Oleg Khaykin: I mean — and it varies. So when you start production you tend to do a lot more tests. So you have a lower number of ports per, let’s say, a million dollars of equipment. As you get with experience curve, you start — and you feel comfortable, you start de-contenting the test, so you spend less time on a tester. So yes, the equipment predominantly goes into the production line. I mean, you have all these factories in China and other places that are building these modules. So when they start, they generally use more intensive testing. And then as they get more comfortable, they start reducing and doing more of a sample testing or less extensive testing per module. So that’s why it’s continuously moving target. So I mean, in one case we’ve kind of did a one project and we have gauged, it came out to about $0.40 per module of CapEx, per module of capacity.
So if you’re doing a port, you need about $0.40 investment per port on the capacity line. So you build, let’s say, 1.5 million units a week line so you probably — I mean, I think it works out to about $0.40, about $300,000 for that capacity that you got to invest. But it’s only one data point and other people do it differently so they spend more. So it’s still very early to tell.
Alex Henderson: So we’ve already seen very significant ramp in productions of both the NVLink and InfiniBand products going into the AI clusters. And from what I can tell, you really haven’t seen any meaningful contribution from that at this point. So should we then think that as we move into the second and third phase of production ramping that the sampling rates actually go up and therefore, we shouldn’t be looking at the rate of growth in AI as the primary driver of the overall demand curve as opposed to the sampling percentage?
Oleg Khaykin: Well, I wouldn’t go that far, because remember what the first thing they did is they redeployed the same lines that were building telecom coherent business that dropped quite significantly. So they redeployed those assets to the AI data centers. And just when you also think about it, if you’re just doing alignment on like, say InfiniBand, you’re putting photonic integrated circuit aligning with the processor that is really more semiconductor packaging. When you are actually building the actual module with lasers and everything else, that’s where you tend to use more of our optical test equipment.
Alex Henderson: Just going back to the telecom piece for a second. There’s obviously a very large inventory glut out there of telecom equipment that has to be absorbed. Has there been any build in inventory in your product areas or is that just something that didn’t happen because they weren’t constrained as much on those type of products?
Oleg Khaykin: So we have no inventory in the channel. I mean, pretty much the orders kind of got turned off, I’d say, December quarter of calendar 2022. So if anything, a lot of the inventory in the field is getting long in the tooth, and we know that there is a lot of wear and tear and there needs to be a replacement coming up. So we actually are already seeing signs that people saying, hey, I will need to replace, I mean, what would be the terms, what would be the lead times. So in that respect, people will kind of sweat the assets, they will swap the assets, they will cannibalize and then they’ll have to do a wholesaler replacement. So when I say kind of flattish 2024 or like — I just think — I just don’t think — I know they need to do it, I just don’t know if they can afford a massive replacement.
But there could be a good chance that in the second half we’ll see more and more of these type of things popping up. But I don’t have that kind of visibility beyond six months.
Operator: Your next question comes from a line of Meta Marshall from Morgan Stanley.
Meta Marshall: Oleg, you mentioned kind of you were more encouraged about cable spending kind of earlier in the year. Just wanted to get a sense is that DOCSIS 4.0, is that just their networks are running hotter, just given some of the comments you just said to Alex? Just kind of what is the trigger to that investment? And then on the flip side, since you kind of think that wireless may take a little bit longer, just what do you think should be kind of the early signs of wireless resuming?
Oleg Khaykin: So if I look at — so cable, first, I think, we know it is going to be happening. It was actually we were expecting some of the orders start popping up in the March quarter and then accelerating into June. From what we’ve heard, and I’m not going to name any names, but there’s been a delay in some of the core technology development by leading infrastructure providers. So I think that is being pushed by one to two quarters in terms of getting the software ready and everything needs to be working. So I think that’s where we are with cable. But I do see actually cable happening this year. What was your second part of your question?
Meta Marshall: Just on the wireless side.
Oleg Khaykin: So the wireless, we all know, you’ve seen the Ericsson, Nokia, Samsung, and all the deployments with T-Mobile, Verizon, AT&T. So that has slowed down. So the area where we are seeing less is on kind of the field equipment and a lot of the people sales related to deployment. Where we continue to see CapEx being spent is on product development. So we have not seen significant decline in the R&D CapEx for 5G, and now we seeing some of the elements of 6G popping up. It’s just, generally, I think the wireless infrastructure is a bit more muted in terms of aggressiveness, I would say, in Europe and North America. Now that said, India is doing pretty well. But it’s obviously — I wish the margins were better in India, but I think, clearly demand is — right now, India is one of the few bright spots for infrastructure deployment.
