John Dillon: Yep, yep. Okay, I’ll get back in the queue. Thank you so much.
Operator: Thank you. Next, we have a question from an anonymous participant who joined via a phone. I will unmute you now. .
Patrizio Vinciarelli : Hello, we can’t hear your question.
Phil Davies : Either there is these delays in holding. Operator, please remind questioners, they have to hit star six to unmute themselves.
Operator: Yes, from their side as well.
Phil Davies : Okay, let’s go to the next question then operator.
Operator: All right. Next, we have another question from Jon Tanwanteng. Please go ahead and unmute yourself from your side as well.
Jon Tanwanteng : Thanks for the follow up. My question is, Phil, just what do we stand today in the migration 48-volt data centers as a whole? What inning are we in? How much are you seeing that play out in your bookings and pipeline?
Phil Davies : Yeah, so definitely the other hyperscalers that we’ve been waiting for are working on 48-volt native rack systems. In the meantime, we’ve talked about this they are buying basically power distribution boards, incorporating our MBMs to do the 12-volt native to 48 volt level necessary for the advanced AI systems. But we are seeing native rack developments now with the hyperscalars in North America and also overseas, because of the power levels that these the –addition of AI chatbots, whatever into the datacenters are requiring for the 48-volt power distribution. And in the supercomputer area, it’s even higher. It’s up at the 380-volt range where we sell high voltage bus converter. So the power levels just keep going up, John.
And, we have the technology to do the downconversion to 48. And then the 48 the point of load either with bus converters for low power CPU systems, or factorized power for very high power GPUs and ASICs and optical network processor. So we’re setting really well with technology migration in our customers and being able to follow that and and even be ahead of that with no Gen 5 solutions coming along later this year.
Jon Tanwanteng : Got it. One thing I’ve been surprised with is just the total power consumption of the next gen CPUs. They’re approaching the levels that the last generation AI processes that they use your level solutions are at. Do you see the CPU market ever going to your — finding your product necessary to drive the power uses that they’re starting to run it?
Phil Davies : Yeah, I think Gen 5 actually opens up CPU — higher power CPU applications for us. And who knows, maybe even lower power CPUs as the rack goes to 48 volts direct conversion from 48 volts with Gen 5 can be very cost effective, and very dense. So we’re going to go after that market as well when Gen 5 comes along.
Patrizio Vinciarelli : But make no mistake, any system today, high-end system. Let’s say high-end GPU powered 48-volt to allow PDN is handicapped by the power system. It cannot deliver its compute capability because of the limitation of power system.
Jon Tanwanteng : Understood. If I’m asked the question on the quarter, what caused the constraints of your supplier? I know they’ve been pretty spotty in the past, is that going to continue to Q2 and just give us a little color as to what happened in the quarter compared to where you thought it was going to be?
Jim Schmidt : Yeah. And so I think, John, as we said, it was mostly a function of the outsourced manufacturing partner, which is we’re in the final stages, now we’re bringing the production in-house. So, in some cases, the supply wasn’t as prompt as we needed it to for the quarter anticipating the output. So, we’re — as I described, we’re into May now with qualification, June-July qualification, so facilitating the equipment and made and we brought in. So I think as we get into the third quarter, we’ll be clear about our ability to generate the output we need.
Jon Tanwanteng : Okay, great. And then Jim, just one more follow up. Do you expect these tariff claw backs to be relatively continuous as you go forward? There was really spotty and kind of unpredictable. How should we think about that, and then kind of what’s in the backlog for you to claim?
Jim Schmidt : In the short term we can, we’re sort of expecting the same level for the quarter. That’s our anticipation. And then it’ll kind of throttle down a bit because we’ve basically been clawing back what amounted two years ago worth of tariff payments. So that’ll come — that’ll further down over the course of the year.
Jon Tanwanteng : Okay, great. Thank you.
Jim Schmidt : Okay, all right. Next question.
Operator: Yep. And we have again, the person who was about to have a question earlier through the phone line. So please unmute yourself using star and six. I will also unmute you from our end. Please state your name and go ahead with your question.
Richard Shannon: Hi, this is Richard Shannon with Craig Hallum. Can you hear me?
Patrizio Vinciarelli : Hi, Richard. We can.
Richard Shannon: Excellent. I’m glad I was able to make it in this time. Thanks for taking the questions guys.
Patrizio Vinciarelli : We’re not trying to drop you.
Richard Shannon: I was blaming my input skills on my phone here not you guys. I guess want to kind of ask a follow up to a few different past questions, including starting with Quinn’s about how to think about when you’re seeing inflection points and kind of the characteristics of when that happened, in particular understand your response to that. A couple of those has been you don’t have complete clarity in that. So let me kind of ask a multi-parter on this and what you kind of go read like to, is I think it gets the gist of my question here. So not exactly sure when that will be, sounds like could be either with them lateral vertical. I’m assuming these are with AI accelerator type applications. But also could you characterize whether these are designs that are that are done and completed and just haven’t been greenlit at the time? Or are they still designs in flight. Maybe just kind of help clarify the dynamics here, among those lines, please.
Patrizio Vinciarelli : So we’re working with a number, as you point out of AI accelerator car companies. They’re all in different stages. Some of them are actually in early sort of pre-production, others are in qualification. Others are also got advanced developments going on that will be qualification later this year and then ramps later this year, maybe more like early next year. So it’s across the board, Richard really. And it’s a mix of applications, but most of it is lateral. And then we’ve got some lateral vertical activity going on as well.
Richard Shannon: Okay, perfect, that is helpful. Needless asking an automotive question here. And I may have, Phil, I may have missed all of your comments about kind of the latest update here in the last quarter. But maybe I’ll ask you to kind of do a two-parter here again, which is repeat what you said about those designs, and maybe kind of give us the top-down here and how many designs you have, in what functionalities here. Your comments on the press release talking about applications up to 150 kilovolts, which seem to apply inverters in EVs. Maybe even kind of characterize what functionality are supporting them. And just to clarify, this sounds like calendar ’25 timeframe to expect material revenue, is that accurate.
Patrizio Vinciarelli : And that was kilowatts, not kilovolts.
