Cisco Systems, Inc. (NASDAQ:CSCO) Barclays 2022 Global Technology, Media and Telecommunications Conference Call December 7, 2022 3:10 PM ET
Bill Gartner - SVP, GM Optical Systems and Optics Group
Conference Call Participants
Tim Long - Barclays
Hello, everybody. Thank you for joining us here for this session with Cisco. Tim Long here, IT, hardware, comm equipment analyst at Barclays. Very happy to have Bill Gartner with us, SVP, General Manager, Optical Systems and Optics Business Unit. Looking forward to the discussion, pretty hot topic area for Cisco and for the industry.
So, I think Bill's going to read a safe harbor, and then, maybe after that if you wouldn't mind just kind of give us a little overview of your roles and responsibilities at the areas that you're covering and then we’ll get [right away] (ph).
Great. Thanks, Tim. First of all, thank you for having me. And before I start, I will be making some forward-looking statements that are subject to risk and uncertainties as outlined in our disclosures. Have I got all that right, Marilyn? Great. Okay.
My name is Bill Gartner, and I have responsibility for two businesses in Cisco that are related by the fact that they both rely on optical communications; one is the Optics business and the other is the Optical Systems business.
And you can think of the Optics business as the transceivers that we sell with routers and switches that find their home inside a data center or inside a central office or within a campus environment. Those transceivers are used typically to send optical signals on a fiber over relatively short distance, like ten kilometers. That's the Optics business. And we serve all markets with that, that includes the campus environments, enterprise, commercial, public sector, service provider and web.
And then, the other businesses, once you have to leave the data center and now send an optical signal across a city or across a country or even between continents, now it's a much more difficult problem to send that optical signal, and it requires much more sophisticated solution that is classically chassis based. It's a chassis that we have to sell for an optical system to carry these signals reliably over very long distances.
And the other thing that's unique about that world is that, in a data center, when you add a new router or switch, you pull new fiber to every port on that router or switch because you're inside. You can do that. When you leave the data center, now you're talking about crossing the Mississippi or crossing the Rockies, and you basically have to use the fiber that's in the ground. And so, we have to put lots and lots of signals on one fiber.
So, Optical Systems are what we use outside the data center or central office and Optics what we use inside. That's the two worlds that I have. They're very different businesses, very different business models, but they're related by common technologies.
Q - Tim Long
Great. Thank you for that. Good start. So, maybe across the two businesses, talk to us a little bit about kind of your priorities, looking out the next few years, obviously you've done the Acacia deal has been integrated, you got routed out optical networks. There's just a lot going on, right? So, maybe talk about two or three of your priorities, then we'll dig more into it?
So, on the Optical System side, we've just introduced a new optical layer platform called the NCS 1010 that offers some very innovative capabilities for customers to simplify operations. It runs IOS XR, which is our routing operating system. So, it's common for customers that have deployed our routers, and it supports CNL band. So, massive capacity. That's our -- you can think about as a layer zero solution. We've just launched that.
And the other key thing for the Optical Systems business that we have under development is leveraging something that Acacia announced, which is a new DSP supporting 1.2 terabits on a single wavelength, 1.2 terabits on a single wavelength. And we'll be trialing that in second half of next year and we'll have that available in fourth quarter of next year. Those are two key development areas for the Optical Systems business.
And then on the Optics side, we're I think still early stage on 400-gig deployments. So, a lot of effort in terms of getting 400-gig out there to our customers. We are also very focused on selling Cisco Optics, not only for Cisco routers and switches, but for third-party solutions as well. So, when customers want to consider Optics as a buying center and, say, they want to consolidate their optic spend, we want to be considered as an optic supplier.
And then I think the one thing that's crossing the systems world and the optics world is Acacia has a very significant innovation in something called a 400-gig ZR or ZR Plus, which is effectively taking what was classically delivered in a chassis as part of an optical system and putting that into a pluggable, a pluggable form factor. And that is a 400-gig ZR or ZR Plus. And that's part of our router optical networking architecture. So, that's an important thrust for really the Optical Systems business and the Optics business.
And if you give me a minute here, Tim, I actually brought some show and tell, I'm going to try to make that a little crisp for you, because I know you guys don't live in this world. This is a line card that goes into an optical system. We sell this to all the web players, service providers. This is -- this supports 1.2 terabits of capacity and it goes plugs into a chassis with a bunch of other line cards that plug into a chassis. So, for 1.2 terabits, the customer basically can put -- get two trunks, if you will, or wavelengths, and they can combine up to 1200-gig interfaces. So, this 1200 gig ports here, and out comes two 600 gig ports. That's where the 1.2 terabits comes from. This is part of an optical system.
And what we're doing with ZR pluggable is effectively now using -- taking what's in there, largely speaking and putting it in here. Now, it's not quite apples-to-apples. This is a 400-gig pluggable, that's 1.2 terabits. So, you'd need three of these to get to one of those. But from a cost, power, space perspective, this is way, way, way more efficient for customers than that.
And so, if you think about it, I own this business, which is part of Acacia. I own this business, which is part of our Optical Systems. I'm going to cannibalize a part of this business in order to make this business successful. And we're okay with that because we own both businesses, but we think there's more potential for this business over time. Does that help?
Yeah. I'm glad you didn't have to take an airplane to get here. That might have been a challenge.
By the way, my supply chain lead always is terrified when he sees me walking around carrying his stuff in a shopping bag. But we're not going to plug it into any customer's network.
So, maybe we'll start with this routed optical networks, ZR, ZR Plus. So, ZR, ZR Plus admittedly has been slow, right? Maybe talk to us a little bit about why it hasn't developed as quickly as most in the industry had expected? Why is that going to be different? And touch on the length power -- some of the key issue -- technical issues that need to be tackled or were tackled.
So, let me disagree with you on one point there, Tim. I think for the web players, ZR and ZR Plus -- not all the web players, but many of the web players are deploying it in massive volume. And that's characteristic, I think, of the web players. They are quick to adopt new technologies. They don't have a lot of overhead in terms of processes and operations to get in the way. They don't have a lot of legacy. So, when there's a new technology that offers significant benefits in terms of power, space, cost, they can jump on it very quickly. And so, we were at capacity for much of the last year trying to serve that segment of the market with ZR and ZR Plus.
For the service provider market, which is traditionally deploying like chassis-based solutions, there's -- we're on a journey that is not going to happen overnight in part because they've deployed these systems which have a lot of life left in them. And so, they're not going to jump to a new architecture overnight. There's also operations -- some operations differences that they have to accommodate in moving from a world of chassis-based solutions to a world where you plug this into a router. But we're seeing very good traction there. We've deployed to over 20 customers now. I'm very confident that over time -- this is going to be a five-year journey. This isn't going to happen overnight. But over time, this pluggable is going to replace that transponder in many, many applications and service provider markets.
So, I don't think the journey has gone any slower than we expected. I think we anticipate that for service providers, it is always a much slower transition to a new architecture. And this is a new architecture. It's not just sort of a new technology. And -- but I think we're on pace with where we expect it to be.
Okay. And part of the architecture is to remove the full optical system from the network?
So -- yeah, let me just outline kind of at a high level what routed optical networking architecture is about. Because there's a lot of FUD out there that my competitors love to throw out. There's some misinformation on that, too. Part of it is replacing this with this. That's part of routed -- and it's a big part of routed optical networking. Some think that's all it is. It's not simply that -- although this is where a lot of the CapEx and OpEx and power savings arises, is from this.
But routed optical networking really came about when we looked at the scale of routers, what's happening in ASICs that's allowing routers to scale so dramatically. It wasn't too long ago. When I came to Cisco -- I've been with Cisco 14 years. And when I came to Cisco, we had a 40-gig line card on a router, and it had 14 ASICs. And now we've got one ASIC, one ASIC that does 19 terabits of capacity. And that’s going to go to 25 to 50 over time, one ASIC instead of 14. And nominally, you can kind of think of it as the cost of an ASIC of the cost of an ASICs. So, every time like I take 14 down to seven, down to three, down to two, down to one, I'm getting cost savings, but I'm also packing much more capacity now into that one, so that's driven down the cost per bit on a router very, very dramatically over the last 10 years. It's also driven down the power per bit, because now we have one ASIC rather than a whole bunch of ASICs. So, the cost per bit on a router has come down dramatically over the last 10 years.
When we started building networks with an IP layer and a DWDM layer and sometimes an OTN layer, the motivation for that was that the router was the most expensive resource in the network. It was by far the most expensive thing in the network 20 years ago. And what we did is we -- as an industry is we basically built layers of the network to bypass routers whenever you needed to. So, if you had to go from A to B to C to D to E and you had some demands from A to E, it was very expensive to go through B, C and D to get to E. So, we went around B, C and D with an optical layer using things like ROADMs. And that made sense economically. That really made sense from a technical and an economic perspective. But now the cost of the router has come down so dramatically that it's actually more expensive to go around those routers than it is to go through them.
So that was -- that's one key insight that drove routed optical networking is it's no longer more cost effective to go around the router than it is to go through it. And in fact, what we did as an industry is we built a lot of these bypass wavelengths that have very little capacity on them. So, we can now take advantage of the IP layer, aggregate a whole bunch of demands and basically go through routers rather than around them. And so that simplifies the network in a very significant way because you can simplify the DWDM layer. It doesn't go away. To be clear, the DWDM layer is still there, but it's simpler. It can be much simpler. You can take advantage of these pluggable optics. And if you can take private line services like a T1 service or an OTN service and now put it on the IP layer with something called Private Line Emulation, you can take those private line services that were traditionally served with custom products, put that now on the IP layer, now you can get down to one layer in the network. Instead of having IP, OTN, DWDM, you can have just the IP layer. And that simplifies operations, it simplifies planning, it simplifies life cycle management.
So that all together, it's Private Line Emulation, it's the idea of rethinking how traffic moves through the network, it's pluggable optics, and now it's automating all of that with an automation infrastructure. It's really those things that make up the routed optical networking architecture.
Okay. Great. Maybe sticking on the system side for a little bit. I'd say if you look at Cisco's industry share or anything like that, it's not where it is in routing. Talk a little bit about owning a lot more of the IP and the optical layer, does that help new products like the NCS 1010. It's real catchy name you got there, that one. Just talk a little bit about kind of that vertical integration and what that can do for you even outside of routed optical networks for the traditional systems business?
So, let me say something you probably don't hear a lot out of somebody from Cisco is, I don't aspire to be number one in optical. That's not my goal, to be number one or number two in that market. The optical portfolio for Cisco is more of a portfolio play for Cisco. When customers want to buy optical and routing from one vendor and want basically an integrated solution, we're there for them. That's not to say we don't sell optical standalone because we do, but it's opportunistically that we go after those standalone plays. I don't have an objective to be number one in optical.
In fact, as I mentioned earlier, when we do -- when we replace this with this, this is the profit pool in optical. This is the most profitable part of the optical system. There's other elements like the ROADMs and things like that, that are really common infrastructure that don't go in with high margins. This is where the margins are, and we're going to replace it with this. And when we sell this, we're going to count it as part of our routing sale. So, effectively, I'm going to take down the Optical business in support of routing. And we have strength in routing that we're going to leverage.
So, this is very much a portfolio play where we're looking -- I'm wearing a Cisco hat and saying it's good for Cisco to leverage our relative market strength in routing. And I may have to cannibalize the Optical business in order to do that, and we're willing to do that. And we think that's the right thing for our customers. We think it's the right thing for Cisco.
Okay. Maybe just last on this topic. The Optical System vendors that are trying to add routing as a software layer or something, why does that not work as a solution?
I will never say never, and I don't discount our competitors. We've got 25 years of investment in routing with a couple of thousand people writing software. Hard-won lessons and building very large-scale networks around the world, and it's a hard problem. So, I wish them luck if they're undertaking that.
Yeah. Okay. Good. The -- maybe back to kind of the Optics side. You talked about still being relatively early in 400-gig deployments. Kind of talk to us about that evolution and how you see the next timing and scale for the next few versions.
Yeah. Well, first -- one thing I would want to be clear about is depending on which market segment we're talking about and even within market segments, different customers, the lifecycle for a given technology can be very, very long. But we're still selling a ton of 10-gig optics, a ton of 10-gig optics. And 10-gig was around 20 years ago. And so, I think the tail for things like 100-gig and 400-gig is a very, very long tail.
And without generalizing too much, what you see is web will adopt a technology like 100-gig very quickly and jump to 400-gig very quickly, and then jump to 800-gig, and they'll jump to 1.6T. The service providers are going to be -- to have a much longer timeframe for deploying that technology, easily ten years for something like 100-gig, easily. And they are going to generally be slower in jumping on a 400-gig bandwagon or an 800-gig bandwagon. They'll be slower jumping on it and then have a much longer time of deployment.
And then, when you look at something like enterprise, it's much, much later and much, much longer. So, like most of my 10-gig or 1-gig, we still sell a ton of 1-gig, is going in enterprise applications. So, these technologies have a very, very long tail. And if you ask like what's Google going to do or what's Facebook or Amazon are going to do, you get a very different answer than if you ask what Bank of America might be doing or what AT&T might be doing.
Okay. And then talking about the big hyperscalers, what kind of trends are you seeing there? And I'm assuming as capacities go higher, there's got to be optical -- much more optics around. We know each generation of switch has more optics. So, maybe talk about the trends there, and what you think that means for the business.
So, let me talk both inside the data center and outside the data center. The -- inside the data center, 400-gig is pretty well being deployed right now by the web players. That's, I would say, entering maturity. It's still relatively early stage, but entering maturity. 800-gig is probably coming in the next couple of years. And 800-gig will be a little different than what we've seen in previous technology jumps in that 800-gig will be -- on the router port, it will support 800-gig, but the optic itself will likely be two 400-gig side-by-side, packaged into one optic. And we do that for technology reasons and cost reasons. So, it will be a longer time before we see 800-gig on the optics side.
It's also an issue of compatibility. If they've got a lot of 400-gig out in their data center and they put an 800-gig optic that has two 400-gig ports effectively on it, they can connect it to an existing 400-gig. So, there's a life cycle management issue there as well. Pretty mature inside the data center for 400-gig, but I'd still say, there's a lot of growth there still ahead of us.
Once you leave the data center, the web guys have metro networks and long-haul and subsea networks. So, those are three very different markets. I think the metro markets for the -- or data center interconnect type market for the web players are largely going to go to this. For a couple, they're already there. Like this is exclusively what they're deploying. For others, I think they'll get there. Once you leave the metro area and get into long-haul or subsea, then I think this, which has higher performance than the pluggable supporting -- this can support many-thousand-kilometer applications, this was maybe up to 1,000 kilometers today, they'll still deploy something like this, a chassis-based solution.
The other thing I would say is we're -- for the last 20 years, the industry has been sort of a game of leapfrog of let's go from 2.5-gig to 10-gig to 25-gig to 40-gig to 100-gig. And every time you do that leapfrog, you get more capacity on the fiber. We are now approaching the point where we're just -- we're hitting what's known as Shannon's limit. We're just going to be out of gas on the fiber. So, we can't play that game anymore. Like our next-generation DSP coming out of Acacia will deliver 1.2-terabit on a wavelength, but the total fiber capacity that can be supported isn't moving that much. So, we can get a little bit better economics with a 1.2-terabit wavelength and maybe a 600-gig or 800-gig wavelength, but we're not really moving the needle in a significant way in terms of the total capacity. And then, if you ask what's beyond that, it's very little incremental gain that we can get in terms of the total capacity you can put on a fiber. So, then we have to turn our attention to things like power or cost, and say, look, the game is going to be who can drive to a lower power consumption on these things. It's not necessarily a game of capacity gain.
Interesting. One of the priorities you talked about was third-party for Optics. How do you go about that? How difficult is that to really start moving the needle on that business?
So, I think we've made good -- I think we're early stage in there. We've made very good progress with some very key logos. When a customer -- and now, I'm talking primarily about service provider customers, because web is a little bit in a different category. But when a customer decides, for instance, that they want to consolidate their optics spend because the optic looks the same to them for whether they're plugging it into Cisco or Juniper or Nokia, we want to have a seat at the table for that conversation, because we put optics through a certification and qualification process that is absolutely unparalleled in the industry.
No supplier of optic, no other vendor, whether it's Juniper or Arista or Nokia, does the level of certification on an optic that we do. So, we can give our customers very high confidence that when they buy the optic, it will work in any host and it will work under all operating conditions. That means temperature variation, humidity variation, voltage variation, all these different permutations, we test for.
And we have a diverse supply chain. We make sure that even when we're sourcing the optic ourselves, even when we design and build the optic ourselves, we still have second sources for either all the technologies that go into that or the optic itself. So, we can take that supply chain management issue away from our customer, and say, “Look, we will guarantee you that there's diversity in the supply chain. We'll guarantee that when there's typhoon in Thailand or an earthquake in Japan that takes down a good part of the optics supply chain, that we've already thought about that.” And that gives our customers comfort.
So, it's more than just does the optic work, because the optics are fundamentally commodity. We make sure that we can guarantee it's going to work under all operating conditions and that we can diversify the supply chain on behalf of the customer. So, with that value, I think we have a good selling proposition to customers.
Okay. You mentioned supply chain, I didn't, but now I have a follow-up.
So, maybe talk a little bit -- it's challenging, it's whack-a-mole and golden screws and all that stuff. So, where are you guys now looking -- in your business, how are you feeling about supply chain and volumes?
So, I would separate -- Optics, I think, is in pretty good shape. We're heading down to back to like four-week sort of lead times. There are hot spots. So, we're not there yet, but we're heading there. And for many optics, we are -- we're within four-week lead time right now. Optics has not suffered, in general, from some of the other areas like the high-capacity ASICs. The semiconductor industry has not hit as much in the optics area. And things like power supplies have not been a real constraint for the Optics themselves.
The Optical Systems are still on pretty long lead times, like 35, 37-week lead times, but we are seeing those come down as well. I would characterize it as, I think, there's daylight. We see daylight. We see improvement ahead. We're not out of the woods. This is not a mission accomplished statement yet. There is a whack-a-mole issue going on where there are some trouble spots we're still dealing with, and then there are some just pop up randomly that we still have to deal with. So, we're not out of the woods yet, but we're in a far better situation now than we were a year ago.
Okay. Just curious, obviously, you have Acacia and some real base level technology and IP. Could you talk a little bit about the broader Cisco Silicon One and having silicon capabilities in addition to the optics? How does that better position you relative to maybe some optical or optics pure-play companies?
So, one thing, I think from an Acacia perspective, we worked really hard to vertically integrate as much as possible on the design and development side and really own all the key technologies there. And I think we're at a point where we can say that for all the key technologies that go into that optic, we've got ownership of that technology. That gives us control over the design, the performance, ultimately, the cost and it gives us comfort that we can manage those trade-offs between the various pieces in the optic.
I think it was December of 2019 that we made a pretty big announcement that we were going to shift our business model to support a component business model. So, in addition to doing our traditional systems business of selling fully integrated routers and switches with software and hardware and services, we were going to meet our customers where they want to be met. If they want to buy just the optic from us, not buy any of our systems, we'd sell them the optic. They want to buy just the silicon from us, like Silicon One, none of our systems, none of our software, we'd sell Silicon One. If they want to buy just our software or just our hardware platform and put their own software in it, we would do that. Very different business model for Cisco, has supply chain implications, has inventory implications, cash cycle implications, and it's fundamentally a different way we think about managing a business.
I think that, that opened up business for us with the web players, in particular, who were the main target for that whole announcement to say, look, if you want to go build your own, we want to be part of that solution with you. We don't want to be on the outside looking in. So, if you want to use our silicon, we'll be happy to work with you. And we now have customers buying only our silicon, only our optics, only our hardware platform with no software, putting their own software on or putting something like SONiC on it. Every combination you can possibly imagine, we are now to some customer offering. And I think it's opened up a lot of possibilities for us that we don't see many of our competitors being able to match with both silicon and optics.
Q - Tim Long
Okay. Great. Perfect. End time here. So, thank you, everybody. Bill, thank you so much. Really appreciate it. Thank you.
A - Bill Gartner
Tim, thank you very much. Appreciate you having me. Thanks, everybody.