The need for integration in data centers

00:39:48 – Suresh Venkatesan: The need for integration in data centers is pretty clear. You talk to data center providers, they don’t really care what you use, all they really care about is dollars: „Please move my data from point A to point B for as few cents as you possibly can. I don’t care how you do it.“ So it’s up to us to figure it out – how are we going to be able to provide that capability to these data center users that gives them the lowest dollars per gigabit that they can get.

00:40:16 – Typically they don’t even want to talk to you unless you have a new solution that can provide them with this cost differentiation which, thankfully, we have. So it is very easy for me to have this conversation with them. I’m not talking to them about business as usual. We’re talking to them about a very different and differentiated way to do business, where we can lower the total cost of adoption and the total cost of solution.

00:40:38 – It is true when we had the conversation with the BB Photonics IP where we talked to them about it and they’re like „Oh, I can see how this is going to lower my costs“. So they don’t want to talk to you unless you’re there telling them how you are going to lower their costs.

00:40:50 – And that’s why integration is important. Because you go back 30 years in semiconductors: if you can integrate and lower your costs, if you can integrate and lower your power, it’s all about integration. And so I think that’s the biggest reason why we want to position our company as an integrative photonics company, because the moment you say integrative photonics, that also means lower costs, lower power, and size, weight, and all of the things that come with it. But I think it usually has that connotation with customers that says we’re here providing a differentiated solution on cost that heretofore didn’t exist and no one else can provide it at the levels that we can.

POET offers a superior solution relative to copper

00:41:37 – Suresh Venkatesan: We talked about active optical cables (AOC); I spent some time talking about it before. Our strategy is to create AOCs at virtually the price point of copper cables. That is really what our initial strategy is. And why is that important? If you look at a copper cable that’s 10 gigabits capable today; first of all there is no copper cable that can do 25, but even at 10, these things are fairly thick and inflexible. I don’t know if you’ve actually lifted one of these cables. They’re about a quarter of an inch in diameter, they’re extremely heavy, can’t move, they’re not very flexible. They cost, I don’t know, $15 or $20 a pop. Whereas an optical solution today costs $60 or $70 dollars at 10 gigabits per second. So people are still stuck using copper, they just can’t, for the volume that they need, for all of these server to server connections and server to top of the rack connections for the data centers, optics is just not cost competitive today.

00:42:41 – So what we thought about initially was, strategically saying, why don’t we take advantage of the fact that we actually have a fully integrated solution and attack really this pain point, get it on a cost curve that makes sense, and now you can take advantages of optical communications in terms of lower power and dissipation of heat. Copper cable typically would consume about two watts per link, call it, and we can do it at about a factor of 10 lower than that. So when you multiply it over a number of links, it adds up to a lot of watts and that’s a fair amount of money for operating the data center.

00:43:24 – At least in talking to most of our potential customers or partners that seems to be an application that’s resonating, you talk to them and like „ah, yeah, I get it“ and „if you can do it there’s a market for it“. In fact, some of the people that we’ve talked to, analysts, potential investors, it’s pretty black and white: „What you have to sell is what the market needs.“ So we don’t have to sit and worry about creating a market. A lot of people have to worry about „oh I have a technology, which market do I go after?“ This is a market that absolutely needs what is is that we want to sell, so we just need to make the breakthroughs necessary in the factory and get to these products that we have a vision to be able to demonstrate.

Value proposition

00:44:29 – A direct attached copper cable (DAC) is typically what is used. These are not the Ethernet cables, like we have here [points to them], these are not RJ45s, these are actual communication links, they sit in a QSFP package or a SFP package. These are links that go behind the server. These are not RJ45 links. They typically burn about 3 watts of power per link, the cost is X, the form factor doesn’t matter, because it’s just copper. Weight and flexibility is a big issue with them, they’re extremely heavy. If you see these data centers, they are about 20 foot tall buildings, and if you’ve got servers up on top, you’ve got to carry these cables up, you’ve got to have elevators and lifts, it is not easy. So people want to get rid of these copper cables, they just want to get rid of them at the right price point.

00:45:24 – So conventional AOCs that exist today is a very good product: a half a watt of power. The problem is it costs about three times more than copper, which is too large. You need to be within the gravitational pull of the price point of copper, so then your solution gets attracted in. If you’re too far out, then it’s good, but it’s not interesting. So what POET does, is it brings that price point in, so it gives all that advantages that a typical AOC has, but brings that price point in. Of course, we do that by integrating. Typically an AOC could have four chips, separate, we can do it all together in a single chip. When you take four chips in about 25 square millimeters we can do it in less than five. That’s where the price comes in, the features are similar, the power is similar, but the cost is what we make a big advantage of and in so doing get a lot more interest from the market in terms of cost. That’s really our value proposition, if you will, in AOCs.

00:46:30 – For those of you who might or might not be experienced in photonics: The cost of photonics is driven up by packaging. So, yes, the lasers are expensive and the detectors are expensive and all that. But really the packaging is what blows you away. Because when you put these pieces together, they have to be individually assembled and then somebody has to go in there and align the fibers to them, and so that’s what drives up the cost. When you have this integrated onto a single chip the testing and assembly costs go down dramatically. Silicon has done that for us, thankfully, we don’t have to reinvent the wheel then. Silicon has really driven down the cost of wafer-scale testing and wafer-scale manufacturing.

00:47:12 – So the moment you take an integrated solution in photonics and apply it into a wafer-scale testing or manufacturing, the costs come down dramatically. That’s the advantage. So if people talk about integration, it’s not just about the number of chips, it’s the fact that now that I have these chips integrated there are other benefits that come with riding the cost curves that the silicon world has given us over the number of years.

POET – component cost against competition

00:47:37 – This is to give you some perspective of cost. We’ve run some cost models, we know what foundries cost us, we know what our epitaxial cost is, we’ve kind of factored in what our yields would be. If we just take a typical cost against a competitive solution out there, if we were to sell a single detector, the cost to manufacture that for us is about 33 times lower (33x) than what a competing solution costs. So this is just to give you a sense when I say „hey, why is this a value proposition, why is it important?“ In the worst case we are 2x lower and in the best case we could be up to 33x lower. So we are not talking about small changes of 10 percent or 20 percent, we are talking about 100 percent or 1000 percent.

00:48:23 – So that is what to me, integration does, and that’s what integration, whether it be indium phosphide with DenseLight or indium phosphide with BB Photonics or POET gallium arsenide, integration helps drive costs, and driving costs down drives adoption up. And so that’s really the wave we want to go down and that’s the wave …

00:48:44 – I mean, others are doing it, too, it’s not like we’re the only one out there thinking that this is the wave of the future. There are a lot of people trying to work with silicon photonics, for example, it’s a fairly popular technology out there, a lot of people working on it and with the same thought processes: Find a way to integrate photonics together. There are a lot of companies, Acacia just went IPO with silicon photonics solutions. The good thing about silicon photonics and indium phosphide is every silicon photonics chip requires an indium phosphide laser. It puts us in a really good growth position even from that perspective, because we’re a laser provider. We can provide into the silicon photonics industry while, at the same time, making our own integrated chip to compete for those same pockets. It gives us kind of a foot in both camps, if you will, in terms of integration.