Jul. 18 at 5:45 PM
$VICR $NVDA $AMD $GOOGL
VICR is slated to announce earnings Tuesday before the market opens.
To give people here some more perspective regarding the opportunities with VICR (disclosure: VICR is my largest holding), I wanted to provide a couple of questions and answers coming out the Annual Shareholders Meeting VICR held last month. The questions were actually asked by my subscription channel partner @JoeDab12 who was there in attendance.
In addition, I will provide a link to the full shareholder meeting at the end of the post. If you are an investor in VICR, or considering investing in VICR, IMHO watching the full annual shareholder meeting presentation is a must.
Question:
Can you confirm how many OEMs and how many hyperscalers, not counting Cerebras, you have committed to VPD?
Answer:
They're all going to use VPD. If we look out two years, an AI system without VPD just doesn't work. So let me be more specific with respect to this.
Nvidia has not deployed a system with VPD. They suffer major constraints in terms of power delivery capability and GPU performance becuase of the lack of VPD. Cerebras, in terms of its power system, has a solution that is much, much more advanced. The king of the hill with a dominant position can, for a certain period of time, to the extent that they can sell everything they can make -- or they can sell twice as much as they can make --take, and do so very logically, the position that we're going to commoditize, we're going to drive multi-source, we can do without advanced power system technology in order to accomplish those other goals.
And it works for a while, until the competition that needs to catch up with the leader by bringing about better hardware and better solutions, gets far enough down the path to make the strategy of keeping it simple and keeping it multi-source no longer viable. Whether it's Nvidia, AMD, any of the OEMs and, most importantly, the hyperscalers -- because the hyperscalers, as you know, they are competing with the OEMs, right? Google is an example with GPU. That's actually the one hyperscaler that took the lead with respect to first generation VPD, paying the price not just in terms of very expensive VPD solutions, but also solutions that that are not at all reliable, very difficult to assemble.
All the other hyperscalers are going to have to go down that path in order to be able to deliver the kind of performance that the Google TPU has, which is quite good. It's enabled by VPD. Without VPD their system would not be as good as it is.
Question:
In October of last year, you mentioned you got to a
$90 million run rate in licensing and you had a line of sight to 'double that in the next couple of years' was the phrase you used. Do you think that's accelerated now?
Answer:
Yes. I think the
$200 million,
$300 million contribution from licensing is a conservative number. I think frankly that could get a lot higher. Just one hyperscaler, just one might lead to that number becuase the current consumption of these devices, the value of VPD per ampere delivered, the number of GPUs or TPUs, when you do the math, the numbers get very very large, very quickly even in 2026 and 2027 is going to be a big step up relative to 2026.
So, needless to say, as with everything, you got to walk before you can run. Obviously, our licensing practive is well established. We've got a methodology but we have to start at a certain level in order to make it easy enough to adopt. Obviously, we got a pace now of actions at the ITC. We're going to be accelerating that pace. Our strategy, very simply, is to make it a no brainer for either hyperscalers or OEMs to take a license from Vicor.
These licenses are typically a couple fo years. They come up for renewal and the logic of having them as two year licenses is that, for one thing, particularly in AI, it's very hard to forecast what their consumption is going to be down the road. But, beyoind that, is the number of patents that we successfully assert and the overall coverage in terms of power system technology gets expanded. The value that we can deliver grows. So we have an opportunity to bring about more revenue, licensing in particular, out of this initiative.
Think about it this way. There are hundreds of billions of dollars in AI hardware. Couple that observation with the proposition I made earlier, which is that without VPD there isn't going to be any hardware. And that's a piece of technology that Vicor pioneered and patented. You couple those two statements and you can begin to do the math with respect to what the licensing business can be worth.
Link to full presentation:
https://vimeo.com/1202920258/b8750092eb?share=copy&fl=sv&fe=ci
