Okay, let's get started. Good morning. Welcome to Baird's 2025 Global Consumer Technology and Service Conference. I'm Tristan Gerra , your Semiconductor analyst at Baird. I would like to introduce Navitas, a leading supplier of next-generation power solutions. We're honored to have with us today Gene Sheridan, Co-founder and Chief Executive Officer. With that, let's get started. I think, Gene, you have a few slides for us.
Sounds great. Yeah, thanks, Tristan. Glad to be here. And glad to give you guys an introduction to Navitas. If we can kick off the slides, or do I give a little push? Sounds good. My green button. Not yet. Oh, here we go. Tell me to bounce back. Okay, great. So Navitas, I started Navitas with my co-founders 11 years ago. I've spent my whole industry in Semiconductors, and primarily in Power Semiconductors. For 30 years, we developed Power Semiconductors using silicon. In the last 2 decades, 2 new materials have arrived that show not only great promise, but significant commercial traction to replace silicon in power semiconductors. Power Semiconductors are really the chips that handle power, volts, and amps. All forms of electronics need power supplies to power those electronics, and they need power chips, power semiconductors, to operate those power supplies.
gallium nitride and silicon carbide, while totally different materials, have very similar properties that are dramatically better than silicon. The name of the game in power supplies is mainly energy efficiency. How efficiently can we move that power from one form or place to another form or place? Usually power density, which is how much power can we pack in a small size, lightweight, and ideally low cost. Power density is usually driven by switching speed. Most of these power applications are switching power supplies. The faster you switch them, the more power you can deliver in smaller size, lighter weight, and ultimately lower cost.
Gallium Nitride, combining these two different materials together, Gallium and Nitrogen, or GaN, and silicon carbide, combining silicon and carbon to create silicon carbide, both possess a strong ability to handle high voltage and high current with small, tiny chips that are super efficient and very fast switching. I'll keep coming back to those same 2 themes: fast and efficient. Silicon carbide actually took off first. It's already a $3 billion-$4 billion market. Gallium nitride is kind of a new kid on the block, just really got commercialized in the last decade and commercialized by Navitas. We were the first company to come along and drive commercial adoption in one mainstream market, which was mobile chargers. We'll talk more about that later. Now we're moving into data centers, electric vehicles, and solar inverters. We acquired a silicon carbide company a few years ago.
Now we're actually the only company on the planet that has gallium nitride, silicon carbide, without the distraction or dilution of silicon-based power systems, like a lot of my competitors are selling today. Big spotlight today. Everybody's talking about AI. It's a big deal, not just for AI, but for the power electronics world. You've probably heard about it. Not only is it massive processing power, these are massive chips requiring massive power electronics, power semiconductors. Silicon is quickly running out of steam to deliver the power and energy efficiency that AI processors need. This is a perfect fit for silicon carbide and gallium nitride. Kind of perfect time, perfect place to really solve a big power problem.
When NVIDIA showed up with Hopper, Blackwell, and Rubin, the series of 3, we're literally going from CPUs of the past, which would be 300 watts, would be a lot of power for the processor. Blackwell is already 1,000 watts. Rubin is going to be 2,000-3,000 watts. These are crazy, crazy numbers to deliver to a little tiny chip on top of the memory chips. Then you put dozens of those processors inside the server rack. You're going from a server rack is like the building block of a data center. Used to be 10,000 watts or 20,000 watts for the entire server rack, which is like 50 shelves and lots of processors. That would be a crazy amount of power. Today, we're already designing 100 kilowatt server racks to put in all those processors.
Now we're trying to go to 0.5 megawatt or even 1 megawatt processor. Crazy, crazy numbers we never imagined a few years ago. That is a crazy good challenge for my company to technically use silicon carbide and gallium nitride to solve that problem. We just launched our GaN chips for this market late last year. We have started ramping. We're in early revenue days. The numbers are not big yet, but a lot of this is brewing for 2026 and 2027. We'll talk more about that. I sell my chips to the power supply companies. The power supply companies then resell them to the big data center companies, the names that I show here. We had 40 design wins last year, which is a really good start, 70 total customer projects that we're trying to ramp.
Just last week, we were announced in an NVIDIA announcement about the collaborations that they've created in a next-generation data center that takes that server rack to the megawatt level and beyond for really late 2026 and 2027 and beyond. It's the fastest growing part of my pipeline, even though we're doing really well in other segments, which we'll talk about. A little bit more of a deep dive here because it's such late-breaking news and it's so exciting for our company. Traditional data centers shown at the top, those were powered with silicon. We call them 12-volt data centers because that was the primary bus voltage inside the data center, relatively low voltage to transfer the power around to those processors. If there's one technical thing you want to know about power electronics, it's Ohm's law. Power is equal to current times voltage.
If you want to deliver a lot of power, you want to deliver it with the lowest amount of current possible. Because when you try to bus high current around through cabling and over PCBs, you get a lot of power distribution losses. It is very inefficient to deliver a lot of power with a relatively low voltage because voltage times current equals power. If you want a big power and you have a little voltage, you have to distribute a lot of current. That is not good. 12 volts was fine as a low voltage when we were only doing a few hundred watts per processor or maybe 10 kilowatts in the server rack, as I show there. All of a sudden, here comes AI. Now it is going to 1,000 watts per processor, and it is going to 100 kilowatts per server rack.
We had already been working in 48-volt data centers. We all quickly moved to 48 volt. That's a 4 times higher voltage. What's cool about that, which means the current can be four times less for the same amount of power. And guess what? The distribution losses are proportional to the square of current. That's the second technical thing to note today. Instead of being four times less current, it's 16 times less power distribution losses by going from 12 volts to 48 volts. That's a great step. We didn't invent this. We're following the trend. We're supporting the trend. It just means you need higher voltage power chips, which plays into the strength of gallium nitride and silicon carbide because they're really good as you go up in voltage. Now, last week, NVIDIA announced the future of data centers is 800 volts.
Imagine 48 volts going up almost 20x or 15x. That's 15x squared in terms of reducing the power distribution losses or making a lot of power get delivered much more energy efficiently. It also means more high voltage power chips, more gallium nitride, more silicon carbide. What I love about this, too, is they're even hooking up farther into the grid. The grid's AC power starts at super high voltages. We have to step it all the way down to when you plug in the wall here. Grid power, when you plug in here, is 110 volt or 220 volt, depending upon Europe or U.S. We had to take that AC all the way down.
If you want to do an 800-volt data center, why would you take it all the way down to 110, 220 just to then boost it back up to 800 to power your data center? That would be very inefficient. Instead, NVIDIA and everybody else is going to hook up into the grid, starting at 13.8 kilovolts, step that down to 800 volts, take that right into the data center. We get a really efficient, super powerful megawatt data center. Guess what? You need super high voltage silicon carbide chips to connect to that very high voltage inside the grid, something that silicon carbide is good at. In particular, Navitas's silicon carbide is the best at 3 kilovolt, 5 kilovolt, 6 kilovolt. We love that this means a lot of power chips. It means a lot of gallium nitride and silicon carbide.
It plays into Navitas's unique strength that our silicon carbide is very good at these super high voltages. All of this is rolling out over the next two years. A lot of work to do, a lot of R&D going on. It is not just Navitas. There is a whole ecosystem working on this move to data centers. Just a quick financial snapshot. We launched the first GaN chips in 2018. As I mentioned, that was the first mainstream adoption of GaN chips. We started with mobile chargers. I am talking a lot about data centers today as the next big thing. Mobile chargers was our first big thing. Our GaN chips are literally used in the top 10 or even top 20 out of top 20 mobile players in the world. We started with aftermarket guys like Anker, Belkin.
As Amazon then quickly moved to every mobile player, Apple, Samsung, Dell, HP, Lenovo, Xiaomi, Vivo, you name it, they're using our GaN chips in mobile chargers. We're not the only one, but we're a leader in that space. That revenue ramped really quickly. We saw a slowdown in the last two years. Last year and this year, the silicon carbide market slowed down because of inventory corrections and a slowdown in solar, industrial, and EV. We felt that slowdown last year. We're feeling it this year, even though gallium nitride last year was at an all-time high with those mobile chargers still ramping, 50% growth last year. We see our growth really kicking into high gear at the end of this year and ramping into next year. As mobile grows, gallium nitride is going into solar for the first time.
Gallium nitride is going into EV for the first time in the next few quarters. Most importantly, ramping that data center, gallium nitride and silicon carbide for data centers, especially with the 800-volt data centers coming in late 2026 and into 2027. That is a super fast overview and tutorial on power electronics, power semiconductors, gallium nitride, and silicon carbide.
Great. Thanks for the introduction. Maybe just a first high-level question, Gene. When you started the company years ago, what did you see that, I guess, some of the traditional analog companies did not see because they are just starting to do a little bit of what you guys have done for several years, and obviously, they are really behind. What did you see and what technologies did you think you have that ultimately gave you that market advantage? I think a few years ago, you talked about having about a two-year leeway relative to competitors in gallium nitride. When you started in silicon carbide, we are getting industry feedback you had the best solution. What did you see at the beginning? Maybe just a quick retrospective on what you have achieved so far as far as your early vision for the company.
Yeah. Our founders, we actually have worked together in power semiconductors for much of the last three decades in our career, starting with International Rectifier, who was really a lead inventor of silicon-based power transistors. When I talk about power semiconductors, we're often talking about the power transistor, the thing that can turn on and off those volts and amps and deliver a lot of power. Gallium nitride, we started the first gallium nitride program at International Rectifier over 20 years ago, but we knew it was very early days. This is new material. You have to work on manufacturing, reliability, cost structure, of course, getting the performance better. It takes a long time. When we started Navitas, we felt like the time was right for the remaining challenges of gallium nitride commercialization to be solved and really start ramping that technology. That was 11 years ago.
The biggest remaining challenge was a GaN is very fast transistor, but super hard to drive it and control it. It's like having a Ferrari engine, but not having the suspension and braking system around it to handle and harness that incredible power and speed. We knew when we started, we had to solve that problem. Many people were trying to do custom silicon drivers outside the GaN chip to try to harness and control and drive that really fast transistor. We, in a nutshell, figured out how to integrate the driver and control circuitry directly into GaN, which even to this day is a mystery to most of our competitors. Very hard to do. Ultimately led to over 200 patents being filed or issued in how to integrate drive into that GaN chip. Once we solved that remaining problem, the rest of GaN was pretty well solved.
Reliability, manufacturability, yield. I mean, there was work to be done, but that really created generation one of mainstream adoption of GaN. We targeted it on mobile chargers, quite simply because there is a simple value prop. We pack a lot of power in a small size in a market that everybody can relate to. Who does not want a fast charger that slides in your pocket to charge your laptop, your tablet, and your phone? Super lightweight, lots of power, multiple outputs if you want it. It adopts fast. It was not looking for a decade of field reliability that you would need for some of these more conservative industries. We targeted on mobile chargers. The rest is sort of history. You can see it ramped up really fast.
We're moving to that next wave of growth as gallium nitride goes into all new markets like data center, solar, and EV.
Great progress so far. Good to hear that you've had this 30-year experience in the field at a time when companies are trying to figure this out. Going back to the NVIDIA announcement, can you give us, and I know that it's an engagement at this point, it's not exclusive. How should we look at this in terms of timing, in terms of potential content? Of course, we're getting questions about potentially what does this mean at the revenue level. There might be ways to maybe tie this to a content per kilowatt and timing, potential market share, anything that you can give us that helps quantify that opportunity as it starts ramping.
Yeah. First, to clarify, we have very good opportunities here in this middle section. This is happening today, 48 volts. We really have high voltage gallium nitride, 650 volt gallium nitride, and silicon carbide fits right here in this box. We just started sampling earlier this year the low voltage GaN. We are just opening the doors to go into this second power conversion stage. It is a 48 volt DC to DC converter. This is already going to happen. We are starting to ship these this year. This will grow next year. This is sampling this year and will start shipping next year. This is great. This was already a great opportunity.
When you go down here, it's just the opportunity explodes further into using our super high voltage we call UHV, ultra high voltage silicon carbide, when we connect up to that grid at that really high voltage. That's brand new. I'd say first commercial ramp would be a little bit late 2026, but mostly 2027 where we ship that. Between now and then, we've got a lot of development work to improve the products and hit the specs and deliver the samples and do the system development work with a lot of the partners. At the same time, the high voltage will continue here. As I said, we're expanding with the 100 volt, which will go and start ramping in 2026 and 2027.
It's really an expansion going from 48 volt shipping this year, 48 volt shipping next year, I mean, in this rung, and then expanding rapidly late 2026 and mostly into 2027 and 2028, where the full 800 volt data center rollout will happen.
Great. Typically, what we've seen is NVIDIA innovates, and there are new solutions being implemented. Ultimately, it expands into other AI platforms and then eventually in traditional data center platforms. It really drastically expands the unit opportunity. Do you see a potential for that? Do you see an interest from hyperscalers or people talking to the industry? I mean, you've talked about a number of engagements, and the number of those engagements has increased as of last quarter. Maybe if you could give us a recap of your other engagement, what it is for, what's been the progression, and the opportunity that you see comparing those engagements with what you're doing with NVIDIA today.
Within the data center world or.
Within data centers.
Yeah. That's a good point because everybody talks about NVIDIA, but data center and AI is not just about NVIDIA. The move to 800 volt data center has been talked about and early work in academia and other places for quite some time. That also is not unique. I think all the hyperscalers are going to move to these really high voltages. Some will call it plus or minus 400 volt or 800 volt like NVIDIA is describing. I think the trend is clear. The opportunity is much bigger than NVIDIA themselves. Frankly, it's bigger than even data centers. AI starts in the cloud because that's where a lot of this really intensive processing, learning, and inference will occur. It's going to go from cloud to the edge to the client. There's just early talk.
There is so much more to happen as AI goes into your car for true complete self-driving and learning goes into all of your mobile devices, goes into medical devices, goes into robotics. I mean, AI is really just at the very beginning, starting in the cloud. Wherever AI goes, not only goes massive processing, but massive processing power, which is a great opportunity for, again, gallium nitride and silicon carbide to follow those chips. There is a lot to do. Right now, I think data center will keep us very busy for the next two years, NVIDIA and far beyond that, Google, Facebook, Amazon, AWS, Intel, AMD, all of them, of course, are moving in this direction. All of them will need more powerful chips for the future data centers.
Do you see opportunities to expand your content further from what you're showing us in this slide here at the module level or maybe some regulation chips or anything that could be an opportunity later on to even increase that one-stop-shop solution?
Yeah. It's a good, it's actually a really good point. I spent a lot of time talking about gallium nitride and silicon carbide. These are the powerful power chips that are kind of the heart and the muscle, if you will, of these power systems. We also do silicon controllers, which are like the brains. When you switch to gallium nitride and silicon carbide, you want to switch a lot faster, as we talked about at the beginning. A lot of things in that power system have to change. We've gotten very good at the brains, which need to think faster, not like AI GPUs, but the brains of a power supply called a silicon, usually done in silicon, not in gallium nitride or silicon carbide. We're doing silicon controllers.
We're doing isolated drivers, is another thing that we do in silicon that you can't yet today do in gallium nitride and silicon carbide. There are adjacent chips, complementary chips that are very important to these systems that we're also doing to enable the full system. It also increases our content. We sell it as more of a chip set. As you also implied, Tristan, there's the chance for some of these to be, and I'm sitting right in front of the picture for everyone to see, some of these could become modules themselves. Each of these blocks are a power system. They could be designed as a power module, which has the gallium nitride, the controller, the isolators, the magnetics, which we're good at designing high-frequency magnetics. There's potential even for Navitas to move into power modules that would increase our content dramatically.
We're not making any announcements yet on that plan, but certainly that potential is there. MPS is a great example of a company that's a silicon controller company. Doesn't do the gallium nitride and silicon carbide today, but they've transitioned from selling the chips to also selling power modules in this traditional 12 volt and 48 volt market. It is one example of somebody, a great financial success that's operating at the module level. That could be a path for Navitas over time.
One last question on my side, at least about data center. Where do you think the highest moat, competitive moat, and IP is? I mean, obviously, you're ahead of the competition with what you're doing here. The module and the silicon controller, what is the IP behind it, and what type of value are you adding eventually if you were to move in that segment as well?
Yeah. I think the gallium nitride, the fundamental advantage we bring comes back to even the early beginning of the company. It's that creating not just GaN transistors, great, fast, and efficient transistors, but GaN integrated circuits. Not only are we now integrating the driver, like I said at the very beginning, which is what really helped to catapult not only the company, but the industry to bring GaN into its first mainstream markets. Now we've gone multiple, we're on our fourth generation, going to fifth generation, which includes higher integration, integrating drive, control functions, sensing functions, protection circuits. The latest GaN technology that we use here, generation four, is called GaN Safe for high reliability applications, including data centers, because it integrates not only drive and sensing functions, but protection functions to make the most safe, protected, and reliable GaN IC.
Of course, data centers and applications like this are expected to last 10 to 20 years. That GaNSafe technology, that GaN integrated technology is not only getting us the best performance, high integration for a small chip size within the system, but also the highest reliability. In a nutshell, for GaN, we continue to have the highest integrated GaN IC, which gives us performance, reliability, and even system cost benefits. Over 200 patents protecting that GaN IC position. In the world of silicon carbide, Navitas and others are all discretes. It is a vertical structure, so it is not easy to create the integration that we did in GaN, but we have set the industry benchmark for the highest voltage and the highest reliability. The highest voltage, of course, plays perfectly into hooking up into this very high voltage grid, 13.8 kV.
Now our silicon carbide chips are the highest voltage today in the market at 6.5 kV. That's not 13.8, but you take 6.5 and you stack them, and you can manage that high voltage, stepping it all the way down to 800 volts. Compare that to most silicon carbide competitors in the world. They have 1.2 kV, maybe 1.7 kV. We're doing 2 kV, 3 kV, 4 kV, up to 6 kV. Ultra high voltage, ultra high reliability is our silicon carbide advantage. Ultra high integration, driving performance, size, and reliability benefits. And GaN is our big GaN advantage.
Great.
Can you talk about competition? There's a lot of providers here that are in the 48-volt area right now that have GaN capability that, quite frankly, no one will have made 100. So who do you fear the most as it relates to scale and technology?
It's interesting. Kind of the biggest, the leader in power semiconductors overall when you include silicon is Infineon. In fact, Infineon bought International Rectifier, which is where I and many of the founders started. I would point to Infineon as the biggest competitor. They have the silicon, so they do a lot in the 12 volt, and then they have low voltage and high voltage GaN, and they have the silicon carbide. That could be the most feared competitor, if you will. Here's an interesting twist. We signed a broad cross-licensing agreement for gallium nitride and a covenant not to sue for the rest of our patent portfolio, which includes silicon carbide, with Infineon. At first glance, you may say, that's a little strange. Why would you do it?
What we're finding is most really high volume applications really want to go to gallium nitride and silicon carbide, but they really do not want to do it with a sole source solution, especially an early stage company like Navitas, inherent risk profile. We find Infineon to be actually one of the closest in terms of technology capability and broad range, as I described, to address this market. We are turning a competitive threat into, I think, a collaborative opportunity. It was, frankly, one of the reasons why NVIDIA announced Infineon and Navitas as part of this collaboration, because they love the dual sourcing nature. Just for the reason you described, the 48 volt GaN, they want to go big with a lot of GaN inside that box. They do not want to do it with a sole source approach. Navitas and Infineon have this cross-licensing.
We're proactively setting up common footprints, common specs to make dual sourcing easy for our customers.
Do you see Monolithic Power and MACOM?
MACOM can be confusing. MACOM is a leader in gallium nitride for RF, but gallium nitride for power is, although the materials are the same, the application knowledge, the device structures, the technologist skill set is actually quite different. We do not see, and we will not be going into RF, just like the RF guys, even Qorvo looked at going again, or silicon carbide in the power field, ultimately sold that business off. I think RF and power do not always mix too well, so we will keep that separate. MPS is a good one. MPS does not have gallium nitride or silicon carbide in-house, so they actually partner with people because they are making the modules, and the modules need their controllers, and MPS has great silicon controllers, but then they want to have the gallium nitride inside that module. That could actually be another partnership opportunity for Navitas.
When you say you sell to the power supply companies and then sell to data centers, how are we moving from the conversations with the hyperscalers and how much kind of didn't they?
Yeah, this is a great question, and it's a common ecosystem situation where typically, even in the charger space, we're selling to the guys who design and manufacture chargers or adapters, names you wouldn't tend to know, but those are the suppliers to the Apples and HPs and Dells and Lenovos of the world. It is very common for us to form the relationship with the major OEMs or the end system integrators because we need to know what's the future coming. What power level does Google want to drive their new phones or Apple want or whatever? What are those power requirements coming? Then make sure we work with their suppliers that our chips and our system, our applications engineers can enable it. We're following that same model with NVIDIA. Right now, we don't sell anything directly to NVIDIA at this time.
That could change as they move in this direction, but we talk to them because they're teaching us what's the power requirements for the roadmap so we can get ready and work with their suppliers. It is always a three-way relationship in all of our markets and data centers the same.
Anything more of those integrators?
The power supply guys are, again, companies you might not know, but they tend to be in Asia: Lite-On, Delta, Akbel, Compal, Chicony, Flex has a strong power capability, Advanced Energy. Those are mostly our companies that are our customers that are buying the chips, making the power supplies, and then reselling to the hyperscalers, AWS, etc.
A few figures of merit. I think GaN is the better material. So if you can continue to bring it to higher voltages, do you need the SiC business in the long run?
Yes, because what's happening is as much as we're going to push GaN to higher voltages, it's a lateral structure. You're definitely hitting limits. As you try to handle high voltages, lateral means the current and voltage is being blocked along the surface of the chip, the wafer, which is how almost all silicon semiconductors are done. If you really want high power and high voltage, we tend to flip that around and make it a vertical structure where you're blocking the voltage across from the top to the bottom. That gives you a lot more dimension to work with to block that voltage. Ultimately, GaN will never be as high voltage as silicon carbide.
Even as GaN moves to high voltages, like maybe you can get to 1,000 volts or 1.5 kilovolts, silicon carbide is popular today for 1,000 volts or 1.5, but look where it's going. We're going to 3 kilowatt, 4 kilowatt, 5 kilowatt, 6 kilowatt. Why not 10 kilowatt? Why not 15 kilowatt? We need to upgrade the entire world's grid. By the way, this is not just how you power data centers. It's how you hook up solar farms, wind farms, energy storage. The entire world's grid needs to get upgraded and solid-state transformers. Replacing that big magnetic transformer has no power semiconductors in it. That's what the grid is today. Over time, it will all go, I wrote down SST because that's solid-state transformer. That just means semiconductor-based transformer. That's going to be silicon carbide. That's the best way to handle it.
I think as GaN goes up in voltage, so does silicon carbide, and silicon carbide will always be better at really high voltages because of its vertical nature rather than lateral nature.
Thanks. I think we're running out of time, but I invite you to Gene's available for one-on-ones, or please reach us and we can certainly set up calls to discuss further the company.