AI as a Sovereign Power

**Ariana Salvatore** (0:00)
Welcome to Thoughts on the Market. I'm Ariana Salvatore, head of US. Public Policy Research at Morgan Stanley. Today, why is sovereign AI becoming a policy priority around the world? It's Wednesday, July 15th at 10 a.m. in New York.
The AI controls debate used to be focused on chips. Cutting edge semiconductors are essential to train large AI models after all. But over the past year, the debate has moved well beyond that narrow focus.
The policy conversation has broadened beyond things like which advanced semis could be sold to China. The bigger question now is, who controls the full AI stack? Chips, cloud infrastructure, frontier models, data centers, cybersecurity standards, and the energy systems that support all of it. That's what we mean when we talk about sovereign AI. At the simplest level, it's a country's ability to develop and deploy artificial intelligence using its own infrastructure, data, workforce, and technology ecosystem.
But sovereign AI is also about reducing strategic dependence on foreign platforms and foreign-controlled supply chains. That echoes a trend toward multipolarity that we've been writing about since back in 2018
Countries around the world are prioritizing national security over economic efficiencies. We see that theme applying to AI as well.
So what does all this mean for markets? First, sovereign AI turns AI infrastructure into a matter of national industrial policy. Data centers, power availability, and grid reliability are just a few examples of components that are becoming strategic assets. That means governments are likely to play a larger role in deciding several aspects of the AI buildout. Where it's built, who finances it, and which countries get access to the most advanced parts of the stack. Second, sovereign AI reinforces the shift toward de-risking in a more fragmented international order. The US is trying to promote the export of an American AI technology stack to allies and partners. At the same time, it's preserving national security guardrails around the most sensitive capabilities. Meanwhile, we see China trying to indigenize as much of the technology as possible, from chips to cloud to model deployment. Other countries are navigating between the two. Third, and most importantly, sovereign AI is also an energy story. Who gets to build and benefit from AI increasingly depends on access to low-cost, reliable power. That makes energy availability a competitive advantage, and it also makes energy affordability a political constraint. That dovetails with one of our thematic predictions heading into this year, the politics of energy. We see rising power costs as a more visible political issue. That's led to backlash against data center development. There's more local opposition to new projects and greater pressure on policy makers and utilities to make sure that existing rate payers are not subsidizing AI-driven grid investment. We think that could push AI infrastructure in a few directions. One is toward a conditional buildout. Here offsets like large load tariffs and other cost allocation mechanisms are designed to protect households and small businesses. Another direction is policy support for the lowest cost sources of energy, even where that might create tension with emissions objectives. And a third direction is more off-grid or behind-the-meter power solutions. That would include things like fuel cells, storage and other time-to-power strategies. So data center developers can secure electricity without intensifying local affordability concerns.
The pursuit of sovereign AI comes with many questions around inflationary impacts. Compute and power are both constrained, regulation remains uncertain, and there could be more limitations on things like tech transfers if the government sees a national security edge. So to the extent that countries want to reduce their dependencies, it may cost more to get there. There are, however, companies that can benefit in this environment, but there's also a policy risk. We're left with a more reactive policy environment. Selective access in some areas, tighter controls in others, and ongoing uncertainty around how Washington will treat advanced chips, cloud infrastructure, and frontier model deployment. Now, that uncertainty matters because it affects corporate planning, cross-border investment, and the shape of global AI alliances.
So what does all this mean for investors? More and more governments view AI capability as a source of economic power and geopolitical leverage. That means the AI race is moving from a question of who builds the best model to who controls the infrastructure, standards, supply chains, and energy systems that allow those models to scale.
In our view, that means sovereign AI is one of the most important themes to watch in the next phase of the AI buildout. And we'll be coming back to this topic soon. In the coming weeks, Stephen Byrd and I will talk about sovereign AI in more depth, particularly around what it means for power demand, data center investment, energy affordability and the broader infrastructure required to support the next stage of AI adoption.