2026 Infrastructure of the Data Economy

P3 Infrastructure of the Data Economy
Jean-Marie Guéhenno:
Delighted to have Erika Lonergan with us to moderate the panel. She had a logistics issue this morning, so she couldn't get to Columbia, so she is moderating on Zoom, but I will be a kind of co-moderator in the room. We are also at an issue with a couple of participants.

I mean, it seems that technology sometimes creates its own challenges, but we are very privileged to have Dominique D’Hinnin and Andrew Metcalf with us. They both have great involvement in, I would say on the business side of things. And I said at the beginning of the program that one of the aims of the program is really to connect different communities, the community of diplomats, the community of business people, civil society, and what we know of artificial intelligence is indeed the enormous amounts of capital that are involved in the development of the data economy.

This morning, we just had a presentation on what is exactly AI. And so I think with this panel, we can get into all the physical dimension of the data economy. And I'm very grateful to Erica for, who is herself one of our experts at Columbia and at SIPA, who is looking at the issues from a variety of angles to moderate this panel.

So over to you, Erica.

Erica Lonergan
Great, thank you so much. And again, to all of the participants, I apologize for not being there in person. I had some issue with a flat tire that prevented me from getting into New York this morning, but I'm glad that technology enables us to have this conversation virtually.

So let me briefly introduce myself and then I'll ask my colleagues on the panel to maybe briefly introduce themselves as well. And then I will, the plan is I have several questions that I'd like to ask my co-panelists and then really hoping we can open it up to a lively question and answer session on this topic, which is of deep importance to the economy, to national security and so on. So I'm Erica Lonergan.

I'm on the faculty here at SIPA. My research focuses on the sort of international security perspective on emerging technology. So I do work on cyber, AI, outer space and what these things mean for the future of war fighting, deterrence, strategy.

But of course we know that one of the challenges and opportunities of these technologies is that the private sector plays a tremendous role in developing and maintaining, securing these capabilities that have implications not just for war fighting, but for our economy, for commercial uses, for governments and so on. And so this is really, I think, at the heart of some of the governance challenges and strategic implications. So maybe we can dive into that a little bit, but let me turn it over to Dominique, if you'd like to introduce yourself and then Andrew.

Dominique D’Hinnin:
Yes, good morning, everybody. I'm very happy and honored to be with you this morning. I spent 25 years in different positions in a high technology group in France and then I became an independent board member and then chairman of a satellite company.

The name is Eutelsat for eight years. I left the board almost one year ago, so I'm free to speech. I'm free to talk to you.

And I guess I said the most, what is important. Great, Andrew.

Andrew Metcalf:
Hi, thanks. I'm sorry I couldn't be there in person, but my name is Andrew Metcalf and I'm with a company called CPA infrastructure. We develop digital infrastructure in Europe, which is essentially a large energy campuses for AI data centers and other types of compute workload.

And, but my background is been in digital infrastructure for my whole career. So I was at Google for 18 years. I started as a network engineer and then I quickly went into management and planning for different types of asset acquisitions that we needed in order to develop our Google's cloud infrastructure.

So I created a set of planning teams in various parts of the world where we purchased dark fiber, subsea capacity and what they call co-location, which is a type of data center capacity. So I spent quite a lot of time doing that. And in my last five years at Google, I led the subsea, the submarine cable team and put together the plan for the subsea cable fleet that's out there as well.

So I've kind of bounced between different aspects of digital infrastructure, as we call it today. And that's sort of my focus is on the technical and the commercial side of that.

Erica Lonergan
Great, thank you so much. So let me kick it off with some questions and I'll sort of go back and forth. Dominique and Andrew between asking each of you questions, but if either of you wants to sort of chime in to respond to or build on an answer from the other panelists, please just feel free to do so.

I'm hoping to keep this conversational to the extent that I can. So I'd like to start with Dominique and talking a bit about satellites. I was wondering if you could sort of first to help us sort of set the stage for those in the audience who may not be as familiar with outer space and with satellites to kind of help explain what are their applications?

How actually do they connect the world? What are their vulnerabilities? What are the most important things we need to know about satellites?

Dominique D’Hinnin:
Okay, I'll try to be short.

Erica Lonergan
Yeah, in brief.

Dominique D’Hinnin:
In brief. So I will focus on telecommunications mostly. So in fact, satellites are quite an old technology.

The first one were launched in the early 60s and they have been around for 80 years now. I mean, 60 years, sorry. The first category and most important one historically speaking, is what we call geostationary satellites.

We talk here, we talk about satellites which they look still in the sky because in fact, they are very far away from the ground, 36,000 kilometers, but they rotate around us at the same speed as the planet itself. That's why they look not, you get the feeling they don't move. So in fact, they go very fast in the sky.

These geo-satellites are very high in the sky, as I said. They are very expensive items, each of them. And they are mostly used today to broadcast TV, TV signals.

I mean, direct TV, this kind of service and less and less, but still it has work, communications. I don't know if there are people old enough in the room, but let's say 30 years ago, when you were calling Europe from the US, when you were talking to your counterpart, it needed about half a second for the sound to move from one talker to the listener. This is because the communication went through the geo-satellite.

So it goes at the speed of light, but even if it's fast, it requires 36 kilometers on the way in, 36 kilometers on the way back, plus the computing time and so on. So it's half a second. That's why sea, I mean, cable under the oceans replaced most of the communication needs around the globe, because obviously the travel time is much shorter.

The speed is the same, it's the speed of light and geo-satellites increasingly focused on TV broadcasting. Now, TV broadcasting through satellites is declining as well because of cable, because of fiber, because of all the ground-based networks you cannot have everywhere to have a broadband connection. And so it's a declining business.

The other type of satellite, the second one is what we call LEO satellites, Low Earth Orbit Satellites. These satellites are around 300 and 2,000 kilometers high, so much closer to the ground. And because they are much closer, they cannot cover a very large part of the earth.

So you need plenty of them, and that's why it's a constellation. And they are not fixed, unlike geo-satellites. They turn around the planet all the time.

It takes them usually, I think it's 90 minutes to make a full turn. So when you have a connection through a LEO satellite, in fact, your connection will jump from one satellite to another one as they move around. In fact, it's a cellular network in the sky where the towers are in the space instead of being on the ground.

And they move, and you don't need to move. I mean, you can, but you can be in your home. So this business is new.

It was launched less than 10 years ago by some American companies. First, the first one became, in fact, a European. The name is OneWeb, which was bought by Eutelsat.

And the second one, the best known one, is Starlink, launched by Elon Musk. These LEO constellations have obviously a very large competitive advantage compared to geo-satellites, which is the time lag you need to get the communication done. As they are not very far, usually, I mean, Starlink is 550 kilometers high.

It's not, doesn't take more time than on cable. Now, up to a few years ago, technology was there, but it didn't make sense from a business point of view because the cost to launch these satellites were way too high. It's very expensive to launch a rocket.

But thanks to a significant decline in the cost of launching rockets, due to SpaceX mostly, it became affordable to launch thousands of satellites in a LEO, for a LEO constellation, and make a profit out of that. So these are the two main, I mean, types of satellites. The first one is declining.

The second one is increasing very fast. Today, SpaceX, or Starlink, has more than 9,000 satellites in the sky. So it's massive.

It works very well all across the globe, or almost all in every country, not in China, not in Russia, because they could not get the licenses, but it's another story. So I try to be short.

Erica Lonergan
Yes, thank you. Let me just ask one follow-up question on the satellite front, and then, Andrew, I'd like to turn to you to ask you some questions. So I was, I really appreciate the sort of overview of, in brief, of the different types of satellites and how their sort of economics have changed over time.

And you mentioned Starlink, of course, which is, I think, what many of us think of first when we think about satellite constellations, especially in low Earth orbit. I was wondering if you could speak to some of the vulnerabilities of these satellites, especially, maybe not especially Starlink, but Starlink is such a notable example. I know with Russia's invasion of Ukraine in 2022, there have been lots of reports of, for example, cyber-attacks targeting satellite constellations in LEO.

There's been other forms of interference and disruptions of communications through the electromagnetic spectrum. And so I want later to get to a conversation about how we think about sort of governing this infrastructure and the challenges of the role of the private sector in the government. But sort of first to do that, I was hoping maybe you could help us understand the ways in which these space systems that are so important for the commercial economy, for communications, and then also for governments and militaries, how are they vulnerable?

What are the different ways that perhaps malicious or adversarial actors could disrupt these systems?

Dominique D’Hinnin:
Yeah, it's a very good question. Of course, in the current environment, I would say first satellites are vulnerable to nature. Sun erections, meteorites, it can happen every day.

And it does happen sometimes. It's specific to satellites. Fortunately, it's not, usually you can recover, but there was a massive sun eruption in the 19th century.

At that time, there was no satellite. So it had very little impact, obviously. But if we had the same kind of eruption, it would probably blind or even kill most of the existing satellites.

So that can happen. On the other side, armed forces cannot bomb satellites. I mean, towers can be destroyed.

It happens, of course, in Ukraine and so on. And that's why the militaries turn to satellites because it's much more resilient in a period of war. I would say, obviously, you could argue somebody could try to destroy the satellites.

Technically speaking, it's probably feasible. It is feasible, at least the LEO satellites. Nobody even tried to destroy a GEO satellite.

It's quite far away. No missile can go there, 36,000 kilometers. Maybe with a laser or something like that, but they go fast, they are far away.

It's very difficult from a technical standpoint. LEO satellites, they can be destroyed, but keep in mind there are now about 15,000 satellites in the sky. Belonging a significant part to SpaceX, but for some of them to Europe, to the US government, to the Chinese government, to the Russian government.

I mentioned the main holders. So if you destroy one satellite, you create thousands of small pieces, if it explodes, that would fly around at a very high velocity and destroy the other satellites, including yours, maybe. So it's not a very powerful option from a military standpoint, unless you don't own any satellite.

And some people say, because Russia is behind, they have much less satellites than the US or China, they could think about that, but still they rely on satellites for not only communication, but for Earth observation and so on. So would they do that? Not sure, by far.

We know they can do it. The American can do it, the Chinese can do it, but would they do it? Not sure.

On the other side, what you can do is to try to blind the satellite with a laser or with some kind of malicious airwave, but keep in mind, I mean, Starlink is 9,000 satellites, so it's a lot. And even if you incapacitate, you make it impossible for, let's say, 100 satellites to work, there are still plenty of them. Now, you mentioned some cyber attack at the very beginning of the Ukraine war.

In fact, it was done by the Russians on the day one, because the Ukraine army was using a Viasat, the Viasat Geo constellation to have some of their communication for their army. So the Russians, they sent a kind of malicious code through this Viasat satellite to the terminals, and it destroyed the terminals, but not the satellite. That's also the reason why President Biden at that time reacted quite harshly, because they destroyed, I mean, they made some harm to an American asset.

That was quite offensive, I would say. But at the end of the day, they didn't destroy the satellite. The satellite had nothing, it worked, and it's still working.

The terminals were destroyed, in fact, I mean, from a software point of view, they had to be replaced. And in fact, what happened is, they replaced them with Starlink. And it raises another issue, which is, I'm Ukrainian, let's assume I'm Ukrainian, and Mr. Musk decide to stop Starlink over Ukraine. So I depend on a U.S. constellation. So Mr. Musk could follow an order from the U.S. administration, or Mr. Musk, being Mr. Musk, he could decide on his own to stop the service, which is a double risk, if you want, if you don't own a constellation. And that's why the bigger countries try to have their own constellation, not to be exposed to such a risk, or double risk, when you talk about Starlink.

It's the case with the Chinese government, I mean, they have constellation. The Russian government has a constellation. Europe has one today, with OneWeb.

And, but other countries don't, and they depend on the willingness of some other countries to keep their communications. Nevertheless, it's a major move from a government to stop the communication tool. So I would say, when, it's still much more resilient in a period of war than a ground-based system, that's for sure.

Or even now, than a cable laid in the, at the bottom of the ocean. We all know the Russians made some sabotage to sea cable in the Baltic Sea. Iran is threatening to do the same with the cables in the Strait of Hormuz.

So they are vulnerable, much more than satellites.

Erica Lonergan
Great, thank you so much. That was so illuminating. I appreciate that.

Andrew, let me turn to you, to another piece of the sort of broader infrastructure of the data economy. We've talked about satellites, briefly talked about undersea cables, I want to pull on that thread a little bit more as well with you, given your experience there. And then of course, there's the cloud and data centers.

And I think we tend to think about these digital technologies from the sort of abstract perspective, but as we're seeing from this discussion, they are grounded, either literally grounded or in outer space comprised, or under the sea comprised of physical, actual physical infrastructure, right? And so, Andrew, I was wondering, again, just to start us off, if you could help our audience understand the role of cloud and these other sort of infrastructural components of AI beyond the sort of the algorithms and the LLMs themselves, what are the infrastructure inputs that enable, that power and enable the compute that makes these models so powerful?

Andrew Metcalf:
Yes, so the cloud is essentially just renting somebody else's computer, that's the idea. So the people who are customers of cloud customers, they're renting a computer that's being housed inside of a data center that's then connected to the internet. That's the simplest kind of version of what the cloud is.

And the reality is that it's a fairly mature industry. AWS started cloud computing in 2006, and since then it's been used for, essentially the two problems it solves are availability of applications. So you have a web-based application or an enterprise application within a company.

It provides high availability because you can host that application across many different computers across different data centers at the same time. And then it's also, it's less expensive because now I don't have to own the data centers that are on in my corporate data center or whatever. So those are the two kind of driving factors of why cloud became wildly successful.

And so from a physical perspective, though, the data centers that these servers are in, they're fairly industrial. There's quite a large ecosystem of them. They predate AI by quite a long time, even when I was deploying these things.

And 2006, 2007, most major U.S. metros had a well-developed data center ecosystem. I think the most well-known is Northern Virginia in the U.S., which certainly has been a concern to a lot of people because there's a single grid, but they have multiple sub-grids for that area. But that ecosystem has been present in the U.S. for quite a while. And the reason I mentioned that is because, like I mentioned, part of the point is the availability is one of the key components of cloud. So the question about whether there are choke points or vulnerabilities around cloud, the first principle of it is distributed and therefore there's less vulnerabilities. Although I think we've all seen issues like the CloudFlare DNS issues that have cropped up every once in a while and the grid issues in Northern Virginia where you had cascading failures that caused massive outages.

I think with the way that cloud is designed and the way that customers deploy their applications, they deploy them redundantly. So they'll put within one metro, because usually the way that a cloud works in a metro is there'll be multiple data centers in a metro connected together. And that's called an availability zone.

Essentially, you can host your application across multiple data centers within that metro, like say New York or Geneva or whatever. And if the application goes down to one data center, it gets taken over by another one. But then even beyond that, you can have your application or your enterprise application or your website or whatever it is distributed globally as well as many people do.

And that's what CloudFlare kind of specializes in. So from a choke point perspective, we do see failures occasionally that are kind of fairly disruptive like DNS failures, but generally cloud has kind of increased the availability of these applications over time just because that's exactly what it was put into service for. And now that we have AI, and I'm gonna use the word workloads, I'm not sure if people are familiar with this, but essentially whenever you put something in cloud, it's sort of in a virtual container, let's say your application.

And you'd call that a workload because it's using compute, it's using storage, it's using compute, it's using different resources. So when I say workload, that's what I mean. So now that we have AI workloads, which are very different.

So the previous workloads were, I say CPU bound, which meant that they use the normal CPU processor that you'd have in a computer. And those enterprise applications ran on those. Now with the advent of the GPU, which was NVIDIA's main product, that's a processor that allows lots of really cheap mathematical calculations very quickly.

And it was meant for graphics, but it also turns out for AI workloads, it's very well suited for it. So fundamentally from a physical perspective, cloud has started to shift from traditional CPUs for enterprise type applications to also adding lots of GPU access for, or GPU resources for AI applications. And that has caused sort of, I wouldn't say a division, but a morphing of the market where now you have what are called Neo clouds, not to just add more buzzwords, but a Neo cloud is a cloud provider that's specific to only doing AI workloads.

So they not only have the GPUs there, but they provide what's called an orchestration layer, which makes sure that your workload doesn't go, it's distributed across the GPUs properly. And also there's a failover. So again, that sort of concept of redundancy comes into play.

And so that's where we are now with AI. So cloud is a hundred percent critical to AI. When you look at the frontier models, like OpenAI and Gemini, those are all on cloud services, right?

So Google has their own cloud services and they host Gemini on it. Anthropic, it hosts their services on AWS and on Google as well. So essentially you have quite a lot of redundancy.

However, I think the thing that we're all worried about and that we know is that the ability to actually be able to host more AI workloads is really kind of the main constraint everybody's thinking about these days, right? Over time, I guess the question is, when, first of all, from a competitive perspective, you have multiple players in the world that say, like China's outpacing the US as far as energy production. Energy production is what allows you to have more data centers, right?

So I think those are kind of the concerns that are going on now and whether or not you're gonna have constraint of cloud because you literally don't have the supply chain for it. You don't have the grid capacity for it and things like that. So I think, so I guess it's going away from the concept of a traditional typical choke point because I feel like a lot of that has been solved within the cloud ecosystem.

There's still issues every once in a while. I wouldn't deny that. But at the same time, we see today, it's very hard to get cloud capacity or GPU capacity within a good timeframe right now, right?

So it's a very spicy market, I would say, for those sorts of things. I can speak to the network side of it if you want as far as choke points, but that's sort of my general view. And I'd say US dominates the cloud market.

I think two thirds of all cloud is done by US companies that may shift as resources become constrained. So that's something I think that's a concern for everybody.

Erica Lonergan
Yeah, no, that's great. I was gonna ask you a follow-up about the economics and the relative market share of US firms versus others around the world. But I was also wondering if you could, especially because some in our audience may not be quite as familiar with all of these issues.

You mentioned the potential challenge of as there's increasing demand for AI workloads, AI-related workloads, there are energy constraints, there's GPU constraint. What are the various potential constraints that might make it difficult to scale and keep pace with increasing AI-related demand?

Andrew Metcalf:
Yeah, grid capacity is number one. You see several different types of solutions out there. I wanna say grid capacity, I mean, just to focus on the US, the national US grid, right?

That's broken up into many regions, but the name of the game for providing electricity to your data center is first of all, you need to be able to connect to the grid. And in order to do that, the grid operator has to do a study and make sure that the capacity is physically able to transmit over the grid. It's a transmission system.

And then the second is to actually have a producer of the electricity, the electrons essentially that will come over that grid. So both of those elements have to be lined up and both of them are quite challenging, right? So from a grid perspective, it's more about building substations, building wire, upgrading that transmission capability by itself.

And then on the production side in the US, natural gas has become a much bigger player. More and more, it's very critical to the way that actually that energy is produced. So I think it's helpful to kind of, when people talk about power issues, it's helpful to sort of separate out what the grid and what we call production, which is the generation of electricity.

And I think those are more so issues than supply chain in my view. The thing with grid is, is you're talking about organizations that are not particularly suited towards quick response to, they keep the grid up. They're very good at that.

But as far as expanding the grid, typically if a car manufacturer moved into a state and said, we're gonna need so many megawatts in 10 years, right? There was time to build the grid. There was time to get those things going.

So the cycle time for traditional industry was fairly long for the grids. And now with the data center, you're talking about wanting something within 18 months, right? Which is sort of warp speed for those industries.

So I think there's some structural issues with making sure that the regulatory side of things within that regulate the utilities and utilities themselves have to kind of recalibrate to this new reality, but that requires money and it requires a lot of support in order for that to happen.

Erica Lonergan
No, great. So you mentioned sort of the regulatory environment. I wanna sort of pivot from that and talk about a related issue, which is this question of governance of these infrastructures and these technologies.

And so I guess this is a question for both, Dominique for both you and Andrew for you. Based on sort of both of your remarks, especially thinking about, again, the role of the relationship between public and private actors, challenges of regulation, both at the national level, at the regional level and the EU at the global level, and then the layering of sort of this geopolitical dimension and geopolitical rivalry between the US and China and Russia and Europe, right? So how, for this audience in particular, thinking about sort of a diplomatic and a global governance perspective, I'm wondering what each of you think about what are the most significant sort of governance challenges around these infrastructures and these technologies and what is the state of governance and perhaps what are the most realistic approaches to thinking about governance in these areas?

And I know that's a lot, so feel free to kind of take the question as can you solve all the problems basically, but I guess it's interesting because outer space is a commons, right? Dominique, you talked about the kind of the challenges of sort of kinetic attacks against satellites not being in anyone's interest because it's self-defeating, right? You create debris that risks not only an adversary satellites, but everyone else's and maybe also your own.

So there's this commons challenge in space. We don't have a lot of treaty law governing outer space except the Outer Space Treaty, which I think was in 1967, if that's correct, right? So who governs space?

How should we think about governance? Who are the stakeholders that need to be involved? And then the challenge is slightly different when it comes to data centers, obviously, but they're competing regulatory approaches to these different AI inputs.

And then when it comes to undersea cables, those are owned and operated by private actors and there are not sort of robust governance regimes around a good portion of those cables, right? So there are lots of governance challenges. So I just thought I'd throw it out to both of you, kind of what are your thoughts on this?

What are the biggest issues? How should kind of from a diplomatic and governance perspective, how should we be thinking about these? So maybe Dominique, I'll start off with you.

Dominique D’Hinnin:
Yeah, it's a big question, of course. You mentioned the main text, which is the Outer Space Treaty signed in the 60s by USSR, the USA, China, Europe. So everybody, I mean, a lot of countries and the most important countries is the ones with access to space signed it.

This treaty says there is a free access to space. Everybody can go there. And the space cannot belong to anybody.

So you cannot appropriate anything in the space. That's the principle. From a military standpoint, what the treaty says is you cannot put mass destruction weapons in the space.

It doesn't say you cannot put weapons, only mass destruction weapons. At that time, everybody was thinking about nuclear bombs, of course. So, but you can make nuclear missile travel through space.

You cannot just put them there. That's the only limitation. Up to now, it was all the states complied with that, as far as we know.

Even if some American people from the administration said they believed Russia was either about to or already did launch a nuclear bomb located in space. But nobody knows whether it's true. And maybe they are just saying that to get a bigger budget.

Happens. Now, the space business is still heavily regulated. Number one, access to space.

You need an authorization to launch a rocket. You need another authorization to re-entry the rocket if it's a reusable one from the country you belong to. You need licenses to have part, I mean, the spectrum.

So, and it's allocated by the states if you talk about LEO constellations, but by the ITU, the International Telecommunication Union, which is part of the United Nations system, I would say, if you talk about geo-satellites. And even for LEO constellation, you have to comply with the fact a new satellite, whether it's NEO or GEO, cannot have a negative impact on the existing ones from a spectrum point of view. So in fact, there is a lot of coordination at the ITU level, both on the LEO side and the GEO side.

And anyway, for LEO, first, for example, SpaceX has to get some bandwidth licenses in the U.S. Then the U.S. government has to make sure with ITU it's okay. But in order to sell the constellation in other countries, SpaceX has to get licenses in each of these countries and each time it's two licenses on the way up to the satellite and on the way down from the satellite. So in fact, satellites are heavily dependent on regulation.

Nevertheless, the space treaty, as I summarized it, is being challenged. It was challenged in 1969 when the Americans put an American flag on the moon. I mean, in the 16th century, it meant it belongs to me.

Because of the treaty, it cannot mean that. But, you know, there were private companies in the U.S. selling some ground on the moon. It's bullshit, but still, it shows there was a beginning of something.

Now there is a race to the moon, mostly between the U.S. and China, and the U.S. are quite vocal right now to say they want not to be the owner of the best part of the moon, but to control it. They cannot be owners because of the treaty, but the lawyers are very smart people, and they explain the treaty doesn't prevent you from protecting the assets you put on the moon. So you can put barriers, you can put even soldiers to protect your assets, and the difference with full ownership is quite thin.

So the treaty is being challenged about the moon today. It really is. Mars, planet, I mean, obviously it's the same.

There is a very interesting point about SpaceX. When you want to subscribe to the Starlink Constellation service, you have to sign off with all the legal texts associated with the agreement. Nobody reads that, but I did.

And I would advise everybody to read, it's on the website, the Starlink website. It's Article 11. It says, if there is a disagreement between the customer and Starlink about providing broadband service on the Earth, the law, the applicable law will be the Texas law.

That's okay, Starlink is in Texas. If there is a disagreement about providing the service on route two or on the moon, it's still the Texas law, which could be debatable from an international point of view. And if there is a disagreement between the customer and Starlink about providing the service on Mars, the law, the applicable law, will be defined in good faith, in fact, by Elon Musk.

It's a very, very surprising provision. It's against everything you can think about in the space treaty. In fact, it means Elon Musk wants to build his own empire, not an American one, on Mars, and he will be the ruler.

So it's not for today, not for tomorrow, but we have to be aware of that. And it's consistent with, he issued his prospectus about the IPO, and he says he wants to build an interplanetary company that goes beyond traditional borders and so on. It's consistent with that.

It's very interesting.

Erica Lonergan
That's very interesting. I made a note to myself. I will read that article 11 after this panel.

I hadn't heard of that before. Very interesting. Thank you.

Andrew, let me turn it to you to offer your thoughts on these challenges of governance, understanding that they're different when it comes to these issues you've been speaking to.

Andrew Metcalf:
I'll touch on subsea cables first, because I think it dovetails in with Dominique's things. So subsea cables, they've been around since the early 1800s, so the first transatlantic cable was in the 1850s. And so there's two components really to governance.

I mean, governance means governance, but there's also sort of a self-governance within the industry. So I'll just touch on those two, because I think they're very interesting and critical to understanding the subsea cable world. And I'm speaking from my perspective as far as building them and then putting them in an operational state within various countries.

So from a governance perspective, every cable has at least two sets of license regimes you have to follow, unless it's a cable that stays within a particular country. So typically when you land a cable, besides your sort of physical things like permits, environmental clearances, and all these other sorts of things, you usually have to go through some sort of, at least process to understand where it stands from a telecommunications perspective, because it's a telecommunications service. Sometimes if it's a private service, it can, there can be sort of a waiver or you don't have to be specifically regulated as a telecom provider.

However, you do, you always have to at least go through that sort of checkpoint. And in the U.S. for instance, you do have to have a license in order to land. The FCC has to grant you a submarine cable license.

Not all countries require a submarine cable license, but they at least require a review to understand what is the regulatory position of this particular cable given its owners and its intentions. So that's, and sometimes that's actually tied into the other physical elements of it, like the, like for instance, if you, when you lay a cable and you have a vessel coming into the territorial seas for that particular country, the purpose, a lot of times the way they'll understand the purpose of that vessel is whether or not there's a telecommunications license associated with that cable, or at least some sort of paperwork from the local telecom provider. Otherwise the local maritime authority sometimes will not allow that vessel to proceed. So even though it's not the most consistently regulated element of these sort of digital assets, because it really depends on the countries that you're going into, there are lots of, there's lots of governance around getting these things installed and then operating them.

And obviously you have to adhere to all the telecommunications rules of the countries that you're operating in. But I think what's really interesting is that you have to remember that a lot of these cables, if you look at a map of submarine cables around the world and you see how they're laid, you'll notice that it is not spaghetti, that they actually have sort of a pattern to them. They're separated a certain amount, and there seems to be sort of a courtesy between the different cable providers, or owners, as they've laid the cables.

And that's because there is a long history of cooperation within the industry in order to make sure that cables are operated in a safe way and that they're operated in an economical and reliable way. So a lot of that is done for the telecom submarine cable world at the ICPC, which is the International Cable Protection Committee, which is a completely non-governmental organization that consists of stakeholders within the industry and within the world of subsea communications, including ship owners, cable owners. And there's a set of guidelines that they publish that are created by work groups that govern how, or not govern, that's not the right word, but they recommend how owners should behave on the seafloor, to use the seafloor in a way that's beneficial to everybody, but there's not as much conflict.

So they do things like keep three water depths away. So if you're in very deep water, you're in 3,000 meters of water, you're supposed to keep 9,000 meters away from another cable. And that's because if you need to recover that cable for a repair, it minimizes the chance that when you put a grapnel down to pull up that cable, that you'll then hit another cable instead.

So the industry has kind of pushed itself to do these things because of hard-won lessons, essentially, right? And another thing I think that people think about is the repair as well. And I don't really, I can't really touch, I think it's too much of a big topic to talk about whether or not like repair efforts to create more resilient networks need to be something that's done by any particular government.

But again, from an industry perspective, the way that repairs are done for submarine cables is through what are called zone systems. So essentially, the industry has gotten together and it's either created clubs or there's private suppliers that provide standby vessels that then do repairs within those particular zones. So like in the Atlantic, you have what's called ACMA, which is the Atlantic Cable Repair Agreement.

And then you have the APMA, which is a private version. And so these are, again, cooperative ways that the industry came together to make sure that they could keep these cables up because there's no insurance for subsea cables. It's essentially a self-insured type system.

And the way that they solve that is by creating these repair clubs and these repair organizations to repair the cables. So I think private industry does a very good, I mean, maybe I'm biased, but private industry does a very good job of repairing those. But I have to caveat that that's tuned towards the typical types of faults that you have on cable systems, which are gonna be a fissure or something to do with like a private vessel, and anchoring for legitimate purposes.

And then they're in a storm and the anchor drags or something like that. So it's not really tuned towards massive failures. However, you do see certain events like the mudslides off of the Congo Basin and the earthquake in Japan, where the industry did get together and essentially put together a bunch of resources to do a massive repair operation in order to overcome those challenges.

Erica Lonergan
Great, thank you. This has been tremendously informative and I have other questions, but I'd like to give our participants in the audience a chance to ask questions. So I don't think I can see hands that are raised, so hopefully we can coordinate that, but yeah.

Jean-Marie Guéhenno:
Yes, I can help on that. So introduce yourself.

[Speaker 6]
No, thank you very much. This was an interesting session. I have two questions, one for Dominique.

When you mentioned Starlink and the outer space laws, you said that there are some countries that does not allow Starlink to operate in their country. Given it is remote, what measures can the states take, for instance, to prevent that? Or like technically, is that possible to beam low orbit satellites without the permission of a sovereign state?

Because it goes into the issue of sovereignty. For Andrew, in the cable business, sea cable, do we see like fragmentation among countries and countries using different cables and over the past years, like what is the trend, especially in the context of China, US contestation? Do we see more and more countries relying on their own cables or we still rely on similar cables by the major companies?

Jean-Marie Guéhenno:
Since the question was addressed to Dominique on Starlink, may I complement that question with an additional question on the comparative advantage for launchers? Because obviously if Elon Musk has been able to launch so many satellites, as Dominique said, it's because it's cheaper. And so how is the international scene, competition scene is on the launchers market?

Dominique D’Hinnin:
Okay, so I'm going to answer first the question about local government, let's say. Obviously, LEO Constellation is a global business and satellites are around the whole Earth, the whole planet. And in fact, I mean, you, let's say in Iran these days, I read in articles, the American administration sent some Starlink terminals to some people in Iran so they can have an access, an internet access.

So it means it works even if you don't get the license, the only thing is it's unlawful. So if these people get caught, they obviously will have some trouble, but it works in Iran. And it works in Russia, it even works in China, unless the government decide to scramble the system, but it's very expensive, complicated, and Starlink has a very versatile way of emitting waves.

So it's difficult to scramble. So yes, it can work even if you, it doesn't, there is no license, then it's about, you know, sanctions against people who are not complying with the local regulation. But Starlink cannot be sanctioned in this respect.

And on the other side, Starlink can decide to stop the service, even if it's fully lawful. So it's a very global business. It goes beyond borders.

And it's an interesting point in this respect because it's probably the first time in human history where something can happen even, I mean, above and beyond the willingness of the sovereign states. Now, to answer Jean-Marie's question about the launchers, I will give you a couple of easy figures. When you want to compare price about launching satellites in the space, you take a price in dollars per kilo.

It's very simple. With the American shuttle 25 years ago, the cost was about above $40,000 per kilo for all. Ion 5 and more traditional launchers were about 18, $20,000 a kilo.

So half the price. And by the way, that's the reason why the NASA decided to ground the shuttles. It was way too expensive.

Falcon 9, I mean, Ariane 6 is 12,000. So it's much cheaper than Ariane 5. Falcon 9 today is 3,000 because they have critical mass.

They have hundreds of launchers and it's a reusable rocket, which means it's less expensive to operate. And Elon Musk very often says, well, the reason why I had this idea is because I looked at the airline industry and I thought, well, you could never develop the airline industry if you had to scrap the airplane every time you use it. Now, Falcon 9 is around 3,000.

Falcon Heavy, which is the bigger version of Falcon 9, it's 1.5 thousand dollars, sorry. So $1,500 and the most competitive launchers beside the US is in China. They are not there yet, but a lot of private companies are rushing to the market and they announced prices which are close to Falcon 9 or even Falcon Heavy.

But Starship, the very large rocket, which is not fully operational yet, Elon Musk talk about a $200 per kilo price. So it's a major breakthrough. And here, nobody today is able to compete with that, but it doesn't work yet.

But at that price, obviously you can start launching activities in the space which were not competitive in the past, including data centers.

Erica Lonergan
I think there was, Andrew, did you want to?

Jean-Marie Guéhenno:
Andrew had a question.

Andrew Metcalf:
Right, I think the question was about, do we see sort of defragment or like a fragmentation of cables to various countries? And I guess my thought is that you don't really see many sort of national cables. I mean, in the old days, cables were essentially created through, they were constructed from what they call PTTs, post-telegraph and telegram companies, which are essentially proxies for the national government.

So all cables used to be essentially like bilateral projects between different countries. These days, the hyperscalers are certainly the most active in constructing new subsea cable systems. So that would mainly be meta in Google, but others are doing it as well.

And then you still have carriers, especially in like Southeast Asia, telecom carriers that build cables. And I would say, fragmented, probably not so much. I do think, my worry is always making, is that, so cables are expensive, right?

And so in order for a cable to come into a country and provide telecom services and internet access and those sorts of things, there has to be somebody there that's willing to put up the money in order to extend the cable to that country. So we call that a branch, because you can have a cable that goes past a country and you can branch it out into a smaller country. So you have traditional cable systems.

It's very interesting to look at West Africa, because over time you had like the SAT-1 and SAT-2 cable systems, which were telecom South African cables, and they skipped most of Africa, right? And then you had SAT-3, which was a consortium cable, meaning that it was a cable of multiple investors and they invested in branches all along the West Coast of Africa, right? So suddenly, Nigeria got their first cable, a modern cable, because a lot of these places had telegraphs, and a lot of other countries down the West Coast of Africa.

So over time, as there's been more cooperation, again, at the industry level, there's been more sort of creativity that's allowed this kind of consortium and kind of branch type configurations to be done from an investment perspective. However, if like a hyperscaler, somebody is putting together a cable project and they've got to get a slot to the factory, they're not gonna spend two years talking to every single country to make sure that they can contribute to their own national interest as far as getting capacity into that cable. So I do think about that as far as something that's a concern.

I think so far, you do see very good, you do see quite a lot of cable countries that have not been connected or connected in the last 10 years. Plus you have Starlink, which helps, although Starlink, they do have earth stations that need to have some sort of connectivity besides satellite in order to get back to the rest of the global backbone. But I guess that's a long way of saying, no, I don't really see much fragmentation.

I do think that like regional government and regional organizations should kind of pay attention and make sure that there's a fair connectivity to the countries that are in there because they do have to sort of compete with the structure of how these cable projects are put together from a commercial perspective.

Costantino Tarducci:
Thank you. I'm Costantino Tarducci from Italy, I'm a diplomat. And I would like to thank both speakers for your excellent presentations.

I have a couple of sets of questions. One is on the satellites that have been mentioned and expanded. I wanted to ask you if there is a technical limit, a technological limitation in the number of satellites, both as far as the geostationary orbit, you recall, which is run by the ITU.

Have we reached that limit? I understand it's declining, but I would be interested to know where we stand as far as limitations there. And as far as the low Earth orbit, the LEO constellations you mentioned, if there are also technically some limitations or not, because I'm assuming that from a technological side, but I'm not an expert, of course, not all orbits, although closer to Earth can be used.

And then I wanted to touch bases also on the vulnerabilities you mentioned. I understand that there are some risks, like in any activity, both in satellites and in sea cables, and that there is a big effort by the private sector, by industries in case of damages. But we also mentioned before the sabotage risk.

What is the awareness on this? Is there a role of public and private to prevent it? And on the technological side, can satellites help preventing sabotage to the cable?

For example, the subsea cables, is there a role they could play in that things?

Jean-Marie Guéhenno:
We aggregate a few questions maybe, because otherwise we're gonna run out of time. Yeah.

Isabelle Boutron:
Hello, Isabelle Boutron, UN. Thank you very much for the presentation. Just a question on, especially for Dominique, on the use of satellites by the bad guys.

How to prevent that? I'm thinking about the example of terrorist armed groups in Mali who use Starlink to launch attacks. And you mentioned sanctions before, so I would be interested in knowing more about how to address these kind of issues.

Thank you.

Surprise Malehase:
No, thanks very much, Prof. My question also is directed to Dominique. My name is Surprise Malehase from South Africa. Would you consider the underdeveloped countries would want to embrace this sophisticated technology for their own developmental agenda?

However, with a fear of their national security, that perhaps the private sector may as well use this sophisticated technology to influence sometimes electoral outcome and regime change. But they do have in mind that technology has to be embraced, but with that fear, would you consider that fear legitimate or something that needs to be overlooked? Thank you very much.

Jean-Marie Guéhenno:
I would add one last question, if I may, to Andrew. I mean, Andrew, you mentioned the importance of the grid, the electric grid for data centers. Do you see that as a key feature in the deployment of data centers in the future?

I just saw that SoftBank decided to invest 75 billion euros in France for a data center. We have seen the project for data centers in the Gulf with cheap energy. Is it going to be a key factor in shaping the distribution of data centers around the world?

Andrew Metcalf:
That's a good question. I mean, it remains to be seen, but I believe it is. I think France, when you look at the fundamentals of France with the high, large amount of low carbon energy through the nuclear power, but also the RTE is a very well-regarded grid with very sort of well-understood mechanisms for doing upgrades.

And they have an aggressive investment plan for doing upgrades. So it attracts more data centers. I'd say the same would be true of other places like Finland has excellent grid capacity and excellent grid company.

So yes, I'll just keep it short. But yeah, I think that the ability to, and I think one thing that people are talking about these days is doing generation on the data center sites. I think that that's something to be avoided.

There's lots of inefficiencies with that. It's not a very good solution. It doesn't solve a lot of the problems that people think it will solve as far as keeping utility bills low and things like that.

So for me, I want to see more growth in the grid and I want to see a more flexible grid and I want to see data centers sort of play a role as a participant in the grid and not just a consumer.

[Speaker 10]
Thank you.

Dominique D’Hinnin:
Okay. So I will try to answer the questions first on the number of, is there a limit in the number of satellites? On the geo side, the answer is yes because all these satellites have to be above the equator line around the planet and there is a limit, physical limit to the number of satellites you can put especially because the ITU doesn't want any satellite to interfere with another one.

So there has to be some distance between satellites. It's being managed by the ITU. In fact, it's not full, but not because there is no demand but because there are significant part of the equatorial line where there is nobody, I mean, over the oceans.

So it's quite full over Europe, over the US, over Asia. It's not empty, but not very full over the Pacific Ocean because the market there for TV broadcasting or telecommunication is not very large. So there is still a little bit of room but it's not very attractive from the business point of view.

On the LEO side, in theory, there is no limit because you can put as many satellites as you want. And keep in mind, you are in three dimensions so you can have several layers. And in fact, it's what's happening.

Even Starlink has, in fact, two constellations. One at 350 kilometers high, another one at 550 kilometers high. One web is at 1,200 kilometers high.

So you have several layers where you can put the satellites so they don't collide between them. Nevertheless, the more satellites you have, the more risky it gets to have a collision. It's already a bit difficult sometimes.

Satellite constellations are using AI in order to drive the satellites and avoid collisions because one single collision would be, for the reason I explained earlier, would be catastrophic because of the debris. But there is another limit, which is the bandwidth. Today, the constellations are mostly using two bandwidths.

I'm talking about communication constellations such as Starlink, OneWeb, LEO, the name of the Amazon one. One bandwidth is the KU bandwidth and another one is a KA bandwidth. It's not the same spectrum if you want.

But there is no other bandwidth available if you want. So if you want to avoid interference and the fact satellites scramble each other, there is a limitation to the number of constellations you can launch. Our experts talk about maybe three constellations per bandwidth.

It means total, at max, six constellations because there are two bandwidths. So it's not a lot. And that's also why the existing players, whether they're American or Chinese and so on, are rushing to develop this constellation because there is a first mover advantage.

Once it's crowded, there is no more room. So this is on that. Now, can satellite help prevent sabotage in, well, you have observation satellites you have hundreds of them belonging to the American administration, to the Chinese administration, to the Russian administration, a little bit less in Europe, but still, and a lot of private companies.

So they can monitor ships, whatever, but there is a limit to what you can do. And we talk about most of this constellation in where private companies are not designed to perform defense operation, even if, I mean, the defense department in the U.S. or in China are very well equipped. Now, can we prevent the bad guys from using Starlink?

The answer is, yes, you can, if Mr. Musk is okay with that. He can turn off the satellites when they go over one single country. It's not exactly the border, but it's quite precise.

So he can decide not to provide the service to one specific area. And that's what he threatened to do to Ukraine, in fact. He didn't do it, but he threatened to do it.

And there are pros and cons about that, because it prevents the bad guy, but it could also prevent good guys from using it. Anyway, so yes, he can, but he's the only one able to do that. Obviously, if the U.S. administration ask him to do something, he will likely comply. But, you know, it's interesting because Starlink is, I mean, SpaceX is the most incredible company in the space industry ever. So they depend on the U.S. administration on many things, including money. But the U.S. administration depends on SpaceX for many things as well. So in fact, that was very interesting to see the way it happened when Mr. Musk had this nasty disagreement with President Trump. Nothing happened. And now they are getting along together, more or less, because they are both dependent on each other.

Why the U.S. administration is dependent on SpaceX? Because of launchers, most of that. Also because part of the Starlink constellation is for military purposes, so they use it.

There is a dedicated sub-constellation for the Defense Department, and so on. So, and SpaceX is the only company in the world able to send payloads and astronauts to the International Space Station, along with the Russian. But the Russian have very old technology and nobody knows exactly whether it will last forever.

So the Chinese have the expertise, but they have their own space station, and they are not involved in the International Space Station. So in fact, the dependence is on both sides, which is very unusual and very interesting. So the answer is, yes, they can prevent bad guys, but only if Mr. Musk is okay with that, or decides to. Now, the last question I understand is whether some less technically advanced countries could be reluctant about constellations. I would say, obviously, it's a sovereign driven business. A U.S. constellation, including Starlink, can listen and see everything that goes through the satellites. And the connection with the U.S. intelligence community is obvious. It's the same in China, obviously. So if you don't want your communication to be listened to by the U.S. administration or the Chinese administration, you don't want to go through these constellations. Nevertheless, they bring a service that doesn't exist elsewhere. Connecting people without connection, without broad ground-based connection. There are a lot of areas on Earth that have very poor connections.

And with Starlink and tomorrow the Chinese constellation, be sure you can have this broadband connection, which is now perceived by the population as a basic need. So if you are not able to provide that to your population, you don't have any choice but to go for this constellation. It's a major technical progress.

And it's very difficult not to accept progress. But I would remind you, there is a European constellation as well. The name is OneWeb.

It belongs to Utah SAT. And it's not American nor Chinese. So you have a third way.

But I'm not part of Utah SAT anymore. But it was part of the thinking when we bought OneWeb.

Erica Lonergan
Well, I thank you, Dominique and Andrew for sharing your insightful comments on this very important topic. I think we are over time, which is just an indicator of how much interest there was in this conversation. And I'm sure there are many more questions and topics that will hopefully be covered the rest of the day today.

Thank you again, Andrew and Dominique. And I hope the crowd in the room enjoyed and I apologize again for not being there in person. But it was great to hear all of your wonderful questions.

Jean-Marie Guéhenno:
Well, thank you, Erica. Thank you, Andrew. Thank you, Dominique.

That was a very informative panel. Thank you.

[Speaker 10]
Thank you. Thank you very much. Thank you, guys.

Erica Lonergan
Thanks, everyone. Take care.