How Ciena makes the whole internet work, with CEO Gary Smith

Today, I’m talking with Gary Smith, CEO of the networking company Ciena. You probably aren’t familiar with Ciena — the company isn’t really a household name. But every internet user has relied on the company’s products; in fact, every internet user has used its products. 

Ciena makes the hardware and software that make the fiber optic cables that connect the world light up with data. That’s everything from local fiber networks for broadband ISPs to the massive undersea cables that connect continents. There’s a high probability that this very podcast came to you over a Ciena-powered network, in fact — this company is everywhere.

Gary himself is a fascinating character in this story. He joined Ciena in 1997, the year the company had an astonishingly successful IPO in the middle of the dotcom boom. And he became its CEO in 2001, right after the dotcom crash sent the company’s stock price — and the rest of the economy — into a tailspin. So Gary has had a front-row seat to the development of the internet as we know it, and Ciena has been there every step of the way — as telecoms and undersea cables first brought the planet online, all the way to the rise of cloud computing and, now, generative AI. 

You’re going to hear Gary and I talk a lot about something called WDM, or Wavelength Division Multiplexing, which is the fundamental technical innovation Ciena is built around. WDM is absolutely key to how the modern internet works, and most people have no idea it exists. The basic idea is that WDM uses multiple wavelengths of light to fit more data onto a single fiber optic cable, which allows those cables to deliver more and more information over time — you’ll hear Gary call it “virtualizing” a fiber optic cable, like you’re turning a single cable into two, or five, or 10. This is obviously hugely important as the world’s internet usage increases and data-hungry applications like streaming video and generative AI ramp up.

Ciena didn’t invent WDM, but it was the first company to deploy it commercially in the ’90s — and I wanted to know how Gary was managing his investment in pushing that technology forward and whether he was thinking about the next kind of tech that might disrupt it.

I also wanted to know about Ciena’s customers. There are only so many companies building deep sea cables and giant data centers, and it won’t surprise you that the customer set has largely shifted from telecom providers and ISPs to giant cloud computing companies like Meta and Google and Microsoft over the past few years. Just a few months ago, Meta announced it would pay $10 billion to build its own exclusive subsea cable stretching nearly 25,000 miles. Ciena sits right in the middle of that complicated relationship between industry and infrastructure, and I think you’ll find Gary’s perspective to be pretty unique. 

We also got into the fracturing state of the internet and what it means that people in different geographic locations, dependent on various levels of government control, experience the online world differently. Connecting cables stretching across continents means the people on either end of those cables might have very different ideas about the internet, and Ciena has to take all of that into account.

See, I’m guessing you didn’t know about Ciena before, but almost every single Decoder idea is right here, sitting on the backbone of the internet.

Okay, Ciena CEO Gary Smith. Here we go.

This interview has been lightly edited for length and clarity. 

Gary Smith, you’re the CEO of Ciena. Welcome to Decoder.

Thank you. Good to be here.

I am very excited to talk to you. The Decoder team and I were talking earlier about how we love stories of hidden companies that run the world, and Ciena very much feels like one of those companies. It is a huge player in networking. It’s a huge player in the future of our networks. There’s some AI to talk about. There’s undersea cables to talk about, which is always fun. Quickly, for everyone, describe what Ciena is and what you do.

Basically, Ciena is the leading technology company in the world for providing high-speed connectivity. So if you think about all the networks that have been enabled around the internet and around the world, or, as you alluded to, submarine cables, Ciena is really the driving force behind that high-speed technology. Even with mobile connectivity, terrestrial connectivity, and submarine cable connectivity, we’re basically the best in the world at moving optical bits.

I always think about it as two internets, right? There’s the internet we experience — the digital internet —  and then there’s the physical internet: the actual cables and wires that connect our networks together and connect us together. That is Ciena’s business, running these cables, figuring out how to move data across them more quickly, and figuring out how to move more data across them. 

Ciena’s been around for a minute. How has the dynamic between what people are doing on the digital internet most consumers experience and what you need to do on the physical internet changed? It seems like something we don’t ever really get to talk about. I’m very curious to ask you about it.

I think we consume most of the internet now in a mobile form, on our iPads, iPhones, or whatever. We still have some desktops, but it’s all laptops. The first point of connectivity there is mobile. It’s available to us all the time. But what we don’t think about is that it’s only mobile until it’s not. It needs to get terrestrialized to the nearest area where it comes down into terrestrial —  typically fiber because to get really high speed, you need to use fiber. Wireless has some limitations around that. So basically, until it gets to any mobile application, its real objective is to get terrestrial into that cell tower as quickly as possible.

Then, from the bottom of the cell tower is when all the magic starts, with the physical fiber cables and the networks that go around the world. If you have an inquiry, depending on what it is, that could require a different continent. That goes down through the cell tower and out across the sea and back to you. But that’s invisible to us all. We’re able to do it at such high speeds now that it’s transparent to the applications that you’re using.

That’s the part I really want to dig into. Intellectually, I think I’ve always understood that. I think most people understand that there are servers and data centers elsewhere and we are communicating with them somehow. But at some point we need to hit a fiber line and actually talk to those servers. 

What I’m trying to understand is that there’s the consumer application layer of the internet — we’re all going to watch a bunch of Instagram reels or something. Then, there’s the internet we need to build to support that. What’s the relationship between those two? How often do you think, “I see the next set of applications that are coming on phones, I see the next set of devices that people may want to use. or I see the rise of streaming services. We better start building an internet that can support it?” That dynamic often seems subsumed in some cell carrier or data center conversations. It very rarely comes to the forefront.

Because it’s the piece that connects all of this stuff. Everybody focuses on things like AI, Netflix, or whatever right now. How do you get those things to manifest themselves to the consumer and to enterprise? The reality is that it required a massive build-out over the last two decades for a fiber infrastructure that basically connects all these data centers together, that connects them through submarine cables, and connects them across countries. It has intelligence on it so it knows where to go to get that information. That has evolved over the last two decades. It continues to evolve because if you think about some of the business models that we now take for granted, like Uber, that wouldn’t be available if you couldn’t get instant response to it.

Twenty years ago, that would not have been possible as a business model. There was an interesting point in the early 2000s where there was this massive overbuild of fiber capacity. It was sort of the telecom nuclear winter where we had more capacity than apparently we’d need for the next two decades. People built out all of these large fiber networks and no one came. But that enabled business models to take large amounts of bandwidth at low cost. That created business models like Amazon, which weren’t able to get stimulated and be sustainable until they had large amounts of dependable connectivity. So it’s constantly evolving with applications that are demanding much higher bandwidth. We’re in an era of that now with AI for sure. The networks have to catch up and facilitate that.

Let me ask about the networks. Let’s say I want to build a network in some rural part of the country. Do I just call Ciena and say, “All right. We’re going to get to work. Go dig some trenches”? Do I run the network? Do you run the network? How does that work?

We don’t run the network. Those are the service provider carriers. If you take North America as an example, there are hundreds of carriers that do the local last mile piece out in the rural areas. Then, there’s the Verizons, Comcasts, and AT&Ts of this world that then facilitate all of that connectivity to the consumer. We provide the technology that powers their networks basically. For them to get that high-speed connectivity, Ciena’s technology goes on the ends of those fiber cables that allows the speed and throughput to continue to drive the capacity.

What we basically do is virtualize a single piece of fiber. So instead of having that little thread, which is about a hair thread in terms of weight and distance, we can multiply that many, many times. That’s hundreds of virtual fibers because of the technology we use around it. That’s enabled the whole scaling of the internet. It’s enabled the scaling of this global connectivity phenomenon that we have, and it’s at the center of scaling for AI connectivity now and into the future.

That core piece of technology where you’re multiplexing and virtualizing the single piece of fiber, that was Ciena’s big technology bet. I believe it’s called WDM. Explain that to people very quickly.

Basically, it’s virtualization. Where you’ve got one piece of fiber, [WDM] makes it look like multiple fibers. Depending on the distance and what you’re putting down, it can be hundreds of fibers, which is actually physically manifested as a single fiber. So you don’t need to put more fibers into the ground. With the technology we have on either end of it, you can just scale it up. 

So, it’s the virtualization of fiber that has enabled this almost seamless marriage between the applications, to your point, and the driving of all the network traffic with the actual network. Through our technology development, we’ve grown the ability to go do that faster than Moore’s Law. We now have a new piece of technology, called WaveLogic, that can do 1.6 terabits of data down a single piece of fiber.

It’s always been explained to me that the fiber is the fiber, and we’re just sending light back and forth. It’s the pieces at the end that actually create the additional bandwidth to create that additional capability. Is that more or less true? Are you just focused on the bits at the end that talk to each other?

Absolutely. There are technology developments within the fiber that assist that, for sure, and there’s been a lot of that development from companies like Corning in the fiber world. But basically, the magic is the technology that goes on the end of those fibers. When you think about it, if we had to scale by digging up the place every time we wanted to put more fibers in, it’s just not practical. We can get a tremendous amount of expansion and virtual capacity even on old fibers just by putting our technology on the end of it.

You were talking about scaling faster than Moore’s Law, which is roughly a doubling every 18 months. Decoder listeners are very pedantic. They will tell me that’s not actually what Moore’s Law means, but I think that’s how you’re using it. I appreciate our listeners very much, but I don’t think that’s what applies here. You’re talking about a doubling of bandwidth every 18 months. 

You’ve been doing WDM since 1997, right? This is the technology bet of the company. Is it the capacity of WDM where you’re using multiple wavelengths of light across a single fiber? Is that what’s doubling?

Yeah, it’s our ability to continue to drive that. When we first started, we could multiplex and keep it very simple. It was a single piece of fiber. We thought we could put four virtual fibers down there. That was revolutionary at the time. Many of the carriers we went to with that said, “Oh, that’s fantastic. We’ll never need more than that.” Then, we figured out how to do eight, then 16, then 32, then 64. And we’re way, way beyond that now. 

So what that means is you have that single piece of fiber that looks like 64 fibers. That’s enabled this massive scaling of architecture around the world. Then, we deploy that technology not just with local connectivity, but  across the country and whole continents — 8,000, 10,000-mile submarine cables. We can support that kind of capacity across the submarine cables.

Do you just sell these boxes to the telecom and data center providers and walk away? Do you have ongoing support?

No, it’s a very integrated partnership. Most of the major carriers —  I think 96 percent of the major carriers around the world — are Ciena customers. These are collaborations that we’ve worked on for multiple decades. We help manage the networks. We do a lot of the installation work for them. We do the planning. We have the software that actually manages the network and looks for predictive issues or challenges and switches automatically. If there are problems on certain fibers, it will detect that and switch it through. So, there’s a lot of underlying technology and collaboration between us and all the service providers. 

Also all of the hyperscalers. Pretty much all of the hyperscalers use Ciena to connect their data centers around the world. We partner with them as well on technology developments that are specific to their networks. You look at the hyperscalers, some of the largest networks in the world, and they’re very complicated, multi-continent, metro, long-haul, submarine cables. The biggest players and the biggest owners of capacity on submarine cables now are actually the cloud players. That used to be the traditional service providers, but now it’s the likes of Google and Meta that really own a lot of the global capacity on those submarine cables.

This is the thing that I was talking about at the beginning. It seems so hard to understand, and it’s so interesting to talk to you about it because you’re in it. The internet we experience is obviously changing the dynamics of the internet we build. So. I experience my phone and there’s a bunch of stuff happening on some data center owned by Google, Microsoft, or whatever, and that means those companies now own most of the capacity of undersea cables. 

I don’t think that chain of events is intuitive unless you’re paying attention to it. But just talking to you for 13 minutes here, that seems very intuitive to you. It makes sense to you that they would have the capacity because I’m not sending nearly as much data as a consumer across the ocean. Do you see that changing? How fast has that grown? Describe that dynamic to me.

Take the submarine cables as an example. It’s a good proxy for what’s happened here. From the 1980s to 2010, this was all funded by the service providers. It was a way of getting transatlantic phone conversations largely to Europe and a little bit to Asia. These companies, like British Telecom and AT&T, would have these cables, they would get together a consortium, and, as a syndicate, fund these submarine cables. With the advent of cloud, you had these players that wanted to have very high-speed connectivity. Take for example Meta, Google, Microsoft, or any of these players. If they were going to be in the European market, they needed to have very high speed, dedicated capacity. So in about 2010, you began to see the cloud players purchasing the capacity of those submarine cables.

In the last five years or so, you’ve seen Google begin to own many of these cables, with Meta and others too. So, it evolved from, “Hey, it’s all this service provider and it’s just phone calls.” When you want to reach your aunt in Europe, in India, or whatever, then it’s just a phone call, which is a very small amount of data. The amount of data that’s now being required by these cloud players is multiple more times than that. You’ve got a lot of graphics going over it as well, which takes a lot of bandwidth. So the need for bandwidth has grown dramatically. It makes sense that the cloud players want to connect their data centers. If they’re putting a data center in India and want to connect it to North America, they want to own the connectivity for that, and it’s a much higher speed. So they’re not just buying the capacity. They’re actually taking control and owning the cables.

So, there’s been sort of an evolution of that over the last five years. The first 10 years was, “Hey, we will take the capacity now.” Their networks are made up of a hybrid of owning the cable, owning some, or they rent it from the service providers. But you’ve now got, I think, 600-plus submarine cables around the world. You’ve got about 1.5 million kilometers of fiber at the bottom of the sea facilitating this, and the cloud players are really driving that marketplace now.

That’s going to lead us into the inevitable conversation we’ll have about AI, but I just want to stay on this for one second. How has the verticalization, from “what we sell is virtualized compute capacity” down to “we will now own the cables in the ocean and verticalize our business and make sure we own it tip to tail,” changed your company? 

We’re looking at, “There’s WDM in 1997. This is the core innovation that lets us get more capacity out of the fiber networks. We’re initially going to sell that to telecom companies to do more phone calls or the early internet.” Now, multiple generations in, you’re selling it directly to Google to connect data centers so that it might run AI workloads.” It feels like that would have some pressure on your roadmaps and your go-to-market. How is that expressed?

The evolution has been such that the first appreciators for that technology were obviously the service providers. As the internet began to gain traction in 2003, 2004, you began to see the cloud players get more interested in the network because if they’re going to monetize and get to the eyeballs that they wanted, the network actually becomes important to them. So, there’s been a steady increase in appreciation of that even prior to the AI piece, and they’ve become much more active. I mean, we really didn’t sell to the cloud players until about 10 years ago, and we had zero market share there. Now, we’ve got the main market share with a leading player in connecting data centers and cloud players around the world.

Submarine cables are the same kind of thing. We really didn’t focus on the submarine cables with our technology because there wasn’t that much capacity being required. We’ve now got number one share in submarine cable connectivity around the world because they value very high-speed capacity. So the cloud players have gone from being a very small percentage of our business to over 50 percent of our business now being, directly or indirectly, cloud and hyperscalers. So it’s a very significant shift in the business, and that’s really accelerated in the last five years.

I think this is a great time to ask the classic Decoder questions. You’re describing a shift in the business and a shift in your customers. How is Ciena organized now? What does that org chart look like?

It’s been easier for us in some ways because we’ve always been the leading technology company, so it’s really about who appreciates that technology and collaborates. This shift to the submarine markets and to cloud players has generally been a good fit for Ciena because we move very quickly, and we’re at cutting-edge technology. The latest technology we have is actually on three-nanometer chips. We were the first company in the world to have a 3nm chip actually be used in the telecom space, and we’re now actually working on our second. We’ve got that out, it’s carrying traffic around the world, and we’re working on our second version of that. We’ve always been at the leading edge of technology, and it really is a natural fit with these cloud players.

So, we’re organized very simply. We have a large-service provider team around the world, a dedicated team for the hyperscalers and cloud, and an engineering team that works with them on co-development and stuff with that kind of technology. I’m a great believer in simplicity of organizational design. So, we have a very simple organizational structure that’s evolved to address that. That’s how we’re able to scale as quickly as we have.

You’ve got the two big markets — the telecoms and the hyperscalers —  and you have the engineering organization. I’m assuming the telecoms and the cloud providers want different things. How is that expressed in what the engineering organization goes off and builds? 

It’s interesting. There’s a lot of commonality about what they want. If you think about it from an engineering point of view and the latest WaveLogic technology we just developed, we actually take that same technology and we deploy it in slightly different variants to both markets. So, we get a better scale of development over that. Obviously, we develop a single chip and then do variants of it for specific applications that they want, but that same technology is essentially consumed across the service provider world because they all want the same thing: better economics across the fiber that they’ve got and increased bandwidth.

The particular sensitivity, I would say — what’s really the difference between the two markets — is the cloud players are much more focused on power, which we all hear about from a data center point of view. It manifests itself exactly the same way in the telecom networks. We do variants that are power optimized for the cloud players, but that’s the predominant difference. That’s just taking the same core technology and doing a different variant of it that’s prioritized around power. That enables us to scale the development and the rest of it. So, we have a single global engineering team focused on this core technology.

Ciena has always been a very focused player. We’ve been very focused on just high-speed connectivity. This has enabled us when most of our competitors are trying to do mobile or other kinds of things. For the past three decades, we just wanted to be the best in the world at moving optical bits at a high speed. The opportunity we have now is increasingly into the campus and potentially even inside the data center with that same core technology, and that’s all in front of us. We get zero revenue from there right now. But it’s the same customers, so that’s an exciting opportunity for us in the future.

I want to talk about that a little bit more, but can I just ask one more question? You said power. That suggests to me that you’re building a data center, you have an energy budget, and then you hold all the vendors to some number on the line to make sure you hit your energy budget. You’re feeling that too, even as the person is like, “Here’s the box that connects to the fiber line.”

Yes. It manifests itself differently obviously, but everybody talks about the data centers and the power and the rest of it. Everybody is very focused on that for obvious reasons. But that same issue deploys when they come out onto the network too. You put amplifiers on the fiber to do that, and around these amplification sites, they want super low power because they are constrained around power and space.

I’m assuming they would rather spend their power budget on running the GPUs as hot as they can. 

Exactly. They don’t want to spend it on the communications part, or they want to spend as little as possible on that. They want it optimized. We’re able to do that for them in terms of the footprint of the technology. We can now get down on a single plug what used to be a rack of cords. Quite literally, a six-foot-wide rack on a single plug, that same capacity. That’s what we’ve evolved over the last two decades with. So, we’re well placed with these guys to continue to drive down the power usage on the network, not just in the data center.

How big is Ciena? How many people are on all of these various teams?

We’re close to about 10,000 people worldwide. We’re very distributed. This year’s consensus revenue is about $4.4 billion, something like that. We’re profitable, and we have a strong balance sheet. We very much invest directly into this technology. Because we’ve been such a focused player at it, we felt that this high-speed connectivity would be appreciated eventually, and it’s really coming into its own now. You have to make these investments for the long term. Some of our technologies take five-plus years to evolve. We’re fortunate in this time right now. It’s a good intersection point where high-speed technology is super appreciated.

Is that an even split? Is it 3,300 in engineering, 3,300 in hyperscaler sales-

No, it’s basically predominantly engineering. As you can imagine, this kind of technology is more about engineering, so more than half of our team are engineering talent. We’ve got applications engineers at the front end who help make sure we understand the applications and work with the customers on that, but more than half is engineering. As I said to you earlier on, it’s a very simple structure.

One of the things that really strikes me is there is this core enabling invention, WDM. The company is built around it. Ciena’s made a few acquisitions over the years. You’ve led a few of them. There was one in 2019 for Centina Systems. Is WDM still the heart of it? Is that proprietary Ciena we’re building around it? Is there a next way to use fiber that’s coming?

No, it’s basically the journey that we’re on. We’ve made multiple acquisitions, but it’s all been about the core mission, which is high-speed connectivity and the things we need around that. A lot of the acquisitions have been around vertical integration of that, so we have more control over it. It’s not just the core technology. It’s the things around it that you need to do. Most of those surrounding technologies were where we could focus and provide competitive advantage by doing something to them. 

Over the years, we’ve worked very carefully at acquisitions and organic developments that make us vertically integrated so we can control that. There’s advantages to control and scale, but what really drove us was that we think we can put competitive advantages in each of the technologies that we gather around the core technology, if that makes sense.

The reason I’m asking that is because often you might want to buy a company that’s early to the next turn, but it seems like WDM is still the heart of what you’re doing, and you don’t see a finish line there. 

Basically, it’s this technology. The actual variant that we focus on is the coherent technology, and that’s what has enabled this DWDM to scale so much. But there are other technologies around it that we’ve developed, like the SerDes — the Serializer / Deserializer — and the kinds of things that are for analog conversion, which is incredibly difficult to do. We’ve got some of the best technology in the world around that. It’s the moat that you put around that, but it’s not the next thing. We think there’s a lot of headroom, particularly because this technology is all deployed right now. Think about it outside the data center. So as soon as it wants to present outside the data center, that’s Ciena.

The opportunities we have with the same kinds of technology are now becoming apparent inside the campus — these campuses are getting super big now — and eventually will be inside the data center. That’s because the  GPU and compute investment that’s going on requires higher and higher speed over longer and longer distances, even in some of these data centers. First of all, they’re enormous. You’ve got fiber going around them, sometimes multiple kilometers within a single data center with that fiber getting wrapped around. 

We think there’s an intersection point for our technology there. The technology there right now, which is just called direct-connect, is required to connect all of the racks within the data center and the intra-data center. The physics of that is getting challenged as these GPUs require higher and higher speed. So, we think there’s a real opportunity for us to intersect that, to your point, with the same kind of core technology that requires some variations over the next 1–3 years.

Let me ask you the classic Decoder question, and then I want to get into that. You’ve been at Ciena for a long time. I think you’ve been there for 30 years. You’ve been the CEO for almost 25 now. How do you make decisions? What’s your framework?

Depends on what kind of decision it is, right? Not all decisions are equal. I very much feel that it’s a team game. When we make strategic decisions — which I think is what you’re  alluding to with how we invest and what long-term things we’re focused on — we’re strategically aligned around what we want to go do. We’re a very focused company. Back to the point I was making earlier, I think people appreciate the best of breed, and we want to be the best in the world at what we do. The north star is very clear to us. Then, it’s about what things we need to do to get out into the future on that.

From that point of view, the decision making is somewhat easier because we know we’ve chopped the field, so we know where we’re playing on. There’s not some great conversation about adjacencies and things that we might want to get into; it’s all based on our core technology. To answer your question, if it’s something strategic around those investments, it’s always a team game. We’re super collaborative. It’s how we work and get cross-functional input into all of those things. So, it’s a process. 

But I would say that I think the role of a CEO sometimes is to say, “Okay. We’ve had the debate. This is the overall perspective of what we want to do. Let’s go.” I think part of the role of the CEO is to facilitate the best conversations and discussions and then help drive to a conclusion.

Let me give you a hypothetical that I suspect you will shrug off, but I’m going to try anyway. You’re looking at two big groups of clients. You’ve got your telecom companies, and you’ve got your big cloud providers who are all racing into AI. You have some opportunity to grow that business by going into the data centers, not just landing the fiber at the doorstep. That seems like the much richer opportunity. 

I look at the state of the telecoms in the United States, for example. They’re not growing much, right? They’re churning customers between them. It would seem very obvious that your entire team would say, “Don’t worry about them so much. We need to put all of our focus on the AI providers because that is where there’s growth using our same core technology.” But then you still have to service the telecom customers, I imagine. How do you balance that out? How do you make that decision to say, “These investments are going to go over here?”

It’s actually much easier for us because, back to the point I was making, it’s the same kind of core technology. When we move the ball down the field for the cloud players, we can apply that to the service provider space because it’s the same core technology. In terms of their needs, there are actually a lot of similarities. Another way of thinking about this is when we start talking about the data center and the rest of it, we provide optical systems where we do the whole thing for the service providers. And we do that pretty much for the cloud players too. So we do all of the systems for the software, what we call the WDM modem technology, and the line systems. We do all of that, and often we’ll install it, support it, and maintain it. We do that for service providers and we do that for the cloud players.

Then, to the point around the opportunity in the data center, it’s not necessarily about providing those complete systems. It’s taking them, putting them in plugs, and then putting them in components. So it’s a different kind of play, but it’s still the core technology. So, anything you do to advance that also applies to our systems business, and our systems business is both service provider and in the cloud. You don’t have to make the “which are your favorite children?” kind of choice, which is where you were going with that. I think that’s another benefit of a company staying super focused on its core technology and understanding what value we bring to the world. That is very difficult for anybody else to replicate,

Well, can I ask you about that? Because it feels like the Trump administration is incoming. By the time people listen to this, he will have been inaugurated and we’ll be off to the races with a new kind of business climate here in the United States. Mergers will be in fashion again. They have not been in fashion over the past four years. If I was running Google or Microsoft, I would say, “Well, this is the core technology that’s going to make my data center more efficient and move data faster on my cable. I’m just going to buy this company.” Have you been approached to be acquired?

Well, I’m not going to–

I’m just wondering because I can see why over the past several years, all of that conversation has chilled. But it’s also remarkable to me that a company like Ciena has remained independent for 30 years.

That used to be the conversation all the time. Think about 20 years ago. I’m not sure some of your viewers will remember some of these names: Lucent, Alcatel, Nortel, even Nokia. The model was all what I call the “generalist model.” They were actually spin outs from their phone companies like Marconi. Northern Telecom came out of Canada. So with the model, we had to be end to end. You had to put switches, routers, handsets, mobile infrastructure, and this thing called transport. We said when we started, “No, we’re going to just focus on the transport piece because we think it can get to scale.” 

What’s happened over the last two decades is that the generalists, because they couldn’t afford to be at the cutting edge in each of those areas, have all struggled. You trace the genesis of all of those companies now and basically they’re all in Nokia. With Alcatel and Lucent and all the rest of them, the music stopped and that’s where they ended up.

The point I’m making about that is the world’s too competitive now for the whole end-to-end thing and being a generalist supplier. It just absolutely is. They used to say to us, “Well, when are you going to get bought?” We used to say, “Listen, we’re a public company. We have all the fiduciary responsibilities that you would expect with a public company, but we think we can get to scale and be a real company just doing what we’re doing.” Now, these things take a while and we’ve been public for 25 years, but we’re at a point now where we absolutely have scale. We’re very strong financially. We can invest in this kind of technology. We’ve got the number one market share in all the key markets that we want to play in. So, we’ve got a good opportunity to drive shareholder value from here.

Do you think that DWDM, the multiplexing you do to virtualize the fiber, is enough of a moat? You were describing the other acquisitions you’ve made and how you surround the core technology with other things to enter different markets as the moat. But is that core technology something that… I don’t know if Mark Zuckerberg woke up one morning and was like, “screw it, I’m open sourcing it,” he could just do it? Or is that more specialized?

There are other companies that have variants of this technology, for sure. It’s just that we’re able to implement it faster and quicker and better than anybody else, and we have the core competencies around it. A bit of an analogy is TSMC. People kind of know how to do [what TSMC does], but they can’t do it, right?

You can go to ASML and ask for a machine and that’s not where it stops.

Yeah. Because it’s people, and it’s the folks that understand that it’s the five, 10 percent of each of these areas that really make the difference to things when you put it and all the things we’ve built around it all together. So like with any of these things, it’s a confluence that makes up the competitive advantage. It’s when you get down to the real technologies. 

It’s all of those things that provide the moat for us. It’s the fact that we’re focused and the fact that we’re now so embedded with these key cloud and technology companies and were part of the collaboration process around the evolution of their networks. It’s the relationships that you build up, that trusted relationship and the technology collaboration over time. That’s why we think we’ve got an opportunity to take that into a whole new market for us with the campus and eventually inside the data center.

I want to talk about two things that seem like they’re going to happen to the hyperscalers, in particular, over the next four, 10 years. One of them has a lot to do with subsea cables and the other one is obviously AI. Let’s start with subsea cables. If I look broadly at our nation’s big companies — Amazon, Google, Meta, Apple — they’re about to enter into a world where an emboldened Donald Trump wants to just reset the international order. They own cables that connect our continents together. There’s a lot of people who want to break the internet apart in different ways. We might ban TikTok. We might not ban TikTok. China doesn’t allow our apps there.

There’s a lot of interest in what is happening on our networks that connect our countries together and who gets what where. Then there’s just the reality that they are wires in the ground. Bad things happen to them. They get cut. There are wars. How much of that do you engage with personally? How much do you think about, “Okay. There are some fraught geopolitics that are about to change. I need to make some decisions to get ahead of it, be aware of it, insulate myself against it, or insulate Ciena against it?” How much of your time do you spend on that?

I mean, there are the headings of things you can control and then things that you can’t control but need to be mindful of, right? It’s definitely in that second list. I could argue that over the last five or 10 years we’ve had quite a lot of geopolitical dynamics going on. We are mindful of that, particularly because we operate in multiple countries, so 80 countries around the world. Our technology is widely deployed. We are mindful of those kinds of dynamics, and I’ll give you an example.

We made a decision in the early 2000s that we would not go in and be a provider to the Chinese market for its major service providers. There was a whole set of thought processes that went into that. One of our competitors at the time was Huawei and it was just emerging. I don’t think people appreciate this, but the first real iteration of technology that Huawei came out with was actually in our space, so way before all the mobile stuff and all the other omnipotent things that it’s now doing. So around 2003 we said, “What’s this company, what are they doing, and how is it achieving price points that are just lower than our cost?” Turns out that we’re being subsidized by the Chinese government, right?

It’s like we’ve seen this movie several times since 2003. 

So one of the thought processes that we had, just to give you an illustration , we said, “You know what? We will not compete with it directly in the China market. It obviously wants us there because we’ve got leading technology and the rest of it. We’re actually not going to manufacture in China either. So we’re going to be the antithesis, if you like, of Huawei in every single way. We’re going to be the alternative that’s not a Chinese manufacturer, and we won’t operate there.” 

That was a very controversial thing. You imagine a young company that’s just gone public and going, “Hang on a minute, why are you not going into the largest market and fastest growing telecom market in the world?” But we had the strength of our convictions. It’s taken a while, but I think largely that was a good decision for us because we didn’t get distracted by that Chinese market.

They’ll give you 10 percent of that market, but they’ll make sure you won’t get anything else, and you’ll partner with a local partner. I spent a lot of time in Asia. I lived in Asia for a number of years before I came to Ciena. And I saw that movie, to your point, play out in so many different ways. I was super sensitive to that, as was the team. So, that is an instance of us doing something because of the geopolitical dynamics at the time and us making a decision. You don’t know how that’s going to play out because things can change., You’ve just got to be smart around what you think is the best thing for the shareholders and your company and stick to it.

Do you mostly follow the lead of the clients? Do you have your own points of view on geopolitics? I’m thinking of things like when you land the cable in Europe and your devices there decoding all the traffic that there are a number of European spy agencies that might want to sit right next to that. 

The NSA in our country would love to sit right next to that. When you land in other countries that have different kinds of privacy and security concerns, they might want to be a part of that. Is that something where you say, “Look, Meta owns this cable. You talk to Meta. We’re going to do what the client says,” or is that, “We are simply not going to build these cables in certain devices”?

We’re actually pretty fortunate that we don’t have to make some of those tough decisions because it’s not our cable. So to your point —

But the data going over the cable, that’s you, right? You’re encoding and decoding it?

Yeah, that’s what we’re doing, but even that technology is owned by the customer. We sold that to the customer, and what they do with it is entirely up to them. We don’t have control over that. It’s their network.. It’s their property. You think about these submarine cables. Someone owns the cable, and we put the technology on the end of it. It’s their technology. They own that, and they run it as well. We don’t run it. So they run those networks. We’re fortunate in that we’ve always been able to stay one step behind some of those dilemma questions. It’s not something we have to opine on.

When you think about the actual cables, they’re in international waters. What’s the longest one that you operate?

I think it’s about 10,000 kilometers. Some of these cables are around the Pacific where you get the big distances, so they are very, very long cables. A lot of them stop off on different islands and places and spur off. Some of the new cables going in are spurring off in different places. We’ve also developed a technology that overlays that provides for resilience. So if you get a cut on a particular cable, then we can reroute it depending on the architecture. 

It’s called GeoMesh. It’s software and technology that’s embedded into our core offerings and enables us to provide resilience across some of those submarine cable links. In the normal run of things, you get I think something like 100-plus faults a year, mainly trawlers dragging their anchors on a fiber cable because they’re buried at the bottom of the sea until they’re not. I mean, they’ve got to come up somewhere. Typically, it’s in the shallower waters where you’ll get some fishing vessel come across it.

Do you ever get angry sharks? Do you ever get a text message like “shark attack brought down the network?”

Actually, that’s a common question I get asked around it. There is a section around fish attacks. I think it’s zero. I’m sure there are some examples of it that someone can point to over the last three decades, but I’ve never heard of it personally. I think they’ve got other things to worry about than these little tiny… it’s about the size of a big hosepipe, so I don’t think it’d be particularly interesting to them.

We have this phrase that we’ve been using at The Verge, my friend Casey Newton came up with it ages ago. He calls it the “Splinternet,” where we perceive one cohesive internet, but really we’re fracturing. We’re fracturing into continent-specific, country-specific internets. 

As I talked to you about the cloud providers, it’s very obvious to me that we now have a number of private internets operating at high speed around the world that are just owned and controlled by single companies. You have a long view of the internet. You are one of the few guests on Decoder who’s ever just straight up talked about Alcatel. You have a long view of the internet, how we’ve grown it, and what it’s for. The telecoms, by and large, were national champions, right? That’s how they have thought about themselves. Even in the United States, that’s how they’ve thought about themselves. That’s where they came from.

There’s a new kind of nationalism in America around our tech giants in the Trump administration. Mark Zuckerberg says, “I want the Trump administration to go fight the European regulators on my behalf.” That’s new, it hasn’t come up before. We’ll see if this jacket fits. He’s trying it on. But the way we build networks is an expression of national capability. We haven’t experienced that too much as the internet. It’s been one global internet that connects all, and then we do whatever we want and set our borders. But we haven’t actually split the networks apart. But now you’ve got big cloud provider private networks. You do have rising nationalism around the world. You just have a pulling apart of the internet. Is that affecting how you are building networks or how your clients are asking you to build networks?

I totally understand the notion, and I think what you’ve got is that a disproportionate amount — this is sort of an English understatement — of the innovation and development for all of this has been done in the US, amongst all these technology companies, for sure. Now, the big cloud players are investing in that network because that’s their business model, and the network is becoming more and more important. When we talk about the network, to your point, we talk about it in this homogeneous form. 

But if you think about this in the context of some of the large markets that the US cloud players are trying to get into… take places like India. When they enter the Indian market, they are dependent upon connectivity from the local carriers. So they can tip up with the submarine cables and the landing stations, but then it’s, “How do I get connected to all the people I need to get connected to in India enterprise and consumer?” They are then dependent on the local carriers, be it Bharti, Jio, Vodafone Idea, some of the wholesale carriers that are also there, and fiber. There’s always going to be an interrelationship because as omnipotent as all of these cloud players are, they’re not going to be able to connect to everybody’s home across the planet. So you’ve always got this interplay between service providers and the cloud players.

What often happens there is that in some of these countries, the service provider will provide a specific network for the cloud player, and they call it a Managed Optical Fiber Network, or MOFN. And we help facilitate that. It’s our technology, so it matches all of their global networks. But it’s owned and operated by the local service provider to support Meta, Google, Microsoft, etc. 

Increasingly, I’d argue you’re seeing that interdependency work. So, I wouldn’t say it’s the counter-argument to the idea that we’re all splitting up because at the end of the day, to get to the consumers around the world, these service providers are not going away in that regard. That’s not going to change. They are the only ones that can provide that connectivity ultimately, be it the enterprise or the consumer.

I’m fascinated by this because it does seem like the overall geopolitical dynamics of the world are changing, and the internet will change with it. I just keep coming back to the question I started with, which is about how the experience we have on the digital internet obviously impacts the physical internet that you are responsible for building. I’m always trying to find the touchpoints between the two. The way we architect networks in India has a direct relationship to the Indian internet that those users experience, and it seems like those things are, if not diverging completely between different countries, certainly beginning down different paths. That, to me, seems utterly fascinating.

I want to end with AI here. As you think about entering their data centers and saying, “We can envision some new high bandwidth, high capacity workloads in these data centers,” which for the past 15 years we have not really come up with massive new ones outside of streaming video. Streaming video is a lot, but that’s basically what we came up with. Now there’s some new ones in AI that might hit all kinds of new industries. 

Do you think that that is just an exponential rise in how much capacity we need, how much power we’ll need to use, how much networking capacity that we’ll sustain? Or is it, to make the comparison earlier to dark fiber, building out a lot of capacity that we’ll eventually use over time?

That’s a great question. Having lived through the previous period of the telecom nuclear winter where we built out all this capacity — and as it turned out, no one came — it did facilitate all these business models that wouldn’t have accelerated without that glut of capacity. I get asked the question if we’re in another bubble around all of these data centers. Specifically around the data center piece, I think there’s an insatiable demand for compute, to build out compute power. These things always take longer than we all think to monetize and the rest of it, so we’re probably going to go through some iterations. But to answer your real question, I’m actually of the opposite view this time round. I think we’re underestimating the amount of traffic and bandwidth that will be required to go do this because we’re already seeing a clip up in cloud traffic. We’re pretty well placed to see that.

Now, how much of that could I attribute to AI? Some of it is, but I’ve got no way of really discerning which bit is which, and it’s not our network. But there’s a definite uptick in cloud traffic just to connect to all of these data centers. We are absolutely seeing that, and I think that’s going to continue because you’ve got a lot more data centers coming online in the next 18 months, or two years around the world. If they’re going to monetize any of that, they’ve got to all be connected for learning, training, and the rest of it.

That’s before you even see the applications for medical use or verticalization, whatever you want to call it. Using your ChatGPTs is generating traffic, sure. But then you start to get to any of that medical usage, for example, and the amount of graphics that are required. Graphics are about 10–20 times more bandwidth-consuming than normal data. So, if you start to get to the monetization of that through these vertical applications, I think we’re underestimating the amount of traffic that’s going to be there.

When you plan out growth for the company — obviously, you’re a public company, so you have to do it out in front of everyone and talk about risk — in your most optimistic view, is it sustainable, the rate of demand that you see now? You’re saying we’re underestimating it. I look at it maybe a bit more cynically, which is, “Wow, we are building a lot of capacity against business models that don’t exist.” OpenAI has yet to turn $1 in profit, and it wants all of the power that the world can generate. It wants every GPU that Nvidia can build. That’s a lot, right? There’s an inherent promise there that this will all pay off. 

Maybe you would call that a bubble or maybe you won’t. I’m just looking at it more mathematically. A lot of companies are going to sell a lot of things to these data centers. We’re going to build a lot of capacity, and maybe at some point we will have built enough or not enough or something will happen, but that will come to an end. Then a company like yours is saying, “Well, we’ve had this rate of growth and now maybe it’s plateaued or maybe it’s going down.” How are you thinking about that?

I think about it like this: Don’t confuse compute capacity with the network. The network is always the afterthought to all of this, right? It’s, “Oh my gosh, I do actually need to connect these data centers.” So what’s the hierarchy of their thought process? It’s, “I need the compute. I need the power. I need the space.” Those are the things that are driving their decision. Then by the way, “Oh my gosh, I need to connect these data centers right now.” I’m being flippant, but it is down there. I’d like to tell you it was right up there, but it isn’t. In the answer to this question, I think that’s actually a good thing. We only get there when they need it. They’re not building out these enormous capacities because we see it. We’re putting capacity out there, and they’re putting traffic down it.

So I take real comfort from that, having lived through 2002 and 2003 when our revenues went down from something like $1.6 billion to about $300 million in one year. I’ve still got the scars from that, so I’m particularly sensitive to this. This is not the same dynamic. They are actually putting capacity out, and as fast as we put it out there, traffic is coming through it. To your point and to what a lot of other folks are saying, is it going to get overheated in terms of the GPU stuff and the data centers and the rest of it? In my personal opinion? Absolutely. Will there be a bit of a tail off in all of that? Absolutely. Will it take longer for applications in the AI world to truly generate the economics? Absolutely. But we’re not banking on that. We’re just seeing this nice steady growth in bandwidth, which I think will pervade throughout.

Here’s another way of answering your question. Over the last two decades, bandwidth has grown about 20, 30 percent a year and is not cyclical. When you think about it, it shouldn’t be cyclical, should it? We’re just putting more and more applications out, we’re using our phones for more stuff, more people are getting phones, and all these devices are now connected. That has been very steady and consistent throughout. Even when you get an economic downturn, it continues at about that pace. What we’re talking about here is the advent of AI on top of that. Exactly what is it going to be? No one really knows. There’s all these models, but no one really knows. But it’s reasonable to assume that it’s going to come on top of that and could push us to 30, 40 percent bandwidth growth per year, and I think that’s going to be pretty sustainable. When you think about all the promise of AI, the real trick to it is you start putting any graphics down it.

Every time you hear [Google CEO] Sundar Pichai say multi-modal AI, I can see the dollar signs floating through the offices over there.

I love it. We’re basically planning our business on what we see, and we’ve taken our growth rates up. Typically, this business has grown. We’ve grown about 7 or 8 percent. We’re now saying it’s going to grow 8 to 11 percent over the next three years, and that’s without any massive entry into the data center space or anything else. It’s just what we see from a cloud growth point of view.

You’ve talked a lot about entering the data center using the same core technology that you use for the undersea cables and the telecom providers. What are the barriers to doing that and what were the things you assessed when you thought, “Okay. This is the market we have to address”?

The barriers to entry there are that you’ve got a technology called direct-connect, which has been well ensconced there really since the start of communication in data centers. You’ve got an ecosystem there that is going to be designed to elongate that for as long as possible, and the customers want to keep it for as long as possible. With the analytics we’ve done on it and working with them, they think that there’s going to be an intersection point where the physics of that gets to the point where it’s not sustainable. We think we’ve got to the point with power and space where there’s a good intersection. So it’s the analytics around that and working with two or three of these players who are already our customers but outside the data center.

Is that them coming to you or are you going to them?

A little bit of both. We have strong relationships with them, so that collaboration has been there. We’re realistic about it. The existing way of working and that direct-connect? That’s not going away. That’s probably still the majority of applications for it. So you’re looking at what the subset of corner-case applications are for this kind of high-speed connectivity. But even a small part of that is massive for us and massive from a scale point of view given the amount of spend that’s gone on in that space. It’s always about taking risks when that market isn’t there right now, but if you wait until that market appears, it’s too late. If you’re waiting for, “Oh, what’s the market share of this and what’s the growth rate?” At the end of the day, you’ve just got to go with what your intuition is and what you’re hearing and seeing from the customers and the rest.

So we’ve got our first iteration of technology for this coming out at the end of the year. That’s a bet that we placed two, three years ago. As you know, the timing takes a while with some of these chip designs and stuff. You have to guess where the market’s going to go, not where it is right now. There’s always an element of risk to it. Listen, having gone through some of this before, you’re never going to get it completely right. You just have to be in the right vector, pay attention, and tweak along the way around it. So we’re listening intently.

Gary, this was an incredible conversation. You’re going to have to come back sometime. I could talk to you about this for hours and hours.

Love to. I enjoyed it, Nilay. Good questions. Good stuff. 

We’ll see. We should have you back a couple of years from now and see how stressed you are about these undersea cables. I think that would be very interesting.

We’re certainly in for some entertainment, aren’t we?

I know what we’re writing about for the next four years. Gary, this is amazing. Thank you so much.