Welcome to Catch Me If You Can. I'm your host, Matt Pearl, Director of the Strategic Technologies Program at CSIS. In this podcast, we take a closer look at the technologies and policies driving tomorrow and how the United States can stay ahead in the global innovation race. Today, we're bringing you a special episode from our studio. Hello, I'm Matt Pearl, Director of the Strategic Technologies Program here at CSIS. Today, we're diving into cloud computing, efforts by nations to assert sovereignty over the cloud and ultimately sovereignty over AI, and why it matters in the global race to AI, both for the U.S. as well as allies and partners. I'm joined by Bill Weinman, who's a Senior Advisor here at CSIS. And Bill just published an excellent paper titled The Sovereign Cloud, Sovereign AI Conundrum. We're going to unpack some of the tricky trade-offs that countries face when they try to build their own cloud and AI systems. So Bill, let's start by offering a quick refresher. What is cloud computing? Thank you, Matt. So cloud is a different technical, economic, and operating model for information technology. It has a lot of benefits. It's cheaper, but one of the most important is that it promotes innovation. You can start up with a small cluster, spin it up, shut it down. You pay only for what you use. It promotes the messy trial and error of innovation. It's becoming core to a country's economic competitiveness, and it's the digital backbone for a modern economy. Specific to AI, you're hearing about all these AI chips. Those AI chips and clusters live in the cloud. So AI infrastructure is effectively the cloud. Yeah, and just a footstomp on that a bit, Bill. I mean, if we look at the U.S. right now, virtually all of our economic growth is really driven by the build-out of cloud and data centers. Isn't that right? A big chunk of U.S. private investment is cloud. In fact, over the last three years, U.S. cloud providers will spend well over a trillion dollars on cloud infrastructure. So Bill that does a really good job of covering the technology at a high level But why are governments pursuing a sovereign cloud and ultimately sovereign AI in all these initiatives that we seeing So at core we seeing heightened geopolitical competition politically economically militarily Technology is at the core of this. In addition, we're seeing countries around the world reassess their dependencies on the United States and hence their dependencies on U.S. cloud providers. The push was this was following Russia's invasion of Ukraine. They saw the effect of sanctions and how it decoupled countries from the economy. And now they're concerned about could they be decoupled from the global cloud? This is transforming into a push beyond things just like data sovereignty, which we've had for many years, physical sovereignty, local data centers in country. And we're seeing pushes for things like operational sovereignty, meaning countries' own citizens are going to operate the cloud. We're seeing a push for independent governance and ownership. So by that, I mean countries are insisting that the cloud be incorporated in country where they have jurisdiction and it's subject to their laws. And in some cases, we're even seeing a push to say it has to be locally owned by their own cloud providers. And so this is putting the cloud in the crosshairs of geocompetition. So I love how you've really broken this down, Bill, right? Because there are all these conversations about sovereign AI and sovereign cloud, but this really lays out some of the specifics. Could you explain, you know, your paper talks about this as a conundrum. What's the conundrum here? Right. So the challenge is when countries do all these things, they get greater control. But it doesn't really give them better technical security. In other words, the encryption in the regular cloud is essentially the same as encryption in a sovereign cloud. And in addition, all these sovereign controls, there's over 200, add a lot of operational friction that raises cost easily 20 to 30 percent. It undercuts innovation. And so at what cost are they achieving this control that doesn't actually provide better technical security? and I think we have to ask, you know, is this sovereignty an illusion? So for example, what happens when a server breaks? Servers are made in Taiwan. What happens when a storage drive fails? How do you do a software update So are you really getting control from a sovereign point of view and at what cost So that really explains some of the trade that countries face in terms of gaining technological control over the cloud versus innovation and security. But the question is, how do we move forward? Because I don't think that some of these things are going to go away. I think that countries are going to continue to pursue this. What's a path forward that can allow countries to achieve both of those goals. So all these forces are pushing countries for their own AI initiatives. And what we're seeing is that over 72 countries, so really most of the world, is now pushing for sovereign AI initiatives. So Bill, you've explained some of the challenges that countries are confronting. How do we move forward in a practical way that allows them to achieve their goals, but doesn't run into some of the problems that you've identified? So we've proposed an integrated set of recommendations. At its highest level, countries have choices. They can pursue strict autarkic nationalistic clouds or they can pursue sovereign clouds with their partner. And there's suggestions and recommendations for things that they should not do. So, for example, we've argued, don't try and build your own national cloud and operate it. There's a long history of that failing. Another thing they should not do is make rules and regulations based on national ownership or country of origin. That just blocks themselves from partners. And there's also a set of things that countries actively should do. So one thing they should do is take a risk-based approach. And what we mean by that is risks that are the greatest, certain workloads have a lot of risk, and that's where they should impose strict sovereign clouds. But the flip side of that is commercial activities that do not pose undue risk should not be burdened with sovereign cloud requirements. Another key thing they should do is training. It sounds easy, but countries can achieve a lot of their sovereign goals in traditional cloud regions just by architecting cloud properly. But that takes a specific set of skills. Another key thing they can do is run confidence building measures with their cloud partners. They can run live cloud workloads and have their own technical people test for security for reliability for compliance And that of course builds trust which is what we trying to get to Lastly for AI specifically countries need to think a lot more about what does sovereignty actually mean. Some are trying to build sovereign AI as an end-to-end stack, and they need to think about what's practical, what's desirable. And our recommendation is focus on the high end of the stack, national data sets, national models, and particular AI applications. That's the heart of sovereign AI. And we would advise them to avoid focusing on IT infrastructure, storage, networking, servers. There's very little about IT infrastructure that's specific to a country's heritage or culture or language. So in sum, despite all the changes, countries still have choices to make. And we would argue that countries should pursue sovereign cloud in partnership with allies and partners rather than pursuing the strict sovereign autarkic nationalist clouds. In many ways, you could compare this to the California gold rush and a huge economic opportunity here in the United States. But the difference was that all people needed in order to do it was picks and shovels. In this case, when we're talking about that IT infrastructure, We're talking about tens or hundreds of billions of dollars. We're talking about massive data centers. We're talking about billions of dollars worth of chips. We're talking about upgraded energy infrastructure. And so that's really where we need allies and partners with the United States in order to have that technological innovation and that ability to make investments to have truly advanced infrastructure. Well said. So I'd like to thank everyone for joining us. I think this does a really excellent job of laying out what Bill's attempt to the solution is here at CSIS. This is an area that's going to continue to be a lot of debate, a lot of discussion. Really encourage everyone both to read the paper and to engage with Bill on the Strategic Technologies Program and to make sure that we have a focus on what this is trying to do, which is to find really practical solutions to some of these challenges that are going to be critical to the AI race. So thank you for joining us. That's it for this episode of Cash Me If You Can. Don't forget to subscribe and follow CSIS for more deep dives into the technology shaping our future.
