Ciena’s Gary Smith: AI is forcing a global rebuild of network infrastructure

I’ve known Gary Smith for nearly three decades and have watched him navigate the telecom industry through serial cycles of hype, disappointment and reinvention. For almost that entire time, Gary has quietly led Ciena, transforming a once-niche optical networking firm into a cornerstone of the modern digital economy. If you want to understand how data actually moves across the globe—looking past the apps and the GPUs to the actual backbone—you eventually find yourself talking about optical networking. And that inevitably leads to Ciena Corp. 

What makes this moment unique is that optical networking, which usually hums along in the background, is suddenly front and center. While the AI explosion has fixated the world on compute and power, connectivity is the other half of that equation. Training models and running global-scale inference demands massive, low-latency network capacity between data centers and out to the edge. 

This shift has been a windfall for Ciena. Its stock (NYSE: CIEN) has leapt like a salmon over the last year—up over 300%—fueled by a new AI investment cycle that follows years of operator underinvestment during the 5G rollout. Simultaneously, hyperscale cloud giants are building infrastructure at an unprecedented scale, fundamentally reshaping the dynamics of both the telecom and cloud ecosystems.

As the longest-serving CEO in the telecom industry, Gary hasn’t changed much over the last 30 years. He’s a little more grizzled, perhaps, than last time I interviewed him in 2015 — leaning a bit more “Hemingway” than “telecom suit”—but still the same cool hand at the tiller.   

I chatted with him about his perspective from the center of the current infrastructure shift: why the AI boom is fundamentally a networking story, why hyperscalers are the likely long-term winners, and why a shakeout may still be coming for newer “neo-cloud” players. We also discussed his unfortunate choice of football team. 

The interview below is largely a verbatim transcript of our conversation, with only light editing for clarity. 

Steve Saunders: Gary, it’s been a while. How are you feeling about things at the moment? 

Gary Smith: It’s extraordinary times. I don’t think anybody really predicted the network impact of AI in its various forms. It comes after a challenge in the market through COVID, through supply chain, through underinvestment in the networks in the last five years, in my view, generally across the globe. So it’s certainly welcome. 

Steve Saunders: A lot of the success at Ciena is predicated on hyperscaler business and the build-out of data centers and the interconnections between them. Is this a case of being careful what you wish for? A lot of the customer commitments are debt-based and planted in special purpose vehicles, which disappear if the hyperscalers change their mind about needing all of these LLM training facilities. Are you worried about that at all?

Gary Smith: If I look at the overall environment that we’re seeing, first of all you’ve got service providers. That’s half our business, and that business is very solid and growing at a nice clip now. That’s underinvested. Most service providers around the world have been so distracted by 5G, which has been a bit of a bust, that now they’re saying, “Oh my gosh, I’ve got to invest in the optical core infrastructure,” which they haven’t really invested in for the last five years. So you’ve got that dynamic going on, which is a super helpful dynamic. 

Then you’ve got the hyperscalers. What we’ve seen with them generally is that for the last five to ten years, we’ve been connecting their data centers around the world, collaborating with them on their submarine [cables], because they now own more submarine capacity than anybody else in the world. That’s how the relationship began with them.

When we think about this whole AI piece, it’s often classified as homogeneous. I’m not sure that’s really the right way of looking at it. You’ve got two broad categories. You’ve got the hyperscalers, who are well-funded, well-capitalized and super profitable, and they’re the biggest users of AI in the world themselves. Then you’ve got the emerging players, the neo-scalers, which have a different set of financial dynamics.

Not everybody’s going to be a winner in this. Some are going to win, some are going to lose. The larger hyperscalers have the scale and the technology and the applications to be the winners. Not every language model is going to survive. They’re not all going to get funded. And how are they going to monetize? That’s also an issue. So you can’t really look at it in one big homogeneous way. 

I think this is very sustainable. Why do I think that? Because the network has fundamentally been underinvested in, in the context of what needs to be done for AI. Everybody’s been so focused on GPUs and power and all of that stuff for obvious reasons, and now it’s: “This stuff’s going to move out of the data center, both for training and for inference. We need a network to do that.” And by the way, it needs to be super high speed, it needs to be global, low latency. These are all the dynamics in which Ciena prospers.

Steve Saunders: The way I look at it, there may be a bubble in neo-cloud, but that doesn’t mean there’s a bubble in AI in general.

Gary Smith: I think there’ll be fallout here for sure along the way. You can’t just continue with everybody winning and just gaining money and market cap and the rest of it. The world doesn’t work like that. But I do think the underpinning—who are going to be the winners and the rest of it—I have to feel that the hyperscalers generally just have a tremendous advantage. Are they going to get it all right? No. But I do think the underpinning is real. 

Steve Saunders: There are multiple ways to slice up the market. One is core LLM and training, and another is edge inference and application. To me, that edge is the carrier play. Yes, they need hyper-fast networks to carry hyperscale traffic, but the edge is where AI actually makes money or does interesting things. What’s the edge play for Ciena?

Gary Smith: The edge play is multifaceted. One, we’ve got a massive installed base into the service provider piece. Number two, we’re a predominant player with the cloud providers for their own cloud-enablement networks anyway, so we’re already a major participant in that inference and the build-up of that at the edge.

The carriers have an opportunity at the edge of the network here because, to your point, as omnipotent as these hyperscalers are—and they are—they can’t be everywhere. They can’t be into every single enterprise fiber, into every single person’s home around the world. You need carriers for that, and I think it’s a tremendous opportunity at the edge of the network for service providers.

Steve Saunders: Hyperscalers historically aren’t very good at the edge. They’re not interested in being last-mile operators. Without that last mile, we really run the risk that AI stalls. And now the edge of the network isn’t the edge anymore. The edge has moved out into enterprise and industrial verticals everywhere—energy, electricity, all of those things. 

At the same time hyperscalers have unlimited IP, unlimited money and unlimited arrogance. It’s tempting to over-rotate into that. I talked to one of your customers the other day, and they said that after you announced this refocus on AI, they never heard from their Ciena salesperson again, and I think that saddened them. Is there a danger of burning bridges by going after the AI core market because there’s so much money there? 

Gary Smith: The two markets are symbiotic. We’ve served both, and we’ve served both for a long time. I can’t comment on that particular instance you got, but yes, that is sad, and that shouldn’t be the case. We’ve not reduced our service provider support—in fact we’ve increased our amount of service provider resources around the globe, and that business continues to grow. 

The unusual thing about Ciena is that we have a core technology that we develop both for hyperscalers and service providers, but it’s the same core technology. Everybody benefits when we’re pushing the envelope on technology. Before the hyperscalers got into networking, we would develop the technology based around service provider needs and wants. We still do that.

But the hyperscalers are at such a scale and want cutting-edge technology. The service providers are benefiting from that because the technology we develop for them is being used in service provider networks. The economics and the new technologies are available to both. 

Our service provider business is scaling very well. To us there’s no difference. We’re providing the same line systems, the same pluggables, the same technology at the same time into both. I actually think service providers are benefiting from the amount of scale and investment from hyperscalers.

Steve Saunders: That does feel like a Formula One analogy—technology developed for race cars at huge expense eventually gets incorporated into family cars.

I’m also seeing changes in hyperscaler culture. If you talk to the Google Fiber people, they’re extremely pragmatic and focused on reliability and trust. That’s a new look for hyperscalers. At the same time, within the vendor community, there’s a new aggression around going after these markets. 

Where do you end up at Ciena? Do you want to be more like a hyperscaler, or do you want to continue to be a telecom equipment provider?

Gary Smith: It’s a great question. For us, it’s about the technology that we’re developing, and we’re very focused. We’re the only player really that’s just focused on high-speed connectivity. That’s all we do. 

This is great for us because we get to push the envelope. We have massive markets for what we do. We bend physics on a daily basis, and now we’re getting a bigger market for it.

We view it pretty simply. We can push bits faster, longer and with better economics, with lower power than anybody else in the world. That’s what we do, and we’ve done that for service providers and cloud players.

Culturally, Ciena has always been at the cutting edge. We’ve always been very focused. The cultural piece is innovation velocity—that’s really the essence of Ciena. Most of our competitors worry about mobile and other things. We don’t.

I think the world is moving to best-of-breed. When you’re a hyperscaler looking for massive scale, that’s why Ciena is attractive to them. 

Steve Saunders: There is also a broader global ecosystem around AI infrastructure—water, electricity, regulation, sovereignty. The U.S. is clearly leading in LLMs and AI factories but not necessarily in every layer of the stack. What needs to happen for the U.S. to lead across the full stack? 

Gary Smith: I don’t think anybody has a single answer to that, but the U.S. is certainly the innovation engine around core technology, and I would include optical in that. But this is a global business.

If you take optical, for example, where do we get our chips from? We get them from TSMC because we use cutting-edge technology. We use an ecosystem around the world of components to put these solutions together. 

So it’s very much a global business. And when you get into questions around sovereignty and regulation, that’s going to require a lot of input from many countries. It can’t just come from the U.S. It has to be a team game. 

A lot of the training on main language models is currently being done in the US—it’s my understanding that from a network point of view, training is pretty much exclusively in the U.S. But all of them have plans for how to do this in other countries while respecting sovereignty. 

Steve Saunders: Before we finish, I have to ask about something else entirely—your choice of football team. 

Gary Smith: I didn’t really get a choice. My father was born near the ground, my grandfather was born close to it, so supporting Aston Villa was sort of inherited. Supporting Aston Villa F.C. also helps build character, as my two sons will gratefully attest. [Edit note: Gary is a lifelong supporter of Aston Villa, who are based in Birmingham and generally acknowledged to be quite rubbish and lucky to even be playing in the Premiership. Steve, on the other hand, supports the legendary Crystal Palace Eagles of South London, who are universally acknowledged to be a far superior club and beat Villa 3-0 in the 2024-25 FA Cup semi-final at Wembley last year].  

Steve Saunders: You are playing a long game there. 

Gary Smith: You have to be in it to win it.