came from this high-reliability, kind of over-engineered SiC technology that operates very well at high voltage. If you think about the 2028 and beyond the grid transformation, which I talked about, right? Basically moving from old-school transformers to solid-state transformers, battery energy system, UPS, name it. This needs super high voltage SiC technology in modules as well as in discretes. That basically is a sweet spot for those type of technologies that are operating very well at super high voltage. Very different from the EV SiC that we talk about in the automotive EV industry, right? 1,200 V, 1,700 V, 2,000 V, 3,000 V, and above. Very sweet spot for GeneSiC technology because it's high reliability, high density, and high efficiency. You get down to the AC/DC for data center. When you convert mid-voltage AC, 480 V to 50 V today, but tomorrow 800 V, right? That's a very good sweet spot for SiC technology as well at 1,200 V. Okay? As we move up, okay, with the power level of the data center, we also move up in the power level of the PSUs. With AC/DC PSUs are moving from five, eight, 12 kW per PSUs, to 18.5 kW, 27 kW, in the future, 50 kW per PSUs. If you do that, you need to move from your typical one-phase architecture to three-phase, again, converting 480 V AC to 800 V DC. That increase, okay, of density drives a lot more usage of Silicon Carbide, okay? My rule of thumb is when power goes up by two, content goes up by five. That's another, okay, benefit of having a SiC technology. You get to the DC/DC space. DC/DC top of rack, okay, where we're going to see that next year, where you move the 800 V to 50 V, okay? 50 V all the way down to the GPU. You can do that with those DC/DC. You can use GaN, you can use SiC. Some people use SiC on the primary side and GaN on the secondary side, or even sometimes silicon. As you move into inside the rack, so I'm talking about real high density, high power racks, what we call the native 800 V right? Where you get this power distribution, DC inside the tray. Here, you have no choice than to use GaN. Okay? The primary side, the step function in content comes from that with the GaN. All the way, GaN will continue to gain traction all the way down. Why do you have to use GaN at that point? It's switching at much higher frequency. Because of the frequency. Because of frequency switching. Okay. I think you mentioned it between SiC and GaN, right? That's the big difference, okay? I think GaN switch. If you think about the board we announced, we announced two board actually, reference board this year, in partnership with hyperscalers, NVIDIA, kind of showcased that board at the last GTC and also this week at Computex, which is 800 V to 6 V, right? We use GaN on the primary side. Okay, so 650 V GaN on the primary side with silicon on the secondary side because it's 6 V. Down the road, there are low voltage GaN companies. There are low voltage GaN technology that will emerge, I think we'll use GaN on the secondary side. If you look at the 800 V, 50 V that we announced earlier this year, this is GaN on the primary side and GaN on the secondary side. That gets to 1 MHz of switching frequency and 98.5% efficiency. Okay? The level of efficiency per density that you get using GaN, okay, is very much better than silicon for sure. A lot of investors, historically, so this is all getting very complicated. Yes. Right. A lot of investors will say, "Well, what is this company's TAM per rack?" Now you have a compute rack, right? You have a BBU, a PSU, a Solid-State Transformer. You have four different modules. How can you help us with a framework to think about what is your TAM in each one of these? Actually, it's even more complex than that. Okay. Because you have the- I was afraid you were going to say that. Yeah. Okay. It's even more. Let's start with data centers, okay? Yeah. Let's start with the grid, okay? You have the AC, very high voltage grid, okay, to DC. You can do that with transformers moving to Solid-State Transformers. Then, in the interim, before we get to SST, you do that with the 480 V to 800 V DC, so AC/DC PSUs. Inside data centers, you're going to the DC/DC first, and then all the way down, further down to the GPU. It's a lot more complex than that. There is no way to quantify the content per rack. The best way is to quantify the content per megawatt, okay? I've seen more and more competition, okay, more and more industry players are basically referring to content per megawatt. Which I think is the best way to think about this, right? As we scale up and talk about 220 GW being deployed in the next few years, up to 300 GW, that gives you a reference point, right? The content we came up with is the GaN, as it gets into native use of GaN inside the rack, we're talking about $10,000-$15,000 per megawatt of GaN content. That's clearly coming from replacement of silicon on the primary side. This will increase over time as GaN gain content on the secondary side over time as you get closer to GPU. On the SiC side, there's two play, inside data centers and outside of data center. Inside data center is basically the AC/DC. Okay. As I mentioned, the AC/DC PSUs moving from one phase to three phase, going to 10 kW to 20 kW to 30 kW to maybe in the future 50 kW. That's about $8,000-$12,000 per megawatt, okay, of SiC content. Mostly at 1,200 V. You have the SST, okay, which is really the modernization of the grid. Basically, the grid moving from the transformer giving you a steady 800 V DC from the 35,000 V AC, right? That's up to $20,000 of content, okay, in the SiC, right? That's really 2,000 V, 3,000 V, and above. You can see the content overall is growing in both SiC and GaN, and that's why we quantified the TAM, or the SAM in the future, as kind of 50/50 between the SiC and the GaN because both will get traction in this grid to rack or grid to GPU. Mm-hmm. If you're moving to 800 V, then it seems like the growth is coming more on the higher voltage side. Definitely. Is that fair? Definitely. I think first of all, the sweet spot for us on the SiC side, okay, is definitely above 1,200 V. Okay? 1,200 V, 1,700 V, 2,300 V, 3,300 V, and above. GaN side, 650 V and above, okay, and moving to mid-voltage GaN with the release of the 100 V GaN that we announced earlier this year, right? You're right. Basically, the trend is higher voltage, higher power, which is a good sweet spot for both technologies, right? Got you. Okay. You talk about the 120 MW or 220 MW that are going to get built out, and you're throwing around some big numbers here. How do you think about the supply side for you? Can you just talk about your supply chain a little bit and what you do to try to make sure that you're going to be there to supply the product to this rapidly growing market? Okay. First of all, I think the answer is different depending if you're on the SiC side than when you're on the GaN side. On GaN, we've announced in November that we've started a strategic partnership with GlobalFoundries. Moving from TSMC, that is working away from GaN to Global. We started that partnership for multiple years. It's kind of more technology partnership than a foundry partnership, where both team got together to accelerate the deployment and the ramp-up of GaN in a U.S. fab. Okay? The fab is in Vermont, in the United States. The team are very active. We're going to be sampling to customers at the end of the year. We're going to be in production next year. With the support of TSMC, we're kind of building buffer, okay, before TSMC will phase out their GaN process, which allows us to give customer a smooth transition, as well as build buffer to accelerate to support that ramp. Of course, we're working closely with GF team and the team to make sure that the capacity is in place. We'll use capital as well on our side to participate with the investment that the foundry have to make. Same on the X-FAB side, which is our partner also in the U.S. for the SiC side, where we've ramped nicely already. We see a very rapid growth of the silicon carbide already, mostly coming from the grid, but also from the AC/DC, which I mentioned, and we're working with them to prepare the next leg of growth. We'll participate to the capital investment for foundry. We'll stay a fabless company. We believe that we can grow, in the next few years, in that model by participating in the investment with our foundry partners. You said X-FAB. X-FAB. In the U.S. or? Yes. Okay. Yes, it's a fab in Texas. Yeah. Got you. Okay. What are your main priorities here? How do you prosecute these rapidly growing markets? I imagine that's really what you're focused on. Sounds like you have the operations, the supply side nailed down. When I came on board, and it was August, September last year, feels like a few years ago. When I came on board, the first thing we did as a management team is most of the revenue of the company was still in mobile. We talked about this huge opportunity that we had, with SiC and GaN going to a data center grid and high-performance compute. The vast majority of the revenue, and as such, the vast majority of the headcount was still deployed to mobile. The first decision we made is to walk away from that. Big pivot, okay? Redeploying all the resources, pivoting the center of gravity of the company where the customers are back to the U.S., okay? Pivoting the roadmap, technology investment, deployment, go to market, and all resources away from mobile. Walking away from mobile revenue was not just a revenue decision. It was basically an output of manpower and resource impact decision, right? I wanted, as soon as possible, the entire company to be focused on the high-power markets. I think by now we've achieved that. There's not a single soul in the company that's spending any time on mobile and consumer. We support the customers from an order point of view, but when I came on board, it was vast majority of the company. Last quarter was less than 20%, and we said to the investors that by the end of this year, it's going to be less than 10%. Essentially, the entire growth of the company now comes from high power. That was the big pivot, right? Pivoting away from this historical market to deploy this new technology, both GaN and SiC into the AC/DC PSUs, the DC/DCs, serving the grid, the BESS first, with the battery energy system, as well as the SSTs in the future. Also the other application, we'll probably talk about it, but I think people underestimate the importance as well of high-performance compute. As AI goes to the edge, those high-performance computers need very high-power chargers. I'm not talking about 100 W. I'm talking about multiple 100 W, 200 W, 300 W, up to 500 W. That's also a good use case for use of GaN. Got you. We're coming a little short on time here. You've been there for nine months, nine or 10 months or so now. You've spoken, you've taken the time to speak with investors, right? Where do you think is the biggest disconnect between your understanding of Navitas fundamentals and where the investment community is? I think there are two things I would say, not a disconnect, but something that we have to spend time on talking to investors. Number one is, it's a big market. People say, "Hey, you're a small company. How come that a small company would sit at a big table?" I don't think customers, when we were at MGX last week, okay, and at Computex, I don't think customers are having a seat at the table for charity, okay? They recognize that it's one thing to be a big company, but it's another thing to have the right technology. We benefit from, as I mentioned, the heritage of Navitas 1.0. We've been in GaN for a while, okay? We know GaN. We know how to take SiC to a different level. Yes, we have not been in, from a revenue standpoint, in scale, but we've mastered the technology. What's happening right now is customers are trying to move so fast, and the AI acceleration is so fast, even the grid reaction is so fast, that you don't have time to learn, okay? You have to basically master that technology. A lot of the discussion I have with this is say, how do we basically come to the table because we've done that before, okay? We can basically help the customers to do that transition. Customer needs reliable, existing solution, not things that we're going to work on for the future. That's number one. The other disconnect is everybody talks about data center, okay? My days are filled with talking to customers in the data center space, either the ODM and the OEM in Taiwan or the rest of the world, as well as the hyperscalers. One thing that people probably underestimate is AI is a catalyst that changed data center, but also changing the grid. I talked about the SST, right, the solid-state transformers. I think by 2030, when we deployed 200 GW, 300 GW of AI data center, we've only changed a small portion of the grid. The grid, there's a tipping point, okay, where if you replace old transformers with solid-state transformers, you're going to do that for the entire grid, not just for the use case of AI. I think people underestimate going five years out, 10 years out, 20 years out, how much potential the grid for high voltage in particular in United States is a huge market. I think this is where people talk about it, but they think it's far out, I think it's coming very soon. I think the SAM is enormous when it comes to SiC ultra-high voltage in the grid play. I think you are the only one that I have heard this from. Sounds like a very exciting opportunity. Chris, we've run out of time. Thank you. Thank you very much for joining us today and sharing your insights on Navitas. Thank you, Mark. Thank you.