Good, I guess it's good morning still. I'm David Vogt from the UBS Hardware and Networking team. Thank you again for joining the UBS Global Tech Conference, and we're excited to have with us today, this morning, Cisco Systems. With us, we have Bill Gartner, SVP and GM of Optical Systems and Optics Group. Bill was kind enough to come back again this year after subjecting himself to us last year. In the audience, we have IR, Sami Badri, and Ryan Cui as well. So please direct any questions to them later if you have any other additional questions. One final note, we're gonna take questions from the audience through the app. So if anyone has a question, you could submit it through the app. It'll show up on an iPad that we have up here. So Bill, thank you again for joining.
Thank you for having me.
So we were just making a joke that, you know, Optics and optical is kind of the hot sector and kind of back in vogue. So maybe we can kind of just level set, you know, obviously you were here last year. You do a ton of presentations at OFC. You're sort of the go-to guru at Cisco. So maybe just to lay it down for everyone, kind of what your role is, what's under your purview, and kind of your day-to-day strategic responsibilities?
Sure. Be happy to do that. But before I do, I have to—I got one job from Sami and I gotta make sure I do that job, which is to say that I will be making some forward-looking statements and they may differ materially from actual results, and all of our risks and disclosures can be found in our 10-Q and 10-Ks on our investor website. I think I've got that right, Sami?
Yep. Perfect.
Thank you. Thank you for having me. I'd welcome any questions you may have. But let me just start by describing a little bit of my role. I have responsibility for Optical Systems and Optics, and that really comprises three business units. One is the Optical Systems Group. Optical systems are those systems that are used outside the data center, or outside of central office, if we're talking about telco. That environment is characterized really by two things. One is that the distances are long, meaning more than 80 km. We're sending a signal on a fiber across a city or across a country or even between continents undersea. The other attribute is that the fiber is constrained, meaning that we can't simply ask a customer to deploy new fiber between two endpoints. We have to use the fiber that's in the ground.
And that really means you have to put many signals on one fiber. That technology's called DWDM. And in that segment, we compete with guys like Ciena and Infinera, ADVA, used to be ADVA, and Nokia, ZTE, Huawei. So that's one business. Then I have responsibility for our Optics business, which are basically the transceivers that we sell with switches and routers. Those are inside the data center. That world is really characterized also by two things. One is that the distances are short, and short in this world means less than 10 km. And the other attribute is that every port on a router or switch gets its own fiber. We put one signal on each fiber. So that's very different than what happens once you leave the data center, where you have to put many signals on one fiber and send it over a very long distance.
That's a very substantial business for us, and we serve all market segments, including campus, a service provider, public sector, hyperscalers, and commercial and enterprise segments. And then finally, I have responsibility for Acacia. Acacia we acquired about three years ago. The reason we acquired Acacia really, there were a couple reasons. One is it gave us access to coherent technology, which we use in our Optical Systems. That's the technology that allows us to send signals over very long distances and put many signals on a fiber. But much more importantly, Acacia had taken that technology, which is delivered in the form of a line card, a transponder that sits in a chassis today. And they have put that technology into a pluggable form factor, a pluggable optic that can go directly into a router. And we feel that that is a very significant transition in the industry.
Technically, it's a transition, but operationally and economically, it's a transition for our customers as well, and we're driving that very aggressively, so it's really those three businesses: Optical Systems, Optics, and Acacia that I have responsibility for.
That's a great overview. So maybe if we could start with ZR.
Mm-hmm.
If we had this conversation a year ago, has the market developed in a way that you had anticipated? It really feels like it's moving much quicker.
Yeah.
Than a year ago from when we had this conversation. And so how are you thinking about, you know, ZR and Cisco's positioning within that market and broadly defined how the industry is migrating towards ZR, given the needs and sort of the, I guess, bandwidth capacity issues out there in the marketplace?
Great. So I'm actually one of the few guys in Cisco that can carry around my portfolio in my pocket. When I talk about a ZR optic, it looks like this. This is the same form factor that you would have for an optic that's plugged into a router or a switch inside the data center. It looks exactly like this. It has the same electrical interface. The only difference is that a ZR optic can send a signal for classic ZR about 100 km. For ZR +, it can send up to about 1,000 km at 400 gig today. We've just introduced 800 gig variants of that as well. So what this does is this replaces a transponder that's sitting in a chassis as part of an optical system. And, David, you're right.
When we did the acquisition of Acacia, we predicted that this technology would overtake transponders, would effectively cannibalize that portion of the optical market. We were okay doing that 'cause we're a relatively small player in optical, so we were perfectly happy cannibalizing one portion of our business in order to get a much larger share in this business. This business has grown pretty significantly since I was here last year. We're now deploying this in massive volumes with five hyperscalers. They're using that for data center interconnect. Different applications for each one, but all of them are basically data center interconnect, different distances. Some are using it in basically 100 km applications. Others are using it for several hundred up to 1,000 km applications at 400 gig. We've shipped over 350,000 of these transponders, so massive volumes for this technology.
We, I think last year when I was here, we had about 80 service provider customers that were deploying this technology. We're now well over 200 customers that are deploying this technology. My prediction is that within the next three to five years, this will be the dominant technology that's used for data center interconnect and for all metro applications. Metro being like less than 1,000 km for a service provider like a Verizon or an AT&T, and some long-haul applications as well. One of the things that we just introduced is a 400-gig variant that can go up to several thousand kilometers. That can attack the long-haul market segment as well. I think we've had very good success. I still think we're early stage with this. I think there's much more opportunity for us, but the traction's been very good.
I think even looking at some of the competitor responses, competitors who might have tried to put this technology into a very small box initially, I think they're coming around and acknowledging that this is going to be the dominant deployment model.
To your point about the distances and sort of the applications, you mentioned metro next.
Mm-hmm.
What kind of traction are you seeing in the metro portion of the market for pluggables at this point?
So I would say most of the service provider applications, the hyperscaler applications are virtually all data center interconnect. The service provider applications are virtually all metro applications. So I think it is gaining very significant traction there. But I would also acknowledge that those 200 service provider customers are for the most part tier two, tier three service providers. The tier ones have not embraced it yet. I do expect that they will. They're all evaluating it. Many of them have done trials. They have to get their operations tuned for this because it's a different operational model. But there's no question that, in the next three to five years, I'll be sitting here saying that, like, the industry has completely flipped and, metro is this.
In tier one SP, what does it mean for their existing infrastructure footprint? Right? It's a different operational model, but does that mean there's this ton of stranded capital?
No.
No?
No. Important, very important point. You know, we've been through technology transitions before where the proposition is, well, you can take advantage of this great new technology as long as you forklift your old technology out or as long as you abandon your old technology, and strand capital, for instance. That is absolutely not the case. We've introduced a variant of the ZR called ZR+. We call it Bright ZR+. That's the Cisco name for it. That allows a pluggable optic to ride on a fiber right next to a transponder-based optic. So if a customer has started deploying transponders on a fiber, and typically service providers grow capacity over time, they don't deploy all that day one. So they might have a few wavelengths that are being driven by transponders on a fiber.
They can just stop deploying those transponders and start deploying pluggables with zero, zero, stranded capital at that point. So it is a cap and grow, and there's effectively no economic penalty. In fact, it's all economic gain for going to the pluggable optic. Operationally, it's a different operations model, and that's really where some of the resistance comes in because they may have, for instance, an optical planning team, an optical deployment team, an optical management team, and they have an IP planning team, IP deployment, IP management. And now they have to bring those together in some way. And that's really where the challenge is for many of our service provider customers.
Can I ask about interoperability? So I think this plug should be interoperable with your systems as well as competitor systems.
Mm-hmm.
Do you have enough evidence or enough data to suggest that, you know, the interoperability's not an issue? Meaning, you know, a Cisco ZR plug.
Mm-hmm.
Will work well with XYZ system. So there's no, from a carrier perspective, from an SP perspective, there's no limitation in terms of them mixing and matching sort of.
Yeah.
Technological.
It's a great question. So there's really two elements of interoperability. One is that you can have this pluggable at one end of the fiber and another vendor's pluggable at the other end of the fiber, and they would interoperate. That is pretty easy to do, actually. The standard is pretty well defined, so we have high confidence that will work. The other level of interoperability is actually the more challenging one, which is when you plug this optic into a host, whether it's a Cisco host or a Juniper host or Arista or whomever, it has to basically recognize the optic and configure it properly. Most routers today don't know anything about coherent and all the parameters that have to be set. And so that's a growing body of knowledge in the IP world.
And the standards are being defined and plugging holes. So I would say the standard initially came out with a lot of holes in it, like, this parameter could have a number of different values, but every host decided that there would be a different default. And so the behavior was different on startup for every different host. That sort of clarity is improving with the standard over time to the point that we will get the plug and play. We're not quite there yet, but we're gonna get there.
Got it. And maybe just pivoting back to the hyperscalers, obviously the use case is a little bit different. DCI-centric.
Mm-hmm.
It seems to be somewhat of a standard sort of operating to the market here for pluggables. You mentioned earlier that you're willing to cannibalize your old business. So how has that played out from an ASP perspective?
Mm-hmm.
In the market today, and what does that mean for, you know, the DCI market holistically going forward as pluggables obviously are kind of the driving.
Mm-hmm.
Form factor.
Yep.
Going forward.
Yeah. So one thing, just to hearken back to the interoperability question, hyperscalers are using these pluggables in an interoperable fashion today. They have driven that standard very aggressively. And where the standards had holes, the hyperscalers basically worked individually with vendors and said, "Plug the hole. Make it work like this." And so we deploy today in all of the hyperscalers, and we know other suppliers are deploying in those environments as well. And we know we're plugged into Cisco and non-Cisco hosts. That's a done deal for hyperscalers today. In terms of cannibalizing the transponder market, we're about a 10% player in North American markets in optical.
And so we made a very conscious decision when we were acquiring Acacia that we'd be willing to cannibalize some portion of our optical business in order to get a much greater share of this business. And we, I think we've succeeded with that. But the ASPs for a transponder at 400 gig, a transponder at 400 gig are much greater than the ASPs for a pluggable at 400 gig. So if you looked at the sort of the total optical market, and I think at least Cignal AI has done some analysis on this, and you look at the growth rate in the optical market for metro, the growth rate was single digits, like 5%-8% typically. That market is effectively flattening out to declining as a result of DCO Optics coming in because the ASPs are lower for the DCO Optic.
We're happy to have that 'cause in aggregate, we're getting much more business, and much more, much more share. But no question we're cannibalizing what was the profit pool of the optical business, the transponder.
Right. So I mean, I guess maybe to paraphrase, if I look at the industry data, whether it's Cignal AI or other third-party vendors, it generally looks like the metro market's not growing.
That's right. That's right.
If you disaggregate and peel back the onion, effectively it's a substitute or product dynamic where you've made the conscious decision three years ago through the Acacia acquisition to kind of target that market and effectively have a longer-term perspective on where you position.
Exactly. Exactly. And the other thing I would say, and the other rationale for us was if you look at the optical market, you can name pretty quickly like 10 players in the optical market, including some regional players. We wanted to basically move to a market where we could be the leader and where there'd be fewer players competing. And I think there's relatively few players that are playing in the DCO coherent market.
Got it. And maybe just can you remind us about on the Acacia deal? I remember at the time, I think, was it what percentage of their business was ex-Cisco? And where does that stand today?
At the time we did the deal, roughly 20% of Acacia's business was Cisco. 80% was not Cisco. And that 80%, largely speaking, were Cisco competitors. And we made a conscious decision at the time of the deal that we would keep those competitors. We'd keep that business. That's very different than what we've done with almost every other deal. We did a deal a few, probably 10 years earlier. We acquired a company called CoreOptics, and it was a similar dynamic. They had maybe 20% of their business was Cisco, 80% was competitors. And I had the pleasure of going to all those competitors and telling them, "We no longer want your business, so please go find another source." And that's usually what we do.
In this case, in part because it's a semiconductor play and semiconductor costs are determined by volumes, we said, "Let's keep that third-party business so that we can keep the costs attractive for ourselves," and so that, that created some structural issues for Cisco. We had to, we had to effectively create an organization that had a firewall between Acacia sales and Cisco sales. I own the Acacia business, and we sell to my competitors, but I don't, I don't get to see pricing that's going to my competitors. I don't see even what products we're selling to our competitors. I see an aggregated view. So we have very strict high integrity firewalls. We've maintained all of those customers. So we are selling to guys like ZTE that's public. We're selling to other optical vendors. We're selling to other switch and router vendors.
We treat them like customers. They really do get treated like customers. If you pay attention, you'll see there's an Acacia-branded website. There's a, you'll see on Twitter, there's Acacia tweets that come out. We've maintained the Acacia brand largely to serve those third-party customers.
You need that volume though to maintain margin effectively.
We need that volume, but I would say it has become a much smaller portion of the business, and the reason is that most of the business is hyperscalers by volume. Our business has shifted from selling to guys like ZTE to, in the last couple of years, hyperscalers dominate the business.
Got it. So when you think about the hyperscaler appetite for these products, how do I phrase this? Is there a rule of thumb to think about how these deployments happen? These data center deployments happened at the hyperscalers, and then what that would mean for your business within that hyperscaler environment? I know there's lots of ratios out there for different parts of the data center network. Just trying to get a sense for how we should think about how that affects your underlying business if possible.
I don't know if there's a rule of thumb. I would say AI, there's no question that AI is driving huge demands for these guys right now. We've just not experienced upticks in demand like we have seen in the last few quarters, all driven by AI. There seems to be just this relentless demand for capacity. These guys are putting in massive capacity. I would say the unfortunate rule of thumb is that the hyperscalers are pathetically bad in forecasting, whether it's up or down. And so we just kind of have to have a very flexible supply chain, and then deal with the ups and downs.
But there's no specific rule of thumb in terms of like, what's happening inside the data center versus what's happening outside the data center because some of these guys are building networked data centers to deal with their AI traffic. Others are keeping it inside the data center.
So just a distributed environment, obviously a distributed environment would help you immensely.
Right.
Given maybe some of the power constraints in the marketplace today, have you thought about what that might mean for, you know, the DCI opportunity given sort of a distributed data center footprint?
Yeah. There's no question that it's expanding as a result of that. I think it's pretty well known. Microsoft kind of distributes their data centers within a campus today. So already we're seeing like massive, massive impact in the inter-data center traffic as a result of that. And others are following suit there. And it's largely driven for others by power constraints where they're fundamentally, they just can't get more power at the current site. I can speak for Cisco. Cisco, we've been told in San Jose there's no more power available for Cisco. So we've got to shut down labs or move labs out of San Jose 'cause there's just no more power for us. So we know our customers are dealing with that as well.
What that drives them to now think about is distributing data centers not only within a campus but across a region or even across states. That does have a positive impact for the optical business.
You mentioned earlier the business limitations. We've moved from short haul to medium haul to potentially long haul.
Yeah.
So as these data centers become more geographically dispersed, the current portfolio can handle that sort of dynamic today or?
So I would say up to about 1,000 km, no issue. If it goes beyond 1,000 km, then I think the there are more significant constraints. So we offer, for instance, today, no problem for 400 gig up to about 1,000 km. The 800 gig that we introduced will go up to about 1,000 km. Again, it's fiber, depending on fiber type and whatnot, but about 1,000 km is a safe number. We've introduced a 400 gig long-haul that can go about 3,000 km. So if a customer.
Works for the data for the hyperscalers though, right?
All of them are deploying 400 gig today. To the extent they wanted to move to 800 gig, they're constrained to 1,000 km today. They'd have to do that math and decide what really makes sense for them. The trade-off might be, "Well, I need to go more than 1,000 km, so I'm constrained to 400 gig, or I stay within 1,000 km and I can go to 800 gig." The next jump will be to 1.6T.
Since you're a guru on this, I'm gonna ask you about sort of the competitive landscape in the broadly defined optical transceiver market between U.S. and Asian.
Mm-hmm.
Participants.
Yeah.
How do you see that market playing out? It seems to be there's like two and a half players here in the States, I don't know, five, six, seven, eight players overseas. Seems like you mentioned there's this insatiable demand.
Mm-hmm.
For capacity. How do you from your [audio distortion]?
So inside the data center for the transceivers, yeah, I'd say China plays a much, much more significant role there. For some of the hyperscalers, I think they're intimately tied with China suppliers. All of the hyperscalers, in my view, have a desire to diversify themselves. They've always had that desire, but I think there's an increased desire for not only technical reasons but now political reasons as well. It doesn't mean they're gonna abandon their China suppliers. They're very happy with those China suppliers. Ideally, what they would like to get is the China supplier's price point at the U.S. supplier's manufacturing facility. That's not gonna happen, but that's ideally what they'd like to get. So they sort of set the price point with their China suppliers, then they come to the U.S. suppliers and say, "Hey, can you meet that price point?
We'd like to have your diversity." So, we benefit from the fact that we do design in the U.S. We don't do manufacturing in the U.S., but we don't manufacture in China either. So, there are, I would say the big players InnoLight is one of the big players that we see inside the data center. Coherent plays very significantly inside the data center. There's, you know, five or six other players that play sporadically across the hyperscalers. We play in the hyperscalers as well as the commercial and enterprise markets as well.
You mentioned they want U.S. manufacturing capability, but with the margin or the price structure.
Yeah.
From China. So how do you think about the sustainability with hyperscalers being the big driver here of those margin structures? Like, InnoLight has relatively good gross margins.
Mm-hmm.
For basically being an assembler.
Yeah.
Effectively.
Yeah.
Right? So is that sustainable or do you think the margin structure of the industry, that particular industry needs to come down?
I think it's a challenge when you look at trying to do all of the internal design and manufacturing. If you're just doing the assembly, then it's basically, you know, it's either automation or low-cost labor that you're talking about, but you don't have the burden of a huge R&D investment, $100 million, you know, fab.
Right.
Foundry fee. So you gotta look at that balance and decide where you're gonna play. And in part for that reason, we have approached the market in a number of different ways. We sell components into the market in some cases to leverage our R&D investment where we don't think we might be competitive on a module. We'll sell components. In some cases, we sell components to a module supplier. We'll buy that module back, but we've got our technology in there at our cost, not their cost. And then we can resell that to other markets and do very well. And in some cases, we still OEM. So we look at the market in different ways depending on which customer we're trying to attack.
Got it. But is it fair to say that for you, I mean, if the margin structure of the industry is low, I mean, I would imagine these are below corporate Cisco gross margin offerings.
We look at whether it's.
What's your way to frame it?
Our optical business or our cable business or our routing business or our Optics business. We look at it as a portfolio play. Not every business in Cisco is judged by, "Are you gonna be accretive to Cisco?" 'Cause we would, the portfolio would fall apart at that point. So every business we look at, we basically look at a portfolio play. We look at even within the portfolio what segments we're gonna compete in. I don't sell Optical Systems into China as an example, but I sell Optical Systems elsewhere. I'll go sell Optical Systems into India. We'd make decisions about the portfolio kind of on a segment basis and on a portfolio basis where we're gonna be playing.
Got it. Final question. Any gating factors if we have a conversation a year from now, whether it's tariffs, power, cost curve, foundry capacity, laser capacity, you know, chip capacity? What are the gating factors over the next 12 months as you see it for your business and the industry at large?
Tariffs become a concern depending on how the tariffs are leveraged and or levied and where. We've managed through tariffs before. I think Cisco's done an amazing job of de-risking the supply chain as a result of the last tariffs that we went through. We've de-risked China in many ways, but we're not fully. There's still small exposure to China. I think things like, you know, China announced that they're not gonna supply some rare earth materials today. Some of those rare earth materials.
So.
I'm sorry?
Like a Gartner issue?
It's like germanium, for instance, right? So, you know, some of those rare earth materials we rely on, I don't know yet where in the supply chain that becomes an issue, so there are things like factors like that that could impact our business. We try to get ahead of those as much as possible. In some cases, we'll find alternate sources. In some cases, we'll buy up supply until things settle down. A lot of these are political issues, not technical, but I think Cisco's got a massive effort on supply chain to get ahead of those issues and make sure that we have de-risked those.
Got it. I'd like to give the executive a chance to kind of discuss what is not understood about your business.
Mm-hmm.
Particularly given the complexity of your business vis-à-vis other parts of the Cisco portfolio. Today, as we sit here, what do you think is misunderstood about Optics, Optical Systems, ZR, Acacia versus, you know, the business that you've kind of pulled together? I don't think people understand the size of the business from a revenue perspective to start.
Yeah. I mean, we're collectively, across those businesses, we're a multi-billion dollar business for Cisco. Cisco doesn't go out and wave the flag on Optics and say, you know, we're an Optics company. But I think we may be, if not the biggest, one of the biggest Optics suppliers on earth. And the reason is we sell Optics to all of our customers. We sell Optics to our enterprise customers, our commercial customers, service provider, hyperscalers. So we sell across all those segments. And most of those segments, not taking the hyperscalers aside, most of those segments buy Optics from Cisco because it's a safe thing to do. You know, they're not in the business of trying to understand every little detail about the optic.
They're in the business of, you know, making sure that their data center runs so that they can run their applications for their customers. They're gonna do the safe thing and buy Optics from Cisco. So we have a very, very substantial Optics business, and it's actually quite a healthy business as well, a very healthy business, in fact, from a margin perspective. The other thing I would say is, Acacia, I think, has been a marvelous acquisition for us. You know, Cisco has done well over 200 acquisitions. Some are very successful. Others are less successful. I'd put Acacia in the highly successful category of acquisitions that Cisco's completed.
Got it. So I think we're out of time, Bill. Thank you for your time.
Thank you, David.
Informative. And if anyone has any questions, please feel free to reach out to Sami and Ryan. Thank you again, everyone, and we'll see you soon.
Thanks so much.