Third Bridge 专家访谈讨论 Power Integrations 在GaN与高压电源转换中的竞争定位

[[00:01:45]]
How would you describe the competitive landscape for Power Integrations? How is the company positioned within the landscape, especially when it comes to AC/DC [alternating/direct current] power conversion?

[[00:02:13]]
I will speak mostly for the business I was in and for the time I was there. What's happening right now, I'm not too much aligned. Also, as you know, there has been a leadership change there. Competitive positioning, Power Integrations, they are a specialist in high-voltage switches. That has been their DNA for more than 30 years. They are doing very well. They are building very well-integrated solution, and that has been also part of their DNA trying to consolidate BOM, bill of material, into one system-on-chip. That allowed them to cut the component count, but also to increase the efficiency because they have much better ability to control things internally through a combination of hardware and software, and that was very important, adding some digital domain to the analogue and power. Competition, at least, again, the time I was there, the most significant ones, so I can cluster them in two fronts. One is price leadership. Normally, these are from China or Asia, Silan, Chipown. The other ones, of course, the global ones like ST, Infineon, and Navitas. This would be the key competitors we had or I had at that time.

When it comes to positioning, we tried to add more of the system know-how into our offering. Not only sell an end-to-end solution, software and hardware. I will explain why. What exactly was it. While competitors normally would sell a chip high-voltage drive IC, we would try to combine the same hardware plus software and plus the control, which normally was not from Power Integrations, a microcontroller processor so that we can help customers design, not only be part… We are trying not only to be a hardware component selection, but to be part of the design itself. That was really a differentiator for our motor drive market. The disruption, the price disruptors like Silan, they didn't really care about the software. There was just a pure price play. They have gained a lot of share, as you may know, in particularly in China. Yes, in general, you don't want to fight with those guys over price. You still want to go and sell your end-to-end solution, adding value in the system design with software company.

[[00:06:30]]
You mentioned the differentiating factors for Power Integrations vs players such as STM, Infineon or Navitas. Is that a differentiation where competitors are closing the gap? Are they two different strategies that each of these players has taken?

[[00:07:02]]
Good point. If you look historically, ST and Infineon are much more integrated than Power Integrations on the bill of material on the board. They have digital components and software and experience for years, which we lacked at that time. Now the solution play was critical for us in order to be on the similar eye level as those guys. However, when component decision is done on the customer, we learned that customers prefer to pick a player or a supplier, which has more of their bill of material to reduce their supply chain complexity. The other challenge we had was to get into designs where the incumbent was one of these guys I just mentioned, ST, Infineon, and so on because we are new in the market. It was hard to get in. In Asia, it's been, again, an uphill battle because of the price. To answer your question, I think we tried to level up the game vs to the global players like ST, Infineon because they had a significant head start. Secondly, pick the games carefully when it comes to price disruptors because it was very hard to achieve the profitability level we wanted.

[[00:09:12]]
Power Integrations suggests that it currently addresses AC/DC applications with power outputs up to approximately 500W. As GaN [gallium nitride] enables higher-power applications, how much larger can the company's addressable market become over the next few years? Is there huge potential here? Is there something that the company has to do to address that? Does it have the appropriate portfolio as it is?

[[00:09:50]]
Again, I don't know when it comes to new GaN technologies, how much they can go in terms of power, but I'm speaking from my perspective and based on how I see the market. Let's be specific. Let's look at HVAC. HVAC is around 120 million-130 million split units. A split unit is composed of internal and external unit. In terms of, let's say, content for high-voltage drivers, you will have at least two fans and the compressor. Fans are about 100W-150W. Internal unit fan doesn't care about efficiency. External unit fan, its higher power cares about efficiency. This is a potential target. Compressor, this is like 1.2kW, 1.5kW, 1.8kW, they care a lot about efficiency. This would be to me, and now when we talk about volume, so let's not consider the internal fan, which is 30W, let's look only at the external fan and compressors. Let's say, 120 million times two. You have two designs here. It's 240 million potential targets. Then if you look at the pricing, I think a compressor driver is much more expensive than a fan 100W-120W fan. Now again, I don't know what their plans are. I don't know what's happening right now. For GaN, these are two target markets, this is a mature market. We are not talking about a completely new market where efficiency matters, where integration matters and that, I think, GaN players are really looking into.

[[00:12:23]]
As you said, it's a mature market. On that note, is there still quite a lot to go after, despite it being a mature market? I'm thinking in terms of the growth prospects for a player such as Power Integrations.

[[00:12:40]]
Home appliances in general, it's of much interest. We talk about in total. If we look at washing machine and HVAC, and fridge. Let's say, let's first put that in the perspective. HVAC and washing machines are the ones in the highest need for higher efficiency, where GaN can bring faster value. This is maybe a 180 million-190 million unit cluster of markets. The third one, which is coming, I think, relatively strong is fridge, refrigerators. This is a huge market. It's by itself more than 150 million. The challenge here is power is not so high. Fridge yet has a very low or small penetration of advanced drives like the ones we are talking today. Only 20-25% of the fridge designs do use advanced drive techniques, which would require solutions like we are talking today. The other side, 70-75%, still use old and very simple switches like a TRIAC. You don't need this complexity.

There is a lot of potential in fridge to grow from 25% up, but the power need is lower. What happens is the higher the power, the more interesting GaN comes. The lower the power, more options are on the table. For example, traditional MOSFETs or even IGBTs sometimes. IGBTs are losing ground. MOSFETs, I think, are coming into play very strong, and they are already very strong because they offer higher efficiency. They are also, at least at that time, right, if you look also at pricing, they are more cost-effective than GaN. I don't know if I answered your question, but, so to summarise, washing machine, HVAC, I think a lot of potential to grow because of the efficiency. Fridge, also, penetration is very low. You can grow a lot there, but there is also competition from other technologies, not only within the same technology spectrum, but also other technology may be good alternatives to GaN.

[[00:15:47]]
Power Integrations has launched TopSwitchGaN, which I think combines PowiGaN technology with the classic TopSwitch architecture. It's an extended flyback converter output of up to 440W. Would you be optimistic about this? Do you think the market needs this right now, and it should be an important potential growth driver for Power Integrations?

[[00:16:56]]
I cannot comment on that, but I can give you my take. I think Power Integrations has been and was a pioneer in GaN. At one moment at the beginning of GaN introduction, beginning of 2020s, we were number one when it comes to market share. Originally, it was maybe mobile. Afterwards, there has been expansion in other markets. I think Power Integrations is and will remain a key player in GaN. Target markets at my time, definitely home appliances, this is a huge market, industrial. Now recently, I mean, recently, since a couple of years and also with the latest news, we know that they are in data centres. I didn't capture that. I was not able to experience that angle while I was there, and I cannot comment too much. I can tell you more about my time. Specifically to your question, yes, I think TopSwitch GaN is a logical next step to the previous product portfolio. As you know, TopSwitch has been growth driver in Power Integrations based on other technologies, not GaN. I guess they are trying to consolidate their position and build more based on the advantages their unique technology brings efficiency, integration, in particular.

[[00:19:08]]
Based on your understanding of the data centre market and the portfolio that Power Integrations has, do you see a particular area where the company could play an important role and provide value in an AI server?

[[00:19:55]]
Yes, I think in order to be more effective in this call, let's get to the next topic. I cannot comment too much on that. Either, I don't know too much or even if I know, there are some things which I was involved with, I cannot share because I don't know what's happening right now. Again, I haven't really observed the market and what they have been doing lately. Hence, I think it's fair to get to that next topic. I'm sorry, yes.

[[00:20:37]]
Could you expand on Power Integrations' strongest areas alongside consumer? You mentioned appliances such as washing machines, HVAC [heating, ventilation and air conditioning] and fridges. Would you say that, at least within the consumer market, those three are probably the best growth potential areas for the company? Should there be any other honourable mentions here?

[[00:21:16]]
Well, whatever needs more power, I'd say, more than 200W, 150W. Wherever that overlap starts to be more thinner, in between MOSFET, IGBT and GaN. People will take decision based not only on performance, but also on total cost of ownership. When performance, meaning efficiency, is so important, it normally happens if the power is very high or higher than 250W, 300W, 400W. This is normally the domain we talked about washing machine drive and HVAC compressor or maybe an external fan. For lower power, then they have alternatives, which are good enough, and they are more cost competitive. Yes, washing machine, HVAC, fridge has huge potential, but it's a little bit a mix. There are also the other type of appliances like dishwasher, hair dryers and so on, yes, they may have volume, particularly hair dryers, but the average selling price of that unit is so low and the market is so competitive that either these guys will go with some local player, maybe even GaN, but local player or they will just use traditional technology, incumbent technology, MOSFET, IGBT because of the price.

The other markets which are growing right now are, okay, not very significantly at this moment, but if you look 10 years from now, robotics. The problem with the robotics is they need a lot of high-voltage to low-voltage converters. As you know, Power Integrations is very strong in high voltage. You have other spectrum of players in this area. I'm talking about robotics because I'm directly involved in physical AI currently. Here, I think the growth opportunity or the potential is lower for them.

[[00:24:03]]
I think consumer and industrial or automotive, in terms of how Power Integrations groups it, have been very large growth drivers for the last few quarters. In terms of those industrial applications – whether it's the motor drives, renewable energy systems or even high-voltage DC transmission systems, etc – where do you see the largest opportunity in the industrial market in the near-to-medium term?

[[00:24:41]]
There are opportunities in industrial, but that market is very fragmented. Now talking a little bit high level. It's very fragmented. Unless you go after the big guys like Schneider, again, I don't know what's happening, but this is what you guys will need to investigate how well-connected or how well they are designing those guys, then mass market industrial, very fragmented, goes through distribution. They have relatively low mid-level volumes and it's very hard to support constant growth there. Also design cycles are very long. Yes, as soon as you are in, you are in for 15 years at least. It's very stable as soon as you are in. That is something, I think, if I were to look okay, whether they have traction there, I will look at top five, top 10 industrial makers, check or investigate whether they have business with those guys because the tail of the market, it's not easy to access. For that, I think from my experience in NXP and other companies, those guys prefer to use, if they have alternatives, suppliers who own more of their bill of material, this is what they like to do because this simplifies the supply chain. I'm not saying specialised suppliers like Power Integrations have no moat, they do, particularly if they bring something special on the table. Yes, that angle, I just wanted to bring it here…

In terms of industrial application, I think everything high voltage… High power, at least for my domain. I'm not talking about these megawatt drivers or things like this, IGBT drivers we have. I just talk about the switches. For the switches within 1.5kW, there are so many applications, pumps, fans, compressors, actuators, a lot of these things.

[[00:27:20]]
You mentioned how the design cycles can be much longer in what we're discussing, compared to consumer. What are the typical design cycles in terms of length and all the different stages that get involved, if we could do a side-by-side comparison?

[[00:27:50]]
Yes, sure. Let me try to. For home appliances, this has been historically relatively long, 24 months in the past. In the past means, let's say, 10 years ago. Recently, while I was there, the tendency is to drive it down. We need to differentiate in between, let's say, a completely new design, which is maybe a new technology or a product suddenly, we go from a mid-end or from a low-end to a high-end. In that case, the design cycle, when it's a completely new product design, it may take maybe 18 months. If it is just, let's say, a facelift, with adding maybe over-the-air updates or not too difficult to add functionality, this can be maybe 12 months. I think 12-18 months is what I have seen. In general, I think, the design cycles are also going down in Europe. Europe means European-based suppliers. Historically, they have been higher also there. I think due to the changes also in their design chain and supply chain, they are getting squeezed and because they are under threat from other tough competition from Asia. 12-18 months, depending on what type of the design we are talking about. In general, HVAC may take a little bit longer because they are more complex. There are also higher ASP vs some other, I don't know, a fridge, a low-end fridge or low-end washing machine. In particular, when you have a high-end HVAC, that can take a little bit longer.

[[00:30:28]]
At least on the consumer side, in home appliances, those design cycles have gone down or become shorter in the last 10 years. Do you expect that trend to continue in five years? Do you expect the design cycle time to reduce even further? For now, do you think the industry is set to a minimum of 12-18 months in some cases?

[[00:31:01]]
I think it must go down because I'll give you an example. A German manufacturer 10-15 years ago would spend 12 months only to test an appliance. Of course, after 12 months of testing, the claim, which was rooted in facts, was this appliance will last for 20 years. Who needs an appliance, which lasts for 20 years nowadays? Not too many. You want something new, every 3-5 years, you want something new. There are new technologies out there, which some of them are really useful to us. Some of them maybe not. My point is the design cycles are pushed down by the changes also in the consumer perception of what this appliance means to them over the lifetime. 10-15 years ago, 24 months, now 12-18 months, I think, is de facto. I think in the future, it will hold. It depends also how fast people can innovate. I think it will go down. I think this is the trend.

[[00:32:33]]
Do you think it's tougher to see the design cycles going down or getting shorter on the industrial side?

[[00:32:43]]
Yes. That's a good question. During my time and particularly in this industrial domain, I haven't seen too much of a reduction in the design cycle there. I think competition is going to play in. Competition, meaning Schneider, or whoever is, the industrial manufacturers are looking left and right and see what their competitors are offering. Definitely, that needs to be taken into consideration. In the industrial domain, quality, longevity, these have been key criteria. I haven't seen too much of wearing off of these requirements, at least while I was active in this space. I'm mostly now looking into consumer, so it's a different dynamic. I don't know the latest in the industrial anymore.

[[00:34:03]]
We discussed the design cycles on the consumer side. If we look at a similar area, how has the cycle of the design win to revenue been trending in terms of when a player such as Power Integrations can actually start benefiting commercially?

[[00:34:38]]
Time-to-money is a very important topic. I think you need to have… Particularly when you go after this industrial space, again, the reward is tempting because as soon as you are there, you stay there. Getting into is an uphill battle, and it can… What is very important in industrial is the fact that it's not like you know by the last day, whether you are in or out. Normally, the industrial makers will build alternative solutions. They cannot afford to rely only on one player. Next year, after they start production, they are in allocation. They cannot get parts. Normally, they build two solutions. Then we would maybe give share to each player, just to keep them interested and connected to the reality. Of course, then they can modulate the demand request depending on the quality. Okay, my point was if you are participating into an industrial design, there are different stages. For example, a stage in that cycle, design cycle is a prototype build. If you are able to hit that, there is prototype build, which is particularly hardware, and then is functional prototype. Functional prototype involves software. As soon as there is a functional prototype build in place, which is working and it's now deployed to testing, then you already look good. Yes, it may take from there another 6-12 months for testing because these guys are really doing extensive testing, pretty long.

At least you know that there is light at the end of the tunnel. How much revenue you will get? Yes, it's a question mark, but at least you will be in. The margins in the industrial space are you can enjoy higher margin than in consumer, as you can imagine. The effort though to support an industrial design, I think it's a little bit longer and more intensive. In consumer, like washing machine and so on, unless you come with something completely new, engineers kind of rely on what they know and what they have done before. Yes, they need your support. Support intensiveness. I think it's a little bit higher in industrial space. That, of course, needs to be taken into account when it comes to cost and time-to-money and the investment which is initially needed in order to win the socket.

[[00:38:31]]
We discussed the various growth opportunities where Power Integrations could benefit. In terms of the actual end demand right now, there has been an inventory build-up and macro uncertainty, as well as the 2025 tariffs and that whole situation. Thinking about all of that, how would you describe the end market on the consumer side? Do you think there is still a bit of an inventory normalisation that needs to happen? Has it normalised, to an extent, and the excess inventory is cleared up now?

[[00:39:16]]
Fairly speaking, I don't know. No, I cannot comment there. I can tell you that companies like Power Integrations are very careful in entertaining forecast because there are cycles, and those cycles sometimes you can predict based on patterns. You look at the past, you look at when this allocation happened and some market triggers and you can have an idea what is the kind of footprint, which may trigger the next one. Companies are very careful to make the distinction in between forecast and the real demand. We are now in another cycle of this allocation. To your question, at least in my area, again, back to the washing machine, fridge, I don't know. In consumer space, the one I'm taking care of right now, which is mostly mobile handsets, laptops and things like this, there is allocation as we speak.

[[00:41:00]]
At least in the particular areas you mentioned, such as mobile phones and so on, would you say Power Integrations is quite well-placed to address that demand as it improves or rebounds?

[[00:41:21]]
I think in mobile phone, the battleground belongs to the Asian players. It's very hard to… Power Integrations was a key player there, and this is in public domain. Innoscience, Chipown, all of these guys came from behind with the GaN solutions, I think they have the lion's share. Laptops, I don't know. I cannot comment there. Power adapters for phones, for things like this, I think, again, you guys maybe know better. I think it belongs to Asian players because there is, first of all, the ASP is very low. There has been a lot of pressure from the market on that. Efficiency, yes, okay, is important. Again, back to what we discussed, we don't talk about always-on. We don't talk about high power levels. You need to look where the power is high enough and the functionality is required most of the time. This is where, again, something which is more efficient would make sense. Otherwise, of course, also cost of ownership is critical. People may choose some existing technologies over new.

[[00:43:09]]
In terms of the lower ASPs that the Asian players can come in on, on average, how much lower than Power Integrations or western suppliers are we talking about? Are we talking about a 0-5% discount, 5-10% discount or even higher, perhaps in the 10-20% discount range?

[[00:43:32]]
I'm very sorry. I cannot answer this question.

[[00:43:37]]
How much of a threat are the China-based players for Power Integrations in the long term? Where are they the largest threat? What does the company need to do to deal with it?

[[00:44:06]]
I think focus and specialisation is what is needed to fight price disruption because as soon as the market reaches critical mass, you will have that threat. If you look back, let's say, 10-15 years, this is exactly what happened with another type of technology, microcontrollers. Also MOSFETs. Microcontrollers was a business entirely, as you know, global players. It's still, the top three are Renesas, NXP and ST, I think, if I'm not mistaken. They left the low-end market to the Asian players. Before, they have been in every single design. The low-end market belongs to the Asian players, where they add value is where some sort of secret sauce would be needed. Same will apply also for Power Integrations and for other global players in the high voltage. You cannot play based on price. Everything what is low cost, high volume, the disruptors will most likely take it. Most likely, the global players like Infineon and ST will just let it go because you cannot leave on that profit margin.

A compressor design. Also, it needs to be said, I think the price disruptors are, again, they look for high volume because they have the capacity, they have also the low price structure. They can address that. They don't care about low margin because they have the volume. With that, they have the profit. They have profitability through huge volumes. I think on the opposite, you look at a compressor, the volumes, okay, are high, but the technology requirements are also high. You need to sustain very high currents and protection. There are new features, new functionalities, not new, additional features and functionality, which a compressor will need vs a very low power fan. This is the market where it's still an opportunity. That is why at the beginning of the call, I took away the internal fan, the one which is low power because that you just give it to the price disruptor.

[[00:47:28]]
On the consumer side, if we break things down across different areas, how sticky do you think Power Integrations' design wins are once they are qualified into a platform? Could you discuss the typical cycles and how easy or difficult it is for a customer to hypothetically switch?

[[00:47:54]]
Just think about you have a design, an AC/DC or maybe even a motor drive design, which normally uses 50 components. Then suddenly, you reduce it by 25-30%, the amount of components. While if you would choose something more in line with what you used before, you don't really save too much BOM. If you are focused on reducing component count, if you are focused on modularisation on very small designs, I think Power Integrations is very well-suited. Yes, it's very hard then to go from 30 components back to 50. I think the stickiness is good as soon as they are in. This is what I liked about, they drive innovation. You can find that everywhere. They drive innovation and integration like this is in their DNA. This is what they have done with their first TopSwitch product. This is what they did also with the motor drive, the business I was in, integration, integration, integration. I think, yes, this is a good argument.

There are also cons. Always, there are sides of the medal. It's like I have a design, I've been doing that for years. Now I have two options. (1), go with what I have done so far, but maybe change it a little bit with another supplier. In principle, it's the same architecture. (2), go with something which brings more value, cuts components, more efficient or things like this. Then the question is, how am I going to deal when allocation starts, when I don't get parts from these guys. I'm not saying this is happening. I'm saying from a supplier, from a firm's perspective, they look at all of these risks. That is, I think, something they will look into before they're making a decision.

[[00:50:35]]
Power Integrations has historically generated some very strong gross margins, especially against its peers. As GaN probably a larger part of the business, do you expect those gross margins to remain structurally above 50%? Do you think the increasing competition in GaN could hypothetically pressure profitability for any player in this market?

[[00:51:06]]
I don't know, and I just don't want to make a statement which is generic here. I don't know how it will play out.

[[00:51:22]]
No problem.

[[00:51:26]]
What do you think will be the largest opportunity and risk for Power Integrations in the next 2-3 years?

[[00:51:47]]
I will try to answer this question based on the time when I left the company because right now, one year and year after, almost things have changed and maybe even leadership has changed the ship direction. I don't know. At my time, so large opportunity in consumer goods. We talked about washing machine, HVAC, where really efficiency matters, where the technology brings advantages.

Risk is, I think, the resistance to change. Again, we discussed about cons when you help somebody achieve… Help a customer design to reduce component count, make it more compact. People or engineers, designers, some people see that as a risk. Not only the supply chain people, but also the people who really do the design. These are considerations which anyone will need to look into before they make such a decision whether they can change the architecture. This is an architect… Sometimes can be an architecture change. Instead of using a big… I don't know how many components, you cut it, into something which integrates those components and adds a little bit more intelligence, which makes perfect sense, works perfect. Technically and practically perfect. The question is whether those engineers, those decision-makers are willing to go that route. There are considerations. It's nothing about technical side. There's nothing about the quality. It has nothing to do with that. It's just the resistance to change of people.

[[00:54:22]]
Would you like to highlight anything else?

[[00:54:43]]
No. I have nothing else to add. I think it was good and questions were framed very well and the categories made sense to me. Yes. If you have more questions or you want to deep dive into one particular area, we can do that at another time.

[[00:55:08]]
Perfect. I think that's a good note to conclude the Interview on. Let me close by saying thank you very much for your time today. I really appreciate it, and it was a really good discussion. I hope you enjoyed it, too.

[[00:55:20]]
Thank you.

[[00:55:21]]
No worries. Have a good day. Take care. Bye-bye.

Transcription ends at 00:55:27 of the recorded material