You’re Scaling on Infrastructure You Don’t Control

You're making infrastructure decisions on a system you can't see. It's fast, stable, and invisible. That's the problem.

Underneath it, global capacity is planned years in advance, routes are shaped by geopolitics, and resilience depends on infrastructure you don't control. Max Clark sits down with TS Narayanan, CTIO of EXA Infrastructure, to explain the layer most enterprise buyers never see — and why it quietly determines whether your cloud strategy, your regional expansion, and your infrastructure commitments actually deliver what you paid for. They get into why AI isn't driving the traffic spike everyone assumes, what geopolitical risk actually does to a network route, why adding nodes costs more than buyers expect, and where the real constraint is right now (hint: it's not bandwidth).

If you've ever signed a provider contract without fully understanding what sits underneath it, this is the episode that fills in the gap.

WHAT WE GET INTO
00:00 — The network layer your vendor isn't explaining to you
04:00 — Why capacity is locked in years before you make a decision
09:30 — The AI traffic myth: what's actually growing on the backbone
12:30 — Why your cloud usage growing doesn't mean what you think it does
15:40 — Bandwidth announcements and what they're not telling buyers
17:50 — How traffic patterns are shifting — and what that means for your architecture
20:40 — What hyperscalers are doing to the build equation (and what it means for everyone else)
23:50 — How subsea cable routes get planned and why it's harder than it looks
27:50 — Geopolitical risk is real: what the Mediterranean actually shows
31:30 — Why LATAM and Africa are underserved — and what's finally changing
35:20 — Data sovereignty: what compliance actually requires from your infrastructure
41:00 — Capacity vs. latency: which one should be driving your vendor decision
46:50 — The complexity cost of every node you add
52:30 — What you're actually buying when you buy network capacity
57:30 — Control vs. outsourcing: how buyers are splitting on this decision
1:02:00 — Power is the real constraint. Not bandwidth.
1:08:30 — Why infrastructure is always in upgrade mode and what that costs buyers
1:18:30 — Network as a Service: what it actually means and where it's going


WHAT WE MENTIONED

• EXA Infrastructure — https://www.linkedin.com/company/exa-infrastructure/
• TeleGeography subsea cable maps — https://www.submarinecablemap.com
• GÉANT (European R&D network, referenced in EU connectivity context)
• AMS-IX, DE-CIX (European internet exchange points)
• IRU (Indefeasible Right of Use) — the long-term capacity lease structure common in subsea contracts
• OVH, Hetzner (EU-based cloud providers referenced as demand drivers)

ABOUT TS NARAYANAN
TS Narayanan is the Chief Technology and Information Officer at EXA Infrastructure, one of the largest fiber and subsea backbone networks spanning Europe and North America. He's spent his career across enterprise IT, systems integration, and telecom — which means he's seen infrastructure decisions from both sides: the buyer who has to live with them and the operator who builds and maintains them. That makes him unusually direct about where capacity planning goes wrong and what buyers should have asked before they signed.

LinkedIn: https://www.linkedin.com/in/ts-narayanan-62b3617/
EXA Infrastructure: https://www.linkedin.com/company/exa-infrastructure/

ABOUT THE SHOW
Signed is the podcast for buyers in a market built for sellers. Host Max Clark, CEO of ITBroker.com, sits down with CIOs, CFOs, operators, and founders who've lived inside real enterprise tech deals.
New episodes weekly at itbroker.com/podcast.
If you're in the middle of a real tech decision and want someone in your corner, book an intro call at itbroker.com.
Buy tech without regret. Follow: @itbrokerdotcom

FULL TRANSCRIPT

Creators and Guests

Host

Max Clark

Founder & CEO of ITBroker.com

What is Signed?

The IT market is built for sellers, not buyers.

That's why 80% of tech buyers regret their last major purchase. Deals take longer than they should. Teams get locked into platforms that don't fit, contracts they can't escape, and vendors they wouldn't choose again. The pitches, demos, and analyst reports are built to close deals, not help buyers make the right one.

Signed is the podcast for the buyers. Host Max Clark, CEO of ITBroker.com, talks with CIOs, CFOs, operators, and founders who've lived inside real enterprise tech deals — the ones who can explain what actually determined whether the deal worked.

Plus weekly Playbooks breaking down the moments that matter most: renewals, M&A, compliance mandates, office moves, budget cuts, and the specific plays that separate buyers who get it right from those who regret it.

If you're responsible for choosing, negotiating, or living with the consequences of enterprise technology, this show is for you.

New episodes weekly. An ITBroker.com podcast.

Max Clark: [00:00:00] Is this creating a situation, a potential where we're just, you know, overbuilding again, and who knows what happens?

TS Narayanan: It's an interesting question. Again, if you look at the traffic dispersion and the traffic patterns today and the demand and supply, AI is not the highest. You look at all kinds of reports out there on the ether, AI-orientated traffic is still in the low teens when you look at the amount of traffic going across the backbone network.

There's still a whole host of cloud adoption. There's still a whole host of streaming traffic. Based on what we're seeing today, it's still nascent. So is there a danger of overbuilding?

Max Clark: I've got a bunch of questions for you. Go ahead. And, and I've been thinking about how to, how to, how to phrase this and talk about this. And, and the first thing is, is like I'm an old network engineer. I love networks. I, I, you know- [00:01:00] Mm-hmm, mm-hmm ... lock me in a room, give me a whiteboard, let me build networks the rest of my life, I'd be really happy.

TS Narayanan: Mm-hmm.

Max Clark: But for the most cases, network is this like, you know, to our, you know, our sidebar earlier, is like this invisible commodity that most people don't think about.

TS Narayanan: Mm-hmm. Mm-hmm.

Max Clark: But they're so critical. And so as a, as a network services provider-

TS Narayanan: Mm-hmm ...

Max Clark: you know, how do you like relate that to people in terms of your role for your clients and, and the internet at large versus this, this kind of like invisible element that's out there?

TS Narayanan: It's a very interesting question. If you look at telcos like us in particular, we're building the most basic infrastructure, right? We are building the fundamental optical infrastructure that carries about everything. In some ways, it looks commoditized, it looks boring, but it is the infrastructure that helps our clients build and innovate on top of that.[00:02:00]

It is, at one level, the most basic infrastructure, optical networks. We are talking about optical cables, wavelengths, ducts, pipes, subsea networks. Sounds boring. And then you look at the other, uh, aspect, a big chunk of the world's internet traffic flows on this network, a big chunk of hyperscaler traffic, a big chunk of content, everything flows on this network.

So at one level, we are building fundamental basic infrastructure, but what travels on top of that is super critical. So that's one way of looking at it.

Max Clark: Today, networks are-- I mean, fiber, fiber is very resilient. It's very fast. You know, the average experience for somebody, whether it's residentially or commercially, is really good.

You know, you think back even 10 years ago and, you know, redundancy was a much bigger thing for the average corporate user And that's a big testament to the, to the [00:03:00] actual network operators and how good things have gotten. Now, you say, like, you know, exciting, not exciting. Subsea is very exciting to me, and I think that's just part of also because I've, like, watched videos of ships laying subsea fiber.

And, you know, there's, like, this sense that I get with people that this is like networking is a solved problem. What I mean by that is like, oh, there's nothing new and exciting coming or nothing new and exciting that we can do. But then I look at, you know, these maps from, like, Telography, you know, or these da- that show these subsea systems, and you're like, there's lots of ocean out there that we still don't have cable in.

So how do you think about markets and think about cable routes and, and, like, are-- you know, where cable needs to be versus do we invest here? Do we not invest here? You know, is this customer der- demand driven? Is this build it and they will come driven? Like, how, how do you guys approach that?

TS Narayanan: I think it's a combination of a number of factors, and things are getting quite dispersed today.

Max Clark: Mm-hmm.

TS Narayanan: Right? So [00:04:00] what's happening, you have the traditional demand and supply question. And where's demand going? So you, you still have a great deal of cloud adoption. You still have a great deal of streaming. You still have data centers growing all over the place. And then you have the most famous buzzword of today, AI.

So there is a need to build connectivity to where computing is being done across the world. And that drives a lot of demand, and that drives the footprint for telcos like us. So to answer your question, a lot of it is driven-- It's a combination of both demand and our prediction of where demand is likely to be.

So there's, there's, there is a, an element of follow the hyperscalers, follow the data center footprint, and then follow what we believe [00:05:00] could be demand for the future.

Max Clark: Bandwidth consumption from an actual end user, from, from whether it's a user, their device, whatever it is, um, has relatively flatlined on a per-user basis.

And what I mean by that is, like, the densest, um, the densest traffic the average person can consume is streaming media. And there's lots of different reasons for this, but, you know, codecs for video delivery, now this isn't like gigs and gigs and, or gigabits and gigabits of data being flowing to an individual consumer.

And we see coax and DOCSIS networks increasing and going symmetric and GPON-based networks for residential increasing. Now we have 10, you know, GPON X and 10 gig GPON. But, you know, running a s- running a network, a fiber service provider years ago, there was this thing that you knew, which was, oh, you, you could deliver a gig E to somebody or a 10 gig circuit to somebody, and if they were an eyeball, you know, a con- like a consumption, they weren't using anywhere near that bandwidth.

And- You know, equipment manufacturers wanna tout, of course, [00:06:00] hey, we're, you know, 400 gig, 800 gig, terabit plus network links. And the-- But then in the back of my head, I always think about it. It's like, okay, well, what's, what's actually... Are we just talking about, like, saturation of people is increasing? I mean, the United States is relatively saturated in terms of the amount of people on broadband.

I mean, it's gonna increase somewhat. Mm-hmm. But we have other population centers worldwide that are either still non-connected or haven't, I mean, Africa, right? Like very limited penetration of real broadband. Um, but then I see these press announcements of like, "Oh, we've just upgraded our backbone to 400 gig.

We've just upgraded our backbone to 800 gig." And, and a lot of that, I mean, how much of that is just really, you say, like data center and hyperscaler demand of just, you know, bulk data transfer flowing between data centers versus actually getting out to a, you know, an eyeball?

TS Narayanan: I think that's a very interesting observation, right?

And you look at it from different lenses. You have the B2C lens, which is what, like you said, an end user would see, right? Uploads and downloads, streaming [00:07:00] content. And we say, hey, you know what? The end user consumer broadband market, there is only so much demand. There's only so much Netflix you can download or watch, and does that define the ultimate requirement?

And then you look at the B2B space, and there's two elements. Traditionally one element, which is the north-south traffic, which is the traffic between the user and the cloud, and enterprise and the cloud, um, which is, hey, you know what? Um, I'm a B2B enterprise, and I have SaaS platforms. And there's a lot of traffic going back and forth from the data center, um, to the SaaS platform, to the end users, to the SAAS plat- uh, SaaS platform.

We call it north-south. And now, that's contributed to some extent to having to build a backbone network, a high-capacity backbone network to connect [00:08:00] enterprises, data centers to the cloud, to other data centers simply because of the application workloads, whether they're SaaS, on-prem. And that traditionally has been called north-south traffic.

And then now with what's happening with AI, there's the concept of the east-west traffic, is how much traffic, particularly related to AI, is flowing between distributed data centers. It's in early stages, but then you look at the volume of data that needs to be potentially carried for AI training.

Max Clark: Mm-hmm.

TS Narayanan: That's huge, and which is why telcos and operators like us are having to upgrade- Backbone networks, whether they're terrestrial or whether they're subsea, to be able to be ready to carry this order of magnitude of traffic.

Max Clark: There's [00:09:00] a certain fear of this is, you know, we're running into the next big technology bubble, right?

You know, is this- Mm-hmm ... the dot-com 2.0 bubble, right? Mm-hmm. However, it's gonna get- Mm-hmm. You know, this is the AI bubble. And, you know, late '90s there was a lot of capacity and a lot of fiber and a lot of construction, a lot of build, and it was all this thing of like, oh, we've got to build to get ahead of demand, and then a combination of demand not showing up or demand evaporating 'cause it was just too early.

I mean, being too early- Mm-hmm ... is the same as being wrong in tech, right? So, you know, y- you know, this, this building for AI without really, like, a definition of what does this actually mean, you know, is this creating a situation, a potential where we're just, you know, overbuilding again, and who knows what happens?

TS Narayanan: It, it's an interesting question. Again, if you look at the traffic dispersion and the traffic patterns today and the demand and supply, AI is not the highest. You look at all kinds of reports out there on the ether, [00:10:00] they will always tell you that, probably tell you that, and my numbers may not be 100% accurate, but I'm not a million miles away.

AI-orientated traffic is still in the low teens when you look at the amount of traffic going across the backbone network.

Max Clark: Mm-hmm.

TS Narayanan: There's still a whole host of cloud adoption. There's still a whole host of streaming traffic. So based on what we're seeing today, it's still nascent. So is there a danger of overbuilding?

Not right now. Not right now with the current trends, even pre-AI. Right? Now, if AI lives to its promise in terms of being the guzzler of compute and storage and power and traffic, then it's a different ballgame. It's early stages, and I think most of the telcos are [00:11:00] looking and watching and saying, "Hey, this does hold a lot of promise," but there is a need for a lot of backbone capacity even in the early stages of AI.

Because simply, traffic loads from other sources are still growing.

Max Clark: Um, you know, the, the race with subsea cables Global Crossing started was driven, I mean, there, there was a couple components of it, but a big component of it was decreasing latency, right? And now when we're talking with clients about networks and network architecture and design and links, you know, part of the conversation's always latency and part of the conversation's in bandwidth and consumption.

You know, what's the actual speed of the pipe? And as a, as a telco, as a service provider, how do you balance those two? And what do you see from client demand now in terms of like, are they chasing latency? And if you turn around and say, "Hey, we can shave off X percent of latency," you know, that's gonna open up a, a new market.

Or is it really like, "Hey, you know, we need this upgrade from 100 gig to [00:12:00] 400 gig or 400 gig to 800 gig," and that becomes the driving factor?

TS Narayanan: It's a mix of both because it's really dependent on the applications that are driving the need for capacity. And again, you can look at it pre-AI and during AI.

Pre-AI, there's still a need for large bandwidth simply because of the north-south traffic that's still growing, right? Which is the cloud traffic, the hyperscaler traffic. And then pre-AI, there is a need for low latency for specific sectors like trading. Anything that's real time. Doesn't have to be all AI.

So there is a mix. There's a demand for both capacity and for low latency, and it's a horses for courses situation, and say what application requires low latency or [00:13:00] high capacity, and then your networks are designed and structured as such.

Max Clark: There's, there's another interest. For somebody who hasn't built a network or managed a network, right, this wouldn't be front of mind. But in networks, you've got this interesting teeter-totter, right, where the more endpoints and nodes that you have on your network, the value of your network increases. But with each node that you add to your network, the complexity of managing that network also exponentially increases, right?

So the fastest networks have the simplest architecture and have the least amount of nodes on them, but then the utility of that network is less than the utility of a network with lots of nodes, right? So how, you know, from, from like a... How, how do you approach this with, with EXA, right? Because, you know, you've got a, a big build in Europe, a lot of edge, a lot of endpoints, a lot of, you know, markets.

EU's a very interesting animal, you know, in terms of like where the population centers are and what countries are [00:14:00] connected. But from a planning standpoint of like thinking about, you know, let's go put something here now, how does that weigh, you know, complexity versus utility?

TS Narayanan: In our case, because we are fundamentally a service provider at the most fundamental level, as you start bringing in layer two, layer three Ethernet IP, then networks get more and more complex in terms of the nodes. Look at a telco like us. There's a subsea element which is really, at the end of the day, hey You have a big body of water in between, and then you have nodes on either side.

From a number of nodes perspective, it's fairly simple. And then when you look at the terrestrial side, because we are operating at the optical layer, it's really about designing nodes, endpoints, and [00:15:00] connectivity where your demand is. And then you always build networks. It's not, for the majority, a hopping concept.

So I have a multi-city network, but I can build an optical path from A to Z without having to hop through B and C and all the nodes in between, right? If you know what I mean, from an optical path perspective. So there is an element of simplicity and there is an element of complexity. Complexity comes from the need to cover a footprint and address demand.

And simplicity comes from the nature of the network, given we are operating at the infrastructure level. So if you ask me a question for an IP network, the complexity is much higher. And then you look at the traditional mobile networks. You look at mobile networks from 2G [00:16:00] or 3G or 4G or 5G, it's always been about how do you manage density of network elements to provide greater capacity and coverage versus the complexity that comes in, right?

You need more base stations in central bus- business districts simply to have greater capacity and coverage and balance that with the complexity. If you decide to have a f- a simpler network, it comes at with the compromise of either capacity or coverage.

Max Clark: Um, you're building right now, um, a terrestrial route through Turkey.

And that's... There's, there's, this is, this is part-- Let me, I wanna, I wanna think about how I phrase this question, right? Because if you're a technologist, right, there's like this very, um, egalitarian view of the world, right? Like, it's just like, "Oh, of course, let's go out and put a link here because we [00:17:00] need the capacity, we want the bandwidth.

This makes sense," right? But when you start dealing with geopolitics and the demands of geopolitics, and then, uh, we've done a lot of work, for instance, in Asia, right? You know, and like landings in different countries in Asia get very complicated because you have national sovereignty and, and who owns what asset and where does the fiber...

You know, how, how does that impact like ultimately the design and the build and the reach- of a network as you, as you look at your map and f- and figure out, you know, which way do we want to expand and, and what markets do we need to serve. Because there are major financial hubs and major industries that are, that are popping up in other markets, um, you know, that, that more density of connection and route are, are, are gonna be massively valuable.

TS Narayanan: Oh, absolutely. It's a complex decision and yet a simple answer in the sense it's always going to be a function of risk [00:18:00] versus the ability to grow.

Max Clark: This show exists because of what we do at itbroker.com. If you're in the middle of a real tech decision right now, new technology, vendor selection, a contract that doesn't feel right, an M&A event that just landed on your lap, and so on, we help buyers like you get it right.

Independent strategy, sourcing and contract negotiation. No kickbacks, no sales quotas, just someone in your corner. Schedule a call at itbroker.com. Back to the episode.

TS Narayanan: We're not going to be able to wish away geopolitical challenges because that's the way the world is. And then how do you balance and trade that off with the need and the demands to provide a network footprint, provide connectivity, or collaborate with others that come in from other regions of the world to provide an end-to-end service?

At the end of the day, it will become a risk management call in terms of what do we want to do and [00:19:00] how far is a telco willing to go in terms of building that out.

Max Clark: From a multinational client's perspective, you know, a company that has to provide connectivity into different markets and, and figuring out how to, how to do this, um, how much of that is problems that they're working through and figuring out and saying, you know, "I'm gonna go and use network vendor A for this and network vendor B for that," which adds complexity, right?

'Cause now you have contracts and legal overhead management, all these different things, versus being able to come, you know, come to Exa and say, "Hey, you know, this is, this is the footprint that we need to serve, and how do we maximize a relationship with you in order to serve this footprint?"

TS Narayanan: Yeah. And, and it's a...

Clients tend to fall in two extremes, saying, "Hey, you know what? I am going to outsource the problem." And you find a telco like Exa to say, "I'll go and [00:20:00] not only be the dominant network provider and also be the integrator with all the off-net components that you need to build the end-to-end service of a network footprint."

Or the client chooses to be the integrator and say, like you said, "I have, you know... I know telco A is strong here and telco B is strong here, and I'll be the integrator." So both models do work. And it's very difficult to say which one is good or bad. It's again horses for courses. And then both the client and the telco work collaboratively to find the right solution.

Max Clark: One of Jim Cramer's famous predictions was that the data center d- companies were dead because cloud was here, right? You know, of course, missing the fact that the cloud sat inside of data centers. But we see a similar thing with, you know, hyperscaler and big tech companies that have now all entered the market of, [00:21:00] of their own routes and their own capacity, you know.

And pick any of the, you know, the FANG stocks, for instance, right? FANG's not right, but, you know, Microsoft's in the game, Google's in the game, Meta's in the game. Yeah. You know, and, and some cases they're doing their own links, some cases they're partnering to build links, some cases they're buying capacity, some cases they're, you know, they're committing reven- you know, you know, committing revenue for somebody else to go out and build.

How has big tech changed this game for the telcos? Because they're both your client and your competitor, you know, in a lot of these markets.

TS Narayanan: Given where the prediction for demand is, I would actually answer this question in two ways. One, yes, the big tech are making a lot of, uh, investments in network expansions, in building their own networks, equipment, and routes. And then you have the traditional telcos. But the fact is that in today's [00:22:00] world, a lot of it is still consortium, consortium-led.

Mm-hmm. Simply because if you take a big subsea project, it takes a long time and a lot of investment to roll out a subsea route. And even with the hyperscalers or the MANGs, if you would like to call them that these days, are always looking most of the time for the consortium model, where it's always multiple parties working to collaborate to either roll out the network or to share capacity.

And this is particularly true, as you've seen, for the subsea industry, right? The terrestrial world is, is, is probably a little different, where most, uh, individual players build out their own backbone networks and their terrestrial footprint networks. So there's a [00:23:00] little bit of a, um, uh, a differentiation, right?

Subsea tends to be a lot of consortium-led, yes, with big footprints from the hyperscalers, and then terrestrials, terrestrials still are the traditional telcos that build out their, their, uh, terrestrial networks.

Max Clark: You know, terrestrial networks, you're identifying a route, have to buy a right of way or get access to the right of way.

Usually train, you know, like most common train tracks. Follow the train tracks, right? Why, why does a fiber route go the way it goes? Everybody-- If anybody asks me, it's just look at the train tracks, right? 'Cause you need to be able to have conduit run along it. And then you're talking about, you know, now, uh, the ILAs or the regens, you know, the spacing between them is increasing, you know, because technology's better.

But you still have to, like, what kind of physical plant do you need to have at what interval in order to run and power this network? Um, y- you know, so terrestrial kinda makes sense. Like, it's easy. It's like, okay, what do we have to do and build here, right? And then [00:24:00] you think subsea and you're like, oh, we need to pull a cable, a continuous cable between two parts of the planet through the ocean.

And I don't think a lot of people realize that that cable, for most of it, is just like, like drifting in the current, right? Not actually buried. And, you know, the amount of energy it takes in planning and prep and then deployment, I mean, these are multi-year projects. Then, of course, you're talking about, like, what's your life cycle?

What's your upgrade? How do you maintain that asset? So from a, from a capacity, you know, like terrestrial networks, it's easy to send somebody out and, like, go and change equipment. I mean, and like when I say easy, like it's complicated, but it's, it's easy. You know, when you look at, like, your subsea assets, what is your, your, your planning and refresh and upgrade, you know, iteration cycle with this?

You know, how many years forward are you looking? What kind of capacity planning do you have to get into? How do you even-- I mean, how do you take a guess of like, you know, if you're doing a five-year cycle, what [00:25:00] kind of upgrade is necessary?

TS Narayanan: Absolutely. And it is important to get the planning right, right?

So we are always looking at a number of factors. What's the current capacity? What's the kind of contracts we have? Because again, then you get into the commercial realm of what are you doing leases, long-term, short-term. You're doing IRUs. How is your capacity spent, and how are you going to-- What's your customer pipeline?

What does the market demand look like? And then are you going to run out of capacity? And then are you going to partner with other consortium to keep continuously building? Because like you said, if you have a single cable route and it's laid, it's laid. And so these are factors that come into play, and we are constantly having to look to all of these.

It's quite a complex set of parameters, and we have to look at all of these and look at multi-year horizons to say, "This is what it is going [00:26:00] to look like, and this is our strategy for managing capacity, for grooming capacity, for expanding capacity, for partnering with new consortia." And once you put all of these parameters in the mix and then you say, "Hey, you know what?

I think we have reasonable clarity on what a three-year, five-year plan should like look like and beyond. And yes, we've gotta get it right. That's important.

Max Clark: This is, this is part geopolitics and part capa- you know, planning operations, right? So cables going across the North Atlantic or even the Pacific Ocean, let's say simpler geopolitics.

Cables running through the Mediterranean Sea, more complicated politics. Also, it seems to be a lot more accidents of ships dragging, you know, their anchors and, and cutting cables. How do you manage that with a, from a service provider, right? Because you have customer obligations and contracts and, of course, your network stability and health that you have to maintain.[00:27:00]

But there's this, like, pretty big X factor that you have to contend with of what else is happening in these regions.

TS Narayanan: Yeah. And, and I think you can see the trend both from client side and from service providers like us, is in these kind of geopolitical situations or high-risk situations, you'd always plan to build a service with resilience.

And the resilience comes from alternate routes that bypass any potential risky areas or options via terrestrial options to mitigate the risk as much as possible. And that's happening, as we say, across the world, because it's very difficult when you're carrying, when you're carrying critical traffic to not plan to have resilience in terms [00:28:00] of alternate routes or paths, subsea and terrestrial, is not something that customers and the world does not recognize.

So for the region that you pointed out, for example, the Mediterranean, all, to my knowledge, all telcos and clients are building resilient planning into their demand and saying, "I need connectivity from A to B. I know this traverses this region, and what's my backup?"

Max Clark: Western econ-

TS Narayanan: And there are backup coming through the terrestrial network, uh, as well.

Max Clark: Um, Western economies have typically ignored developing economies as relates to, like, network build, right? So, um, North America, very dense fiber. You look, start looking at LATAM, you know, n- y- you know, it's like a network desert in a lot of these markets. Africa, India's, has, has-- going through a lot of development, both data centers and, and network, and [00:29:00] so that's coming up.

But, you know, India and Southeast Asia were also both relatively, like, network deserts. And I'll use that term. And it's always surprised me because having clients in these markets, there is a lot of economy and a lot of revenue and a lot of opportunity, but it feels like there's this massive disconnect of why aren't networks building in these markets more aggressively.

And you look at Africa, you know, and Africa has these subsea networks that kind of, you know, stay and just kind of hopscotch down to South Africa. In South Africa, you can never get enough network bandwidth out because there just isn't a lot of capacity built down to it. And we have the same thing in Latin America, South America, you know, with these networks we've built down in, in Brazil.

Huge, massive population center for, you know, but, but, you know, horrendous networking for many years. Now it's getting-- I mean, South America, Latin America is getting much better now and, and we have, we have a lot of build going on. But, you know, what, what is the disconnect here and, and why, you know, have these locations been [00:30:00] skipped by the, you know, investment decisions to, to build these networks out?

Because there are a lot of people, and there's a lot of cell phones, and there's a lot of commerce, there's a lot of revenue. And, and it, it feels like whoever gets there is gonna benefit from facilitating that.

TS Narayanan: Yeah. I think it's a combination of a number of things, right? Historically, like you said, some of these regions have been, uh, underserved.

And now with the growth of the need for computing and, and data centers being built there. Historically, I would say it's a combination of a number of things. Perceived demand or lack of it, and probably the geopolitics, politics or the ease of doing business in some regions or the lack of it. I think it's a combination of these factors that have probably been a [00:31:00] perceived block, and it's good to see some of those being overcome now, right?

With network penetration and data center growth in some of these regions growing, it's a good sign.

Max Clark: So if a stable government made it really easy for somebody to build a, you know, 10-gigawatt data center project in the western coast of Africa, for instance, you'd think that would drive like all-- You know, pe- people would come and, and consume it, right?

'Cause I mean, you can't put that kind of project anywhere in Europe at this point. You know, the space isn't there, the power capacity is not there, demand's not there. It's an interesting perspective of just the need for space and power driving or, or forcing construction in markets that have been overlooked in the past.

TS Narayanan: Yeah. And, and there's a, there's a new factor that might come into play, and I do not know how much of that will be relevant in [00:32:00] LATAM or Africa or other parts of the world, is the growing need for data sovereignty.

Max Clark: Mm-hmm.

TS Narayanan: Most administrations are coming up with-- Most governments coming up with their own strict data sover-sovereignty requirements, and that's probably going to drive some level of penetration of growth because then that becomes a mandated requirement.

And, and that might also deliver a certain level of push to, um, both data center and network growth in some of these regions.

Max Clark: How--

Data sovereignty is, is an interesting animal, right? You know, and the idea of, you know, whether it's-- Uh, let's just say EU, right? You have an EU dr- mandate for data sovereignty, so we have to keep data in region. It has to be within our, our... But when you're dealing with a, an underlying service provider that then is not-- that has assets or has, has touchpoints with other [00:33:00] jurisdictions, I mean, like how...

You know, it, it feels it's almost like, uh, it's like a wink-wink, nudge-nudge thing. It's like, okay, our data is resident in this region, but people in this other region have access and, you know, and influence over it and, you know, and, and it just, it feels like this is getting messier and messier as opposed to actually easier and better.

TS Narayanan: It is going to get more difficult in terms of being able to access data across regions simply because of the regulations that come in. It's not an insurmountable problem, but the industry is going to have to build additional controls to be able to manage that in a way that meets the regulatory requirements.

Yes. Does it make it more complicated? I completely agree with you.

Max Clark: OVH and Hetzner are becoming very popular, um, you know, public cloud providers for people looking for [00:34:00] compute. Obviously really good for, for Exa in terms of, you know, capacity, demand, and growth, right? 'Cause now you have in-region European-based customer with a lot of stuff coming back to the United States in terms of data capacity, right?

Like, this is your network footprint, you know, it's like your core asset. Do you feel like the EU is, is moving in this direction of, of actually gonna position themselves to be more competitive in the startup space with the United States?

TS Narayanan: I, I think there are some good, uh, signs of that happening. For example, you've seen, uh, that we signed a deal with GEANT, G-E-A-N-T, the R&D organization, and that's for collaborating research across continents. So yes, there is an element of that happening, and I'm sure there's going to be more of that in from initial hesitation to being [00:35:00] able to build- And accept the required controls.

So I think the EU will move in that direction in, in a positive way.

Max Clark: I got on the internet on a 1,200 baud modem. I mean, it's like I'm like a dinosaur, right? And something that I've, I've, I, you know, I, I think about a lot is just, like, how fast this is progressing and how fast e- you know, there were a lot of standards.

There were a lot of different standards. We had ATM, we had all these other things that were all good standards, but then Ethernet just dominated and took over, and everything, you know, all the OC-1, you know, 48 circuits turned into 10 gig n- you know, links, and OC-192 left and, you know. And, and so Ethernet seems to have killed, you know, it is, it is the dominant thing now.

And now we see these, like, really fast iterations in terms of bandwidth, you know. Like 10 gig, 100 gig, you know. I mean, kind of skip 40 gig, but 10 gig, 40 gig, 100 gig, 400 gig, 800 gig. Um, uh, terabit [00:36:00] is, is h- here really soon. You know, like, how, how... What, I mean, like, how... Is it just... I mean, I, what am, what am I looking for here in terms of, like, a phrasing here?

I mean, as a service provider, when you look at this thing, like, it must feel like you're always chasing an upgrade cycle of, like, just never-ending, always planning and deploying faster links and network refreshes and, and, and what's that like, you know, from, you know, like, y- because it's like there's no status or static element, you know, to that.

TS Narayanan: It's true from a demand perspective, and again, uh, um, it's also true from a standards and protocols perspective. Um, but luckily again, again, at the optical layer, today you say, "Hey, you know what? You're upgrading from 100 gig to 400 gig." The technology's also made it easy because [00:37:00] a good example is 100 to 400 or 800 gig is pluggable technology.

Max Clark: Mm-hmm.

TS Narayanan: Right? You have OSPs, right? Small factor pluggables that you plug in to your infrastructure, and then you have your upgrade. Yes, I am simplifying a little bit, but there is an element of certain parts of the network lending themselves to easier upgrades and futuristic technology and, and we're lucky in some sense from an optical perspective.

And then as you grow higher in the layers, there's more and more complexity coming in. Um, but, uh, the 100 to 400 or 800 is a good example.

Max Clark: What always surprises me about, or what surprised me about that is it's not just, like, the cost of the router or the cost of the optic. It's like how much power do these things consume, and what's the footprint of the optic, and how much just power does the optic [00:38:00] consume? And, you know, and, and it's not something you think about until all of a sudden you tr- you take a step back and you're like, "Okay, we need to deploy, deploy 1,000 of these optics across our network.

And how much power do they actually take?" Sure. Um, I, I mean, is the entire tech industry just really a proxy for how efficient power generation and distribution is on the planet at this point?

TS Narayanan: If you look at it-- I mean, there's, again, two lenses to look at it. From what's inside the data center and on the networking side.

Max Clark: Mm-hmm.

TS Narayanan: If you look at what's inside the data center, it's becoming a power game, right? With, hey, how much power per rack in a traditional data center, and what is it going to become in AI? And that's orders of magnitude different. And then you look at the networking side, it's actually becoming better because you look at modulation [00:39:00] technologies.

You look at the capacity, you look at spectrum utilization. They are actually compressing the footprint and the power. So there's one part that is going to guzzle more and more power, which is really your compute and storage, and that's because of the demand. And then the networking or the optical technology is actually improving in terms of the physical footprint, and that makes it better.

So it's a si- it's a, it's a coin with two sides.

Max Clark: Mm-hmm. Modulation is a, and spectrum consumption is, is, you know, it's a good point, right? Because, I mean, how many wavelengths are you running on the average, you know, long-haul route at this point?

TS Narayanan: Yeah. I mean, you, theoretically you could do, um, 80, right? [00:40:00] And, and the technology's constantly increasing, right?

There's a th- theoretical limit, and then you have what's in your network, and then what happens for the future. And that's constantly increasing.

Max Clark: Right.

TS Narayanan: Right? And that's why you say, hey, you, you look at your capacity and say, "Hey, you know, what's your, uh, throughput per wavelength? What's your throughput per fiber pair?"

Which is the number of wavelengths in a fiber pair. And then, "Hey, how many fiber pairs can you pack into a cable?" And then you say, "Hey, I have this big fat pipe."

Max Clark: So that becomes-- I mean, that, so now you get into some really interesting planning and, like, contracts, right? Because if you're putting a cable in, how many strands are in that cable, right?

How many wavelengths do you put per strand? Who owns the strands? How do you cut those strands up? Are you selling wavelengths? Are you selling strands? Are you selling capacity? You know, there's like, that, I don't know, there's probably like a dozen different business models just in that, that, you know, 30-second statement, right?

TS Narayanan: Totally agree. Totally agree. Yes. And it [00:41:00] becomes a combination of a technology and business problem. How much of the capacity do you own? How much of the capacity are you willing to lease? How much of the capacity are you willing to lease on short term? Um, what's your IRU model, right? Mm-hmm. And then, and then how are you going to divvy up the capacity?

How much are you willing to lease out or share? How much are you willing to buy from a partner? All of those come into your strategic planning that we discussed a few minutes ago, right? And it drives a number of things, right? How do you plan? What's your financial planning? What's your technology planning?

Because if you're doing a whole bunch of leases, then, so if you're doing fiber pairs, then you'll, you need to plan equipment somewhere. You're doing a combination, you, you, you plan your network a different way, and that has a commercial model. Yes. All of those are definitely true, and they become a part of the planning process.

Max Clark: Are your clients dominantly looking for [00:42:00] capacity, like bandwidth? Are they looking for wavelengths so they can run their own whatever they want? Are they buying strands? I mean, I'm not talking about like clients like telco to telco, right? 'Cause telcos want control of their own infrastructure, and they want regen, and they want stuff, right?

But like from a, a B2B standpoint, I mean, is the average B2B cl- uh, client moving into wanting to have control of strands and their own wavelength, or are they just saying, "Hey, I need 100 gig from point A to point B"?

TS Narayanan: We're actually seeing a mix of both. There are, because sometimes regardless of the segment of your customers, you will find customers that want, that are quite knowledgeable, that are willing to, that are keen to run their own networks.

They're willing. They need to have the control and the flexibility. And there are c- there are customers that want to, clients that want to have someone else manage it for them. So we're seeing a, a healthy mix of both.

Max Clark: For companies that don't know [00:43:00] that EXA exists, right? Because, you know, it's not something that you guys are gonna see str- you know, on the street installing residentially, right?

And unfortunately, in a lot of cases in business tech, what the residential tech it, you know, influences and comes across. F- from, from that standpoint and that awareness, like oh, what's like your effective, like education or go-to-market style of like, "Hey, we, you know, this is available to you. You can, you...

This product, this service in a different way." 'Cause to us, everything you're saying here is just like, "We're flexible. We support a lot of different ways. There's lots of different options here. People need us differently." And from a customer standpoint, that's a, that's a really good thing to hear, you know, from a potential service provider.

TS Narayanan: Yeah. And, and this, this, this is an interesting marketing perspective, and you look at the B2C industry and the B2B industry. Let's, uh, kind of accept it. It's an industry, particularly [00:44:00] when you look at subsea and you look at terrestrial networks in our space.

Max Clark: Mm-hmm.

TS Narayanan: Everyone knows everyone. Are we looking for a lead that comes out of nowhere from someone that we don't know?

Probably less likelihood. We would love that. But I think it's a reasonable assumption to make that this is an industry where everyone knows everyone. So it's a good thing, and sometimes you end up saying, "Hey, you know what? Are you doing the right amount of marketing?" And then there's another picture that says, "Hey, everyone knows everyone."

And then you, you become known on the basis of your capabilities of what you can deliver, your network footprint, your capacity, and your, uh, focus on your business and the way you deliver your business, is probably my take on this.

Max Clark: Fiber [00:45:00] providers have a very interesting dynamic with data centers at this point, right?

'Cause a data center becomes the on-ramp for, like, real network in my world, right? Mm-hmm. Figure out how to get to your local carrier neutral data center, and then the world is available to you. I would, I would presume that with Exa, you're looking at builds of, like, what data centers do we wanna build to and what, what places can we go meet and interact with the most amount of other networks or customers at, right?

Mm-hmm. But then the data centers are looking at networks and they become your customer. You know, data centers are buying capacity from you to link their data centers together, but then they're buying that capacity to then compete with you because they wanna sell that connection between their own data centers, right?

And it, and it's like this really inter- it's just this, it's this, like, fun kinda like circular relationship of, of what, of what's going on. And like, how... You know, is this, is this like, uh, is this, I mean, you said everybody knows each other, right? Is this just like, you know, just, I wouldn't say like frenemies 'cause that's like really negative, but it's just like, oh, we're all really friendly competitors, but we're selling to them and they're competing with us for our [00:46:00] own stuff that they're buying from us at the same time.

TS Narayanan: Yeah.

Max Clark: Yeah.

TS Narayanan: No, that's, that's indeed true. Yeah. Yeah. That's the nature of the business, right? Um...

Max Clark: How, um, we don't have the same experience in the United States. In Europe you have massive AMS-IX, DE-CIX links, like these really massively dense interconnection locations that are really, it's really impressive how, I mean, just how much traffic flows through these things.

Is there an opportunity for network providers and telcos to move that interconnection out of a data center into something that's more like akin to, I don't wanna say like virtualize it, 'cause it's still very much physical infrastructure, it has to be. But instead of being at the data center, how do we facilitate that interconnection between networks in our own footprint?

TS Narayanan: Yeah. Um, I don't think it's a, it's a simple [00:47:00] problem to solve simply because of what exists today, and you have these big hubs that everyone connects into. Maybe, it would probably take some time and a fair amount of thinking and investment to be able to do that, is, is, is, is, is the way I would think about it.

Max Clark: There's a lot of companies pushing into the space of like network as a service. It's not really defined. I was,

TS Narayanan: I was gonna mention, yeah.

Max Clark: Yeah. It's not really defined what it is. It's one of these like squishy terms that's out there. And, um, I've, I've been trying to figure out and explain or, you know, just like what do, what does network as a service actually really mean for us?

What does it mean for our clients? How do people take advantage of it? You know, a lot of time, you know, in a lot of cases, this idea of NaaS started from like, we need to figure out how to connect to a VPC and a cloud provider easier, right?

TS Narayanan: Yeah, yeah.

Max Clark: Like just that. And like a, a relatively simplistic problem that was really annoying.[00:48:00]

Mm-hmm. But with the proliferation of clouds, cloud regions, you know, where you need to connect, you know, um, uh, going a step, a level up and, and not even talking about connection to a cloud, but connection straight to a SASE platform or, you know, another data vendor or, you know, these, you know, whether it's AI or data lakes or, or you know, um, you know, different processing.

You know, it's, it seems like this is a, um, this is an area that's gonna continue to expand and, and something interesting is gonna happen here. Like, I don't know what it is, it just feels like there's something interesting coming down the pipe that's gonna benefit network builders.

TS Narayanan: I actually would tend to agree with you because the, um, the term has been interpreted in multiple different ways, right?

Saying, "Hey, a NaaS service is," has ranged from reselling to [00:49:00] abstraction to virtualization to whatever you could. So I would think there is evolution in the making, is, is, is my sense.

Max Clark: TS, this has been very fun for me. I have one request, a parting favor to ask, which is next time I'm either on the East Coast or in Europe, I would love to get a tour of a landing station. It's been a while since I've been in one.

TS Narayanan: You're most welcome. And, uh, a good place- An easy place. And this is, this is interesting.

It's an interesting question because when I first joined the subsea industry, I actually found it difficult to go to a landing station because they were all very remote.

Max Clark: Mm-hmm.

TS Narayanan: Right? Dublin is a place where we could take you to a landing station without having to travel to a very remote location.

Max Clark: I went on a tour of [00:50:00] landings on the west coast of California.

Okay. On the west coast of Cal- west coast of the United States in California. And, uh, and San Luis Obispo, you know, there was an AT&T facility, Asia Global Crossing had landing. And when you say, like, everybody knows each other, the engineer that was, that was giving me the tour knew every building, knew every operator, knew every engineer.

It was like they all hung out together and whatever. Yeah. You know, like surfed in the mornings. But what was really funny was the amount of, like, stranded asset. Like, there's a bagel shop, and behind that bagel shop is an ILA for this landing cable that's, like, hopping up the coast, and you're like, "It's just a bagel shop."

And he's like, "No, no." And he drives you, drive around the parking lot and back to the parking lot, and there's generators sitting there . And you're like, "This is very unusual," right? Um.

TS Narayanan: Yeah. Yeah.

Max Clark: The, uh-

TS Narayanan: Yeah, that's, that's true. Yeah. Many of these facilities tend, and rightly should, you know, are, uh, nondescript, if you like.

Max Clark: Yeah. The average person would have no idea what it is or why it's important, which is, which is good.

TS Narayanan: Yes. [00:51:00] Yes.

Max Clark: Unless you just happen to be on, you know, on the coastline when the ship is, has the divers, like, pulling the cable into the manhole, right?

TS Narayanan: Very true. Very true.

Max Clark: TS, um, any, any parting thoughts that we didn't cover that, um, we should, we should talk about briefly?

TS Narayanan: No, I think we've, we've probably covered a, a wide spectrum, right? We've talked about data centers, we've talked about traffic patterns, we've talked about the hype of AI, we've talked about network planning. Lots of things, right? And I think my only parting thoughts would be to reflect on what you said, is to see where, where AI is gonna take us.

Max Clark: If you figure that out, please let me know.

TS Narayanan: And likewise.

Max Clark: Promise. Deal done. Thank you. Thank you very much. It's been a pleasure.

TS Narayanan: Thank you.

Max Clark: That's it for this episode of Signed. If you got something out of this, share it with someone [00:52:00] in your world who's staring down a tech decision. A CIO, a CFO, a founder, procurement lead, whoever.

That's how the show grows. Everything from today lives at itbroker.com/podcast. Show notes, transcript, links to anything we mentioned. If you're in the middle of a real tech decision right now and you want someone in your corner without the vendor bias, that's what we do at itbroker.com. Schedule a call on our website, buy tech without regret.

I'm Max Clark. Thanks for listening. See you on the next one