Fabrinet (FN) Earnings Call Transcript & Summary

Hello, everybody. Thank you for joining. Tim Long here, Barclays IT hardware com equipment analyst. Very happy to have Fabrinet with us today. Seamus Grady, CEO, with us. So thank you very much for joining. Obviously, really hot space, and you guys are in the heart of a lot of the key technology growth areas now.So I got a bunch of questions here. I want to start with kind of the maybe the 4 top ones that we get a lot of questions on, and I'm sure you do as well. So maybe we'll start with telecom and DCI. It's been nice, you guys started breaking that out recently, so we could see it a little bit better. Just talk a little bit about from a Fabrinet side, what products and solutions are really seeing that growth on one side and then talk a little bit about kind of the breadth of customer base and who's driving that strong growth.

Yes, we decided to break out DCI because I think for the feedback from investors was DCI has always been in our telecom number. But of course, it's what's going on in the datacom world that's driving the growth in DCI. So we decided to break it out so that you can clearly see, okay, yes, it's in telecom, but here's the number. You can see it separately. And it's been growing very nicely for us. We have 5 -- most of our DCI data center interconnect, most of it is 400ZR and 800ZR, 400ZR plus and more recently some 800ZR. It's been growing very nicely for us, as you can see from the numbers. And we really have 5 customers that we deal with there. This is one that we don't have that we're working on, but we have 5 customers there. Most of the volume comes from probably 2 of those 5. But it's been -- ZR DCI is a really good solution for these distributed networks where you just can't get enough physical electrical power into a data center. So by having the data centers further apart, you can connect them together with ZR products and really get the full effect of the processing power that you have in a data center. So it's...

So it's kind of a scale across.

Yes. Our telecom business overall, let's say, excluding -- so DCI has been growing very, very strongly for us. And I think the demand seems to be very robust and continues to be. For our telecom business outside of DCI, that has also been growing for us, driven more by, I suppose, share gain somewhat by growth in the business, but also share gain, where we've been winning business with a few companies.

Right. Okay. And then you mentioned 5 customers. I mean 1 is reported as a 10% customer. The other one is kind of in ramp mode, I believe. So is it -- is this the type of thing where looking at all the CapEx announcements and power being committed that you feel there's a very sustainable DCI opportunity in the next multiple years, particularly as those 2 customers continue to do pretty well. And if you get the sixth one, that will probably add to it also.

We hope so, certainly. The demand looks to be very robust. I mean, we only guide 1 quarter at a time, but the visibility we do have, we have visibility, obviously, beyond that. But the visibility, it gives us -- we're very encouraged by the demand for DCI and it just seems to be increased.

Okay. And it seems like the telecom side is strong, the traditional telecom, but that's not going to have the same growth dynamic over time. It's a little bit more stable growth rather than kind of exponential what we're seeing.

Yes, we can see that.

Great. Okay. Great. Maybe second topic here, HPC. So also started breaking this out, which is also helpful. I think it was $15 million or something last year, early stages where you're -- you've got 1 line running and a ramp throughout the year. So maybe if you could start by talking a little bit about -- obviously, it's -- Fabrinet as a company does really well in optical and transceivers and pluggables. This is a little -- obviously, a little bit different. So maybe talk about the technology angle, the learning curve, the ramp first and then we'll get into a little bit more on it.

Yes. So we -- our customer is AWS. It's in the high-performance compute category. Again, we decided to break it out as a separate category because it's a new category for us. It doesn't fit in any of the other categories. It's data -- it's data center, but it's not datacom. So it didn't fit in datacom and it would be too big to be in the other category. And we also think it's a good potential market for us to win other business with other customers. We're producing a range of products for our customer, mostly very, very densely populated, very complex PCBAs. And the approach we've taken is we've put a very cost-competitive solution in place, we believe, for our customer. So far, as you said, we did $15 million last quarter, which was really just qualification type volumes. So we just really got off the ground last quarter. And we have -- we're very optimistic about that business. We think it has a lot of potential to grow. We have to earn our way into the business and do a good job and execute very well for the customer. But we really think with a very high level of performance and execution, coupled with what we believe to be a very cost-competitive solution. We're optimistic we can grow that business nicely over the next 5.

Okay. Great. And you guys are clearly differentiated on the transceiver optical business. Do you think HPC will be something where once you're fully up and running, you'll be able to demonstrate the same level of differentiation compared to the competitive landscape?

We think so. And we certainly think there's a lot of opportunity there for us. Again, we're a pure contract manufacturer. We don't have any of our own products. So that's maybe in one sense, an area of the business where we don't participate. Some of our contract manufacturing competitors do have ODM type offerings and the like, we don't. We're a pure contract manufacturer. So the target customers for us are ones who have those type of products, but where they own the design themselves. We don't want to be a product company.

Right. Right. Yes. And there's obviously news every day about a very large TPU custom ASIC platform. When you say there's other opportunities, I'm assuming something like that is a next phase for you in the HPC business? And what would it take to break into another large program like that?

I think in our business, what it takes to break in is you have to do an excellent job for the customer at a very competitive cost. It's -- our business is not a complicated business, not an easy business, but it's not a complicated business. It's all about performance, delivery, responsiveness, quality and doing all of that at a competitive cost. So -- and we also have a lot of expertise in putting together really excellent manufacturing process. Our first foray into this, we have a completely automated production line, 1 line qualified, 2 lines in qualification. And again, so far, we're very happy and the customer seems very happy with the job we're doing in terms of delivery, quality, cost, et cetera.

Okay. Great. Great. Maybe we'll move over to capacity a little bit. You got the Building 10 coming on next year. You've talked about pulling forward a little -- a portion of that capacity. So maybe walk us through how you're thinking about capacity more broadly and what the space is earmarked for? And obviously, demand is good, the fact that you're trying to move some forward. So any thoughts around that would be helpful.

Yes. So we're building -- the construction is underway, Building 10, which will -- when it's -- so just to maybe level set first, our capacity right now in our current footprint, we have capacity for -- our current run rate, I would say, is we're at about a $4.5 billion run rate. We still have room to grow with our current footprint, but call it a $4.7 billion, $4.8 billion capacity that we currently have. The new building, Building 10, when it's finished, will be 2 million square feet with capacity for an additional $2.5 billion of revenue. We announced our decision to pull in about 250,000 square feet of that to June. So we'll be -- as soon as that's ready, we'll be occupying that fairly quickly while we finish out the rest of the factory. That should see us through in terms of capacity for the next [ 5 ]. Then we have room on that campus to build 2 more factories, each of which would be 1 million square feet. So $2.5 billion capacity with Building 10. And then if we were to build out the other 2 buildings, Building 11 and 12, that would give us capacity for another roughly $2.5 billion.

Okay. And how flexible is the space from a product solution standpoint?

It's completely flexible. The way we build these buildings, the capacity is very fungible. We build the building, we fit it out in terms of facilitization and whatnot. But then the specific fit out for each individual customer depends on what the customer needs from us. But at any point in time, if a customer decided that they wanted to move out of a building into a bigger building, we can repurpose that space very quickly. So we're very flexible and can repurpose and reassign that space.

Yes. So the CapEx that you're going to need to spend is pretty success based. You have visibility into needing that capacity at this point?

And also, the decision for us to expand the capacity is a very easy decision for us to make. So to build that 2 million square foot facility, the CapEx is about $130 million, which we can fund from our own resources. We have about $1 billion of cash. We have no debt. So we're able to fund that out of our own cash with no debt. And God forbid, but even if we woke up one day and it turns out the whole industry collapsed, and we never filled up Building 10, the gross margin headwind if the building were to sit idle is 15, 1-5 basis points. The gross -- so the downside risk is tiny. The upside opportunity is that incremental $2.5 billion of business. Approximately, when you do the math based on our operating margin, about 6 months' worth of operating profit at a full run rate would pay for the building.

It seems like we might be getting into areas where if you have capacity, you're going to win. And if you don't have capacity...

Because all of these opportunities are time-based. If you don't have -- like if you go back a couple of years ago when we engaged with NVIDIA and we had a huge ramp with NVIDIA. If we weren't able to ramp, we wouldn't have been successful. So you have to have capacity. As you know, these days with demand.

Right. Right. Yes. And then if there's more success and the 2 other million square foot buildings, is it kind of 6 to 12 months to get 1 of them up and running?

It's probably more like 12 months.

12 months?

Yes. And we're also repurposing other space in both campuses. We have 2 campuses in Thailand. We're repurposing other space to create more manufacturing space. We're also looking for additional and additional factories as well.

Great. Great. You mentioned NVIDIA, so maybe we'll talk datacom a little bit. Maybe start with the demand profile. It's still -- there's obviously cycles in there, but maybe just high level on demand for the datacom business?

Yes. So overall, demand seems very robust. And if anything, it's very robust and increasing. I think with the rollout of all of these AI data centers, it just seems that the demand for transceivers in particular is insatiable. So we have -- our main customer, our kind of marquee customer is NVIDIA, but we have other customers that we're working on as well. But demand is very robust.

And as far as the -- you guys aren't talking 400, 800, 1.6 much anymore. But where are we in kind of the node progression for the large customer on the datacom side?

Well, we're producing with, let's say, 200 gig per lane EMLs. We're producing 1.6 terabit and 800 gig transceivers. We don't produce much 100 gig per day these days. We're producing a significant volume. We could produce more if we had more components. We're constrained on a couple of components. But if we had more components, we could produce a lot more, but demand is very robust.

Right, right. And that demand doesn't go anywhere if there's a shortage of a specific component because the industry is short.

The industry is short. [ Currently ], the demand is increasing.

Okay. And then you mentioned other customers other than the large one. Where are you in that process? How do you think about scale for the other potential customers?

So we're working with a number of other customers. Nothing really to announce at this point, but we're working with a number of other customers, some of whom would be maybe some of the traditional companies who are planning to get into the space or we're also working on hyperscale, on our ability to supply direct to hyperscale with their design, again, not our design but their own design. So several kind of growth factors that we're working on.

Okay. Back to the HPC business with AWS, obviously, there are warrants associated with that customer. How do you think about. You talked about the success of that program to help you do better in that program at AWS and maybe do better at other HPC players. How do you view your success at AWS for other products at transceivers or other? How do you view that?

Yes. So the nice thing about our agreement with AWS, there are no particular products that are excluded from that arrangement. So yes, we'd certainly be using this business as a kind of a proof point with our customer to show them the really excellent job we can do for them with a view then to expanding into other areas of the business including optical interconnect and various other aspects of the business. But really, it's a step one for us and it's a fairly sizable step one. But we think it will open the door for us to other lines of business with AWS.

Okay. Now when you think about like diversification, maybe from optical strength that you lead with, you had this high-profile HPC program. Is that going to occupy the next few years ramping HPC? Or do you see an opportunity for a different product set, like a switch or a power system or something else?

I think we can do both. We can do all of the above. We have a lot of new program ramps in front of us. We have the HPC ramp. We have -- the big growth with NVIDIA is really in front of us on 1.6 and 800 gig. We're ramping new products with Ciena. Our business with Cisco is growing. So we have a lot of growth going on at the moment, but we're happy to continue to grow if AWS or any of the other hyperscalers want us to do more than just HPC. We're very happy to do that.

Okay. And this -- back to this HPC business, should we think about it as a comparable margin structure to -- is there any reason to believe it could ultimately be better? It's probably -- it's a pretty complicated product generally just like transceiver is as well?

Yes. I mean all of these products we have to produce them at industry standard margins, if you like, are hopefully industry-leading margins. But at the same time, when we get into a new line of business, we're always quite careful to do everything we can to make sure we don't do anything to dilute the margin. So it should be comparable in terms of the margins.

Right, right. Okay. Maybe back to the DCI part, when you talk about having the 5, like what does it take to get the other player, like you're trying to get the sixth customer. Is it like engineering, co-development? Or is there something else that has to happen for you to...

They use a different supplier today. They use another contract manufacturer. So we would have to displace the other contract manufacturer. Most of these other companies are very good. They do a very good job. It's not easy. It's not easy to displace our competitors. They do a very good job.

Right, right. Okay. And then I know you're not the CFO, but when we're thinking about the mix of all this stuff going on, is the model generally maybe the gross margin doesn't move around much. But if you get incremental revenue growth, there's a little bit better leverage on that. Is that kind of the way we're thinking about the margin structures?

Yes, I think that's right. I think we're somewhat range bound on the gross margin. We're in the kind of 12.5% to 13% range, which is -- which for contract manufacturing is kind of leading. Our OpEx is very low. It runs at about 1.6 -- on a 1.6% of revenue, maybe 1.7%, it's very low. It was 2%. And as we've been growing the company, we've been able to add these big pieces of new business with effectively 0 incremental OpEx. So the leverage is really on the OpEx and the operating margin. So we should see -- I think the gross margin will stay in that range and then the operating margin, we should see improve over time.

Okay. Maybe let's touch a little bit outside the AI world for a little while, the auto business. Obviously, everyone's been having struggles with the auto end market. How do you view your participation there? Is there an end in sight? Or what do you see for turnaround in that business?

So our auto business is not that exposed to -- so one part of it is, but it's not that exposed to, let's say, the consumer-facing piece. We have one customer, which is we call it traditional automotive. That's quite steady. So that business has been very steady for us. It hasn't really grown that much, hasn't shrunk that much. It's very stable. And then the 2 other parts of automotive for us, one is EV charging, which is more on the kind of infrastructure side of automotive. So that business has been growing nicely for us over the last while. And then the third part of automotive for us is the whole area, we call it LiDAR and sensors generally. We did capture really all of the LiDAR customers. And as time has gone by and as the number of companies in that space has shrunk as they've either acquired each other or whatever, we really have most of the players there. So I think that part of the business could grow nicely as and when LiDAR as a technology starts to take off.

Okay. And then I'm sorry, I'm hopping around a little bit here. But back to the HPC business because I think there's a lot of attention because it's a huge opportunity. I think you talked about the 1 line. Just walk us through the time line of the ramp, how you go from like the trial line to full production? And is this going to remain PCB only? Or will there be other elements of their HPC program that you could participate in? And I think the main incumbent does the PCB, but they also do the card as well. So is there other -- maybe talk about the ramp and then talk a little bit about not just like transceivers, but related to the HPC.

So we do that as well. We do the main and the...

Okay. You do both. Okay.

Okay. Yes. I think there's opportunities to do more than just the PCBA. For now, last quarter was just about getting qualified. So we got qualified. We had that $15 million of revenue last quarter. So we're qualified now we're up and running. We have -- so we have 1 line qualified, 2 other lines that are in qualification at the moment. And I think we'll ramp to that kind of first plateau of revenue probably by the June quarter, maybe a little bit earlier, but probably by the June quarter. And then really, it's all about how well we perform. If we do a good job, we perform well, we execute very well, and we're easy to do business with and they make significant savings, then I think there's an opportunity to take more share there and grow that business. So it's really down to how we perform.

Okay. And I'm assuming currently, it's -- you don't have a full view on market share, but it's starting out with whatever percentage of the business. And then if you perform then that has the opportunity to move higher?

Yes, yes. But again, there's no guarantees. We have to earn every piece of business and what goes up can also come down.

Yes. And there's been a lot of discussion about that program with going to the third version from the second to 2.5 to 3. Is Fabrinet, with the ramp phase, are you able to participate in evolutions of the program? So you don't have to be requalified. They're...

Well, I mean, yes, the new product would have to be requalified for us. But we are requalified as a supplier. The last time around, there was really 2 things. We had to be qualified and we have to qualify the product. So that first part is done now. So that should be a more straightforward exercise.

And I believe the competitor in that one is more of an ODM model. So from a pricing standpoint, could be beneficial. You said if you -- if there's cost savings that would...

Yes. As far as I know, they do have some ODM business. But certainly, the products that we're making are contract manufacturing where the customer owns the IP. So from that point of view, it's kind of interesting for us to see that kind of overlap between ODM and contract manufacturing that. Yes, in theory, it's ODM, but it's not ODM. It's contract manufacturing. Because ODM is not for us. We don't have any of our own products.

Yes. Maybe a higher level one. When you think about as a CEO, you got the capacity stuff we talked about. You got some pretty massive opportunities, right? This HPC deal could be $15 million now and 10x that in a year, right? How do you plan not just capacity, but infrastructure in the company, investment? How do you manage with uncertainty into just how big some of these revenue ramps can be?

Yes. We take a really long-term view. So we have a couple of processes that we use internally that work really well for us. We have a rolling 8-quarter revenue forecast that's frighteningly accurate because we're very plugged into the customers. And when we build our revenue plans, they're bottom-up revenue plans. We don't build the top-down. So believe it or not, we don't say we have to get to be a $10 billion revenue company. That's not our goal. Our goal is to grow. We always like to grow at 2x the rate of growth of the industries we serve, 3x the rate of growth of the contract manufacturing industry. That's what we like. That's how we like to grow. Over the last 10 years up to the fiscal year ended in June of this year, in the prior 10 years, our compound annual growth rate was 16%. We compounded the earnings 21% in that same time period. So we have, we think, a good track record of growth and success. As we look ahead, like I say, we have an 8-quarter really close in rolling 8-quarter revenue forecast. We have a 3-year and we have a 5-year. So we use that 5-year. You can't be reviewing and sitting down looking at strategy every month. It's something we sit down periodically usually kind of once a year, more frequently if something big comes along, and we set out our plans for the years ahead because these things take -- if you build a new building, it takes 1.5 years to get it off the ground. But we're in a fortunate position is that -- in that our -- if you like, our footprint, our geographic footprint is quite compact. We're financially very strong. So our balance sheet is very, I would say, pristine and strong. We're able to fund our own growth. We really just sit down and look out 5 years all the time and make plans accordingly. So -- and we're in a very fortunate position that the customers are very happy to share their demand and their plans with us. And we're able to take that and use it to build our own plans. And the customers know when they share their plans with us, we're not going to hold them to it. It's not like we're going to turn around in 5 years and tell them, you told me you'd need this in 2029 or something. No, but it's really good to get that information from the customers. And the customers are really good at sharing their plans. What also helps is we're always working with the customers, not just on the current product, but on the next-generation product and the one after that and in some cases, 2 or 3 generations after that. So when you have that kind of visibility, it just allows us to see what's coming down the tracks and hopefully see around corners a little bit.

Okay. Yes, you could feel free to send me that 8 quarter rolling, it would be lovely. But -- okay. I think that we got a minute -- a few minutes left, but I think that does it for me. Thank you so much for the time. Really appreciate it. I know it's a very busy time. So thank you.

Thank you.

Thank you, everybody.

Thank you very much.