I mean the reaction is astounding. And they’re astonished that we can put as much current as we can into that small of a package. And there, if they’re looking at VPD today because that’s where they’re going to go as Patrizio said, all of the conversations now are about VPD, it’s not about lateral. The current Gen 4 is lateral vertical, but Gen 5 is all about vertical. They look at that and they say, “My God, I don’t need to disturb my capacitive layer underneath my board. I can put one of these tiny little current multipliers there. I don’t have to worry about mechanicals, thermals.” Everything sort of fits. And so the reaction is incredible when compared to companies — similar Chinese companies that are trying to — that copied our MBMs are trying to do this thing with VPD copying basically the wrong thing, which is a stacked package, really heavy, terribly terrible.
They’re leading customers down the wrong path and the customers can see that when they see our Gen 5 technology. So the excitement is really incredible, and I’m very confident, as I said, we’re going to take major share here in this market.
Patrizio Vinciarelli: First-generation VPD is an horror show from a manufacturability perspective and from an IP perspective as well. So there too unscrupulous foreign suppliers are putting not just themselves but major OEMs at substantial risk with respect to continuous supply, both in terms of the viability of the solution from a robustness perspective and from the perspective of viability from an IT perspective or non-infringement perspective, I should say.
Jon Tanwanteng: So it’s fair to say that the pull is much stronger on Gen 5 than it was in the Gen 4, Gen 4.5 that you have out there?
Patrizio Vinciarelli: Yes. I think that customers are going to get both performance, right, and power savings. And we talked about all the advantages of reliability, of cost that add on to all of that. And with our scalable ChiP fab, we’re going to be able to work with customers on flexible supply chain solutions. So we’re sitting in a very good place, Jon.
Operator: Next question is coming from the attendee who joined over the phone. So please introduce yourself before you have your question.
Unidentified Analyst: Richard Shannon from Craig-Hallum, again, guys. Phil, I wanted to ask you about one of your statements in your prepared remarks here about the 5G technology and kind of targeting 2 different groups with different priorities here. The last part of that statement was you believe the first group will embrace level of innovation for 5G from a multiplicity of ChiP fabs. Since you only have one, at least that you’ve talked about, are in plan, can you parse the statement a little bit more and tell us what you mean by that and over what time frame we could see a multiplicity of ChiP fabs, please?
Patrizio Vinciarelli: Well, I’ll try to answer that question. Again, given the lack of sharp visibility with respect to that. So bear with me as I answer your question in general terms. As you know, it took many years and a very substantial investment to bring together a first ChiP fab. In some respects, this initiative is similar to the kinds of fabs most of us are familiar with, semiconductor fabs. And as in that case, the technical complexity, the complexity from the perspective of equipment and processes is such that bringing to closure, the first fab takes, again, considerable time, persistence, and investment. But the great news with respect to having done it is that replicating it is something that can be done on a much shorter cycle time with a high degree of predictability.
And the investment involved while substantial, it’s in the hundreds of millions. On — in the right perspective, from the perspective of OEMs wishing to have flexibility from a sourcing perspective, given their system capabilities and value propositions is, in that perspective relatively small investment. So we do anticipate more than 1 fab coming to the market to bring about an ecosystem where large OEMs that cannot, for good reason, rely on capacity from just the first Vicor fab will avail themselves of capacity from power facilities.
Unidentified Analyst: I guess we’ll parse that statement and ask about it in the future, Patrizio. I’m going to follow up on one of your answers to a prior question here talking about, I think, and I’ll probably not get your term or wording correct here, but you’re only engaged with customers that are interested in vertical-only solutions, either first or second gen here. Is that essentially saying — are you saying that you’re not seeing anyone with a sufficient level of power and all the use of 1,000 amps, maybe that demarcation line, tell me if it’s different, that are not looking — are you telling me they’re not looking at lateral at all or you’re just talking with the guys that are only looking at vertical?
Patrizio Vinciarelli: So they are typically looking at programs on different time scales, in some cases, a few months or the better part of the year or a year and a half with different kind of requirements. And our focus is to engage where we can contribute substantial value through, in particular, 5G solutions with their unique set of attributes. And those are particularly differentiated when you get past the lateral, right? It can be lateral-vertical. We demonstrated that with lateral-vertical relative to lateral at current levels of roughly 1,000 amps, you can achieve something at the order of nearly 150 watts of PDN, both direct and indirect power savings. Those are very substantial. And those when applied across data centers, result or can result in the megawatt scale savings that Phil referenced earlier.
So there is a value proposition there. And that can be done with our 4G technology, even though the current density of that technology is a small fraction of current density with 5G. So it is a bet, a lot of vertical. There is a value proposition there, and we can engage there to enable customers to achieve their objectives. But the much bigger opportunity is looking a little beyond the next few months or 6, 9 months. Our systems are going to go into production, as Phil said, late next year or early in ’25 that we’re with 5G MCMs, a few amps per square millimeter, the whole new world of capabilities that are enabled and in turn enable solutions that around go around. They’ve got all the attributes, right? They’re scalable. They are robust from a mechanical perspective, they’re robust from a thermal management perspective and they are more cost-effective.
Phil Davies: So Richard, this is Phil, just to add on to that. A couple of weeks ago, I got invited to a big data center, the company’s strategic supplier day where they had their executives presenting on the future road maps and the challenges of building AI out in their existing data centers as well as adding more data centers going forward. And as you can imagine, as Patrizio mentioned here, we’re talking about data centers of the 20-megawatt plus level, and getting power into those while also trying to get carbon neutral is a massive challenge. So if you look at the amount of compute that’s going to be needed to support AI going forward, both on the CPU side and on the GPU or ASIC side, plus the network processing. Even if you save 10 watts or 5 watts per the CPU, that’s a massive savings.
