Great, I think we're going to get started. Good morning, everyone. Thank you for joining. I am David Vogt. I'm the UBS Enterprise Hardware and Networking Analyst, and we're excited to have Cisco with us today. From the company, we have Bill Gartner, SVP and General Manager of Optical Systems and Optics Group. In the audience is Sami Badri, new head of investor relations, for those that don't know Sami. So, just before we get started, Bill, I'm just going to read a quick disclosure. So UBS, we might mention individual companies here that we have a relationship with, so if anyone has any concerns or comments, please see the UBS Research website disclosures at www.ubs/disclosures. I think we're good on your end, so why don't we get started?
We thought we'd start with kind of maybe discussing Bill's background, his purview, and the optical business. Maybe we'll just start there, and then we'll jump off from there.
Sure. Thanks, and thanks for having me. So I'm responsible for optical systems and optics at Cisco, and let me just frame those two, because those terms are interchanged at times, and I try to keep them separate. When we talk about optics, we're talking about the transceivers that we sell with routers and switches for use in Cisco routers and switches or in third-party routers and switches. It's a very substantial business for Cisco. We have a portfolio that ranges from 1 gig all the way to 800 gig, serving all market segments, whether it's campus, enterprise, service provider, data center, or hyperscaler. That world is really characterized by two things.
One is, you can think of it as sort of what's happening inside the four walls of a data center or a central office if you're in a telco environment. And it's characterized by the fact that when you add a new router or switch, you pull new fiber to every port on that router or switch, and you terminate that into a transceiver, and every port gets one transceiver, and every fiber has one signal on it. That's one attribute. And the other attribute is that the distances are short, and in this world, short means less than 10 kilometers. And for that, we've historically used a technology called direct detect, which basically means if the light's on, it's a one; if the light's off, it's a zero.
That technology is getting a bit more sophisticated as we go to 400 gig and beyond, but that's the world we've been in the past. And I'm fortunate in that I can actually carry a piece of my portfolio around with me. The transceivers that I'm talking about look something like this. This is a typical transceiver. This is an 800 gig transceiver. I'll come back to this. Once you leave the data center or the central office, now the problem is very different because now you have to send the signal across a city or across a country, or even across continents, between continents on a subsea line.
And in that world, the fiber is in the ground, and we can't ask a customer to pull new fiber between routers and switches that are added between New York and L.A., for instance. We have to use the fiber that's in the ground, which means that we have to put many signals on one fiber. Fiber is a very precious and scarce resource, so we can't just pull new fiber to every port. We have to put many signals on the fiber that's in the ground, and that technology is called DWDM, dense wavelength division multiplexing. Basically, many different colors of light on the fiber. And the other attribute is that the distances are long, and long in this world means anywhere from several hundred kilometers to thousands of kilometers.
For that technology, that's a much more sophisticated technology than Direct Detect, lights on to one, lights off is a zero. We have to use very sophisticated modulation schemes that are borrowed from the RF world, the mobility world. In doing that, we have to basically use chassis-based solutions, so these big chassis that house line cards that do all this heavy lifting. So in the data center, we're using transceivers like this. Outside the data center, in optical systems, we're using chassis-based solutions. Now, the one really interesting thing that's happening is really what motivated our acquisition of Acacia just about three years ago. Cisco wrote a $4.5 billion check for Acacia, so it was a pretty meaningful acquisition.
One of the interesting technology trends is that we can now take what was in that line card in a chassis supporting a Metro application or a Long-Haul application, and put that technology into a pluggable optic. Looks just like this. And that pluggable optic can go directly into a router, and we can eliminate the chassis. We can save power, we can save space, and we can save cost. And that's part of our Routed Optical Networking architecture that I'll talk a little bit more about. And we're really excited about that technology because we think that's a game changer for the industry. It will change the landscape of the optical industry, but importantly, it simplifies networks for our customers. It reduces their power consumption needs, it reduces their space needs, and it simplifies networks, so we're really excited about that.
So in a nutshell, I have the optics business, which are the transceivers we sell for routers and switches. We ship over 10 million of those a year. I have the optical systems business, and I'm also responsible for Acacia, which is developing this new really, really cool technology for us.
So that's a great overview. So maybe go back to Acacia and on the systems side as well. Why, why did Cisco think it was strategically, you know, smart acquisition to kind of get into this sort of integration or this consolidation-
Mm-hmm.
- of the technology sort of platform, and as it pertains to sort of the core portfolio as it was configured before the Acacia deal?
Yeah.
Like, what drove that strategic decision?
...So, one thing I can say that, that did not really drive it is we, we did not acquire Acacia so that we could shore up our optical business per se, though that was sort of a by-product, an important by-product, it was not the main driver. The main driver was we saw this technology where we could take what was in the optical systems world and effectively make it part of the routing sales proposition for customers. So we sell this pluggable as part of a router. That effectively helps our routing business. If we're wildly successful with this, we'll cannibalize the optical business because the transponders, which represent about 70% of the optical business across the industry, will be replaced by these pluggables over time.
So we were willing to make a decision to cannibalize the optical business in order to increase the TAM and increase our market share of the routing business. So the main motivation for Acacia was really this, this transition from a chassis-based line card in the optical systems world into a pluggable that would be part of our routing portfolio, and that was the driver for the Acacia acquisition.
Got it. And so as we sit here today, I mean, you know, I think Chuck and the rest of the team talk about the success that you've had. What is your sense in terms of how this replacement cycle or this cannibalization cycle, I don't want to use that word, looks like over the next couple of years? I know there's been a lot of talk about growth in pluggables-
Mm-hmm.
The steep part of the S-curve is, you know, in front of us. But how do you think about the timing and how this plays out over the next couple of years?
So, first of all, I would say Cisco came out with this approach as really coincident with the acquisition of Acacia and said: We think the industry is going to move in this direction, and we're motivated to move it in that direction. We got a lot of resistance from, from our peers in the industry who said, "No, no, no, this is only going to capture, you know, a tiny, tiny bit of the market." Already, we are now seeing that this is effectively being used in virtually all data center interconnect applications for hyperscalers, for instance, and in many, many metro applications. And this, this pluggable, when it's supporting the coherent optical DWDM solution, can support about 1,000 kilometers today. That will grow in the next 12 months to support any long-haul application as well at 400 gig.
So our belief is that over time, the transponder business, that's classically part of the optical systems world, will be cannibalized by these pluggables. That's a good thing for our customers. It simplifies their life, it reduces their costs. We think that's a natural thing to do, but it will take time. It's an architectural transition for customers. They have to think about not only the technology moving from one domain to another, but for many customers who might have an optical operations team and an IP operations team, there are operations issues that they'll need to consider as well. We help them with that through things like automation solutions, so we help overcome some of those obstacles in the operations transition. But I think this is probably a three-five-year transition for customers. This is not something that's happening overnight.
We doubled the number of customers that are deploying this technology in the last year. We're now deploying in over 80 customers right now, including three major hyperscalers. So we have pretty deep belief that this is, this is a train that's moving, and that train is not going to stop. The other thing I would say is that while we encountered a lot of resistance from competitors early on, if you survey the other players in this space, you will, you will hear Cisco talk about Routed Optical Networking. You'll now hear others talk about things like coherent routing and their version of this. But I think everybody has acknowledged that this pluggable is really here to stay, and it's going to play an increasingly larger role in network applications.
Got it. And so you mentioned it's allowed you to, you know, grow or expand your relationship with hyperscalers. You mentioned three of the four. I know Chuck and the rest of them have talked about, you know, the expanding relationship with hyperscalers or the web scalers.
Mm-hmm.
When you think about the needs of the hyperscalers today, how does sort of the optics or the optical systems or the pluggables play into what the hyperscalers are doing today? Like, what need or, or what's the solution that you are solving for?
Right
... effectively with this?
So I'd say, first of all, while we have a deep belief that optical systems are going to be cannibalized by pluggables, there is still a role for optical systems, even in the end game, and that role is typically for the very high-performance applications and for the long, very long-distance applications. So higher, higher bit rate and very long-distance applications will still demand a chassis-based solution. And we have hyperscaler customers who, for data center interconnect, are deploying this technology today in massive volumes. And we have other hyperscaler customers who say: Look, I'm building out a metro network, for instance. I want the most, the highest performance I can possibly get in that metro network because my needs are so extraordinary that 400 gig may not be adequate.
On the system side, we offer today a technology that delivers 1.2 terabits on a wavelength. This is 400 gig on a wavelength. So for many, many, many metro and DCI applications, this is going to be adequate. For the most demanding applications, the customers may still rely on a chassis-based solution. So you can think of it as sort of two vectors that we invest in. One is pluggables and trying to drive more performance, lower cost, lower power into that solution to expand the market for that... but we're also driving higher performance through chassis-based solutions for those customers that may demand that.
Got it. And since you just mentioned sort of the optical system, some of them are, you know, deployed at 1.2. There's 1.6 out there from a competitor. How do you see that optical systems market developing, evolving-
Yeah
-over time?
First of all, I would say that competitor is not shipping yet.
Announced.
They've announced.
I should have said announced.
Announced, and we're shipping. That, but I think that's the dynamic of the market. The optical systems market is a leapfrog game, and it has been a leapfrog game since my early days in this market. My first generation DWDM system was 20 gigabits of total system capacity, and now we're putting 400 gigabits on a single wavelength and delivering tens of terabits of total system capacity. And the game from way back when, and continues to be, that Cisco will announce and deliver a product. We delivered 600 gig. Our competitors came out with 800 gig. We delivered now 1.2 terabits. Our competitors come out with 1.6 terabits. It's a leapfrog game. That said, those are the high, high-performance applications. We are approaching fundamental limits of physics.
We're approaching fundamental limits where we just can't get any more capacity on the fiber. And so the game then will become much more about driving down power, driving down cost, driving down space, which is where the pluggable plays a much more important role. So we're kind of approaching what's known in the-
Shannon
physics world as Shannon's limit.
Yeah.
We're maybe a generation away from the point where there's just no practical gain to be had there. We've been playing a leapfrog game on technology for many years. We're kind of at the end of that cycle for real. There's just physics limits there. And the game is going to become: who can drive lower cost and lower power?
Got it. And so when you talk to customers today, you know, you mentioned power consumption and cost. You know, obviously, they're well aware of the limits of Shannon's limit. And so where do you see that part of the business for Cisco on the system side, playing out over the next couple of years? I would imagine most of the focus is on optics, right, today.
Mm-hmm.
And systems will be less of a growth driver over the, you know, multi-year period. Is that a fair way to characterize on how that plays out?
I think it is in a sense that we are perfectly willing to cannibalize our transponder business in favor of pluggables that really help the routing business.
Right.
So in that sense, we will see a diminished optical business over a, you know, five-10-year period. I would expect the optical business would diminish. But that said, there is always going to be the most demanding applications, like a subsea application.
Right.
I don't see Subsea being served with a pluggable in our lifetime.
Right.
Subsea is always going to require the most demanding performance we can possibly deliver, and that will likely be in a chassis-based solution for quite some time.
Got it.
We'll continue to invest in that space.
Got it. Maybe just sticking with sort of the web-scale relationships, there's a lot of discussion about what Cisco brings to bear by customer, suite of products, suite of solutions, optical systems.
Mm-hmm.
Optics are one part of it. Can you kind of help us understand, you know, when you think about the 500 million orders to date that had been announced, I guess, the previous quarter-
Mm-hmm.
-and then the commentary on the last call about, you know, we have line of sight, I think, on roughly $1 billion-plus orders-
Mm-hmm
-in 2025. How do you think about all the different moving pieces within your purview?
Mm-hmm
and the company in terms of what that means for a customer?
Yeah.
Right, so it could be a complete system, it could be an optical module. Like, how, how are you thinking about how you're addressing those potential orders?
Yeah, that really does cross the Cisco portfolio in a pretty significant way. As I think you know, we have pretty significant investment in Silicon One, which is a technology that we acquired from a company called Leaba. That has really penetrated our routing portfolio in a significant way, increasingly moving into our data center portfolio and ultimately into the enterprise portfolio as well. And with that Silicon One technology, we are able to work very closely with the most demanding customers, who are the hyperscalers, and customize solutions for them that really address their specific needs. We are also delivering a more generic solution to the broader market. Our Cisco 8000 Series of router is the fastest-growing router in history. And that's a basic router that supports the generic core type applications.
So in working with the hyperscalers in particular, we can customize a silicon solution for them. That then takes the form of a switch or a router that becomes part of the AI infrastructure. We believe Ethernet will be an important-
Right
-transition that takes place there, that will favor a Cisco solution over a proprietary InfiniBand solution. Optics, I think, represents a pretty significant portion of the AI market as well. And just one quick tangent on why I have that belief. At 10 gig, when we look at the total BOM for a customer between a switch or a router and an optic, at 10 gig, the optic represented about 10% of the BOM. 90% was in the port of the switch or the router. At 100 gig, the optic represents about 50% of the BOM, and at 400 gig, the optic is actually more than 50% of the BOM today. That may drop, but it's not going to drop below 50%, in my view. So proportionately-
At 800 gig?
I'm sorry?
At 800 gig.
800 gig will certainly be more than 50%-
Right
of the BOM. And so proportionately, the optic has become a bigger part of the customer spend over time, and I think if you look at the AI infrastructure and the interconnection, the optic is actually a big part of that infrastructure spend. So between our Silicon One advantages, which we are successfully positioning with all the hyperscalers right now, the encapsulation of that in our routing or switching portfolio, the optics that we bring to bear, and then the optical systems, which then allow those customers to interconnect data centers. So we've already seen demand arising from interconnection of hyperscaler data centers that's increasing their demand, entirely driven by AI applications. And they'll just, they'll just tell us, "Look, this is unexpected demand that we're seeing that's driven by the AI application." Interconnecting data centers, not inside the data center.
It really crosses the entire Cisco portfolio. I think we are super well-positioned to address that full complement of needs.
Can I just ask a follow-up? So historically, in the last migration, in terms of a move intra data center, there was a big share shift dynamic, and I think the company's been public about, you know, obviously having to rethink about go-to-market, reengineering product. Today, it sounds like you're offering what the customer wants in any capacity or configuration.
Yeah.
Is that a fair way to say it?
Yeah, I mean-
Customer just wants silicon, it's just Silicon One.
Yeah.
They want, you know, complete total system solution that's brought to bear. Is that the right way to think about go-to-market?
I think, I think just to go back and put a little bit finer point on that, David, it... You know, probably 10 years ago, and Chuck has been very open about this, we basically approached the hyperscalers and said, "Here's our ASR 9K product, and, you know, it comes in this flavor or this flavor. Which one would you like?" And the hyperscalers were like: "I really don't want that. I want, I want this piece of it. I don't really want the whole, the whole integrated solution, software, hardware, services, everything else that Cisco would package." And we just were not, we were not in that game. We were not really willing to do that at that time. I think we learned some hard lessons, and in actually December of 2019, we announced a component business model.
And the idea behind that component business model is we're going to meet customers where they want to be met, and that means if a customer wants to buy our silicon, we'll sell them our silicon. If a customer wants to buy just our optics and not any of our systems, we'll sell them our optics. If they want to buy ASICs that go inside these optics, we'll sell them those ASICs. If they want to buy just our hardware platform and put their own software on it, we'll do that. So we have adopted a component model that says: Look, whatever a customer wants to consume, we're gonna offer that to them in the way they want to consume it. And that means that they won't necessarily buy the fully integrated solution, but they might buy a piece of it.
The interesting thing that's happened is, I think, expressing that willingness and going to the customers and saying, "Look, you want to buy silicon, buy silicon. You want to buy software, buy software." That's opened doors in a dramatic way with especially the hyperscalers, who are really the consumers of-
Right
That type of technology. Increasingly, they come back and say, "Look, I really do want to buy the integrated solution, but I'm happy to know that I can get the component if I want to." So it's actually significantly expanded our systems business with the hyperscalers. Even though we're still offering them components, really what they're buying in many cases are integrated solutions.
Integrated solutions. And I know... So we get questions on this all the time over the last couple of quarters, in terms of your AI-related exposure. We talked about the $500 million product orders-
Mm-hmm.
Line of sight to $1 billion. I think there's a lot of confusion or maybe debate in the investment community about what should be characterized as AI, right?
Mm-hmm.
To your point, you've seen, you know, a customer will say to you, "We've seen a massive uptick in traffic across the DCI.
Mm-hmm.
That was from an AI application.
Mm-hmm.
How do you think about what is AI?
Yeah
effectively within the portfolio? Because it could be, you know, legacy workload related-
Yeah
at points. It could be AI workload related.
Yeah
at points. So how do you think about that?
So, I would say, I think, quite honestly, there are blurred boundaries there. There are some things that are quite distinct, like the back end of an AI network is going to be, you know-
AI, right
pure AI.
Exactly.
There's not, not gonna be a debate about that. That will include things like routers and switches and optics. When you get into the WAN side and where they're now deploying more routers and more optical systems because of AI workloads-
Because of more traffic. Because of more traffic, right?
... That's blurry, because I'm sure there's mixed traffic in there. But what we do is take the signal from the customer saying, "This is what's driving our demand," and we can attribute it to AI, but you know, technically, it could be carrying legacy traffic as well.
Got it.
I think, to be fair, there are going to be things that are very black and white and some things that are just in the gray zone.
Got it. And so, you mentioned sort of this Ethernet and InfiniBand, and we didn't touch on this earlier, but maybe just would love to get your perspective on this. I know Chuck's talked about, you know, an Ethernet switch market-
Mm-hmm
- that's going to be AI-related or AI/ML-related of, you know, I think the number we cited is $10 billion.
Mm-hmm.
Obviously, there's going to be share for InfiniBand, share for Ethernet. You are a founding member of the Ultra Ethernet Consortium.
Right.
Maybe just kind of help us think about, you know, why InfiniBand, in your view, isn't sort of going to be the sole provider?
Mm-hmm
- of interconnects and the dominant platform effectively over the next three-five years. Why Ethernet is-
Sure
... you know, having, well, at least according to the specific data that we've looked at, Ethernet's going to have a relatively healthy share of that market.
Okay. I'd like to actually expand your question-
Sure, go ahead.
... to optical as well. So I think Cisco has a long history and a deep belief and, and very high integrity belief around open systems being the right answer for our customers, standards-based open solutions. Even when, even in those cases where a proprietary solution might have provided us with an advantage, we have argued for open solutions. And, and the reason is that the industry benefits with open solutions. Our customers certainly benefit. They get choice that they would not have otherwise, when there's a proprietary solution. The supply chain generally benefits because the supply chain can consolidate around one answer rather than many discrete proprietary solutions. And, and generally, the industry expands.
We've seen volumes in Ethernet that we couldn't imagine if all of the technology choices that existed 20 years ago had developed on their own, we just would not have seen that level of volume. So we believe it's a good thing for the industry. We believe it's a good thing for our customers. I want to just expand on InfiniBand. InfiniBand, I think, was kind of a don't care for the industry for a long time because the volumes weren't that great. There just wasn't that big a demand. Now that AI is kind of exploding, it's become a much more critical issue for our customers to be able to choose, have choice, and to see that there's an alternative to InfiniBand.
Cisco is very much behind that with the Ethernet Alliance, as well as many of our peers in the industry who believe, again, that Ethernet is, is the right, the right thing to be centering on here as a standard. We will have to improve the Ethernet performance in some areas, but that's certainly within sight. I just want to expand on that a bit within optical. The DWDM world is probably the last bastion of networking that is proprietary. If you buy a DWDM system from Cisco, it's bookended. You can only buy the transponders from Cisco. If you buy a DWDM system from Ciena or Infinera, ADVA, same thing. That customer is locked in for effectively five-10 years.
We've been advocates of opening that DWDM interface for years, and making that a standard interface where you could have interoperability, or you could provide the customer with choice at some level. The pluggable optic that we're now advocating, 400G ZR, ZR+, is complying with open standards, and we're very, very happy about that. We're driving those standards. We want to see this become open. Cisco acquired Acacia. We could say, like: Hey, we've got the technology that many others don't. Most of our competitors don't have this technology. We could sit on this, make it proprietary, but we believe that opening this up is good for our customers, and it's good for the industry.
It will, it will help deployment, and there's no question that we've seen customers adopt this technology because it has the promise of being open for them. So they will have choice that they don't have with the alternative chassis-based solution that is a proprietary solution, one vendor only. So whether it's InfiniBand or whether it's DWDM, we will take a stand that open is good for the industry, it's good for our customers.
Can I ask a really... Maybe this is a naive question. So I think at the optics show, there was a ton of demonstration of interoperability on pluggables-
Yeah.
- different systems, different pluggables-
Yep. Mm-hmm.
from your peers
Yeah
... across the landscape. Does that raise the risk competitively of white box? You know, white box having a bigger footprint in this particular market, where historically it's been, not relegated, but primarily more on the switching side of the house.
So the demonstration that you're referring to was actually the 400G ZR-
Yep.
demonstration, with a bunch of optics vendors and routing vendors, Juniper-
ADVA.
ADVA-
You guys
... Cisco, Ciena, all participated in that. And we mixed and matched optical system, optical solutions as well as, as routing solutions, as well as pluggables. That was a good demonstration. The, the industry is not mature enough yet to make that a reality, but it was certainly directionally where we want to go. I don't think it really changes the white box dynamic, especially at the optical layer. The optical business is generally lower, pretty low volume, so-
Right. Mm-hmm
... most of our customers are not looking for white box solutions there. It's just not, not that big a volume for them, compared with something like a server or a switch that might be in a data center, where there's compelling economics. And frankly, I would say if, if a white box industry emerged for optical line systems, I'd be very happy to see that. That wouldn't be a bad thing for the industry.
Got it. And so maybe one final question. So you talked about Routed Optical Networking and competitors-
Mm-hmm
having coherent-
Mm-hmm
rebranded names
Yeah
- or different flavors of it. You know, as a new entrant from a technology perspective, I would imagine that has... It re-raises the risk of potential margin compression as you collapse layers, right?
Mm-hmm.
You need fewer, you know, higher-end pieces of equipment. Is that a longer-term dynamic that you're thinking about and that maybe would affect not just you, but your competitors across the street?
I'm not sure it's a margin compression issue so much as a top line-
Volume? Volume. Okay.
Top-line issue. Like, if you look at the pure play optical vendors, this is a threat for them. This is an existential threat because the transponder is going to cost more. It's a top-line, much, much higher revenue. It represents 70% of the business. That's going to be replaced with a lower-cost element-
Right
... that could be substituted with a third party. So that's, that's a pretty existential risk for a standalone, pure play optical vendor. Cisco has optical and routing, so we can make that trade pretty comfortably and say: Look, we're going to basically move something from the left pocket to the right pocket. It will be lower revenue, but we're going to capture that business. We're going to capture that business as part of our routing portfolio. So, we're okay with that. We can make that trade.
Got it. And then, you know, I think we're running out of time, but I just want to give you an opportunity to touch on: is there anything that we didn't cover that you think is topical, germane to kind of the discussion today, that maybe came up in meetings today or in, in conversations?
The one thing I would say is Routed Optical Networking architecture is an architectural shift that really is driven by these pluggables. We've got now 80+ customers deploying this. We are super excited about this, and I think in five years, you're going to see this as a dominant architecture in the network.
Got it. Well, great. Thank you.
Thank you.
Informative, as usual. Thank you. Thank you, Sami, and thank you, everyone.
Thanks, everybody.
Thanks, Bill.