Will Midterms Test the AI Investment Cycle?

The AI investment boom has been a defining force in markets. Our Head of Public Policy Research Ariana Salvatore looks at whether the U.S. midterm elections could change the spending and policies behind it.Read more insights from Morgan Stanley.----- Transcript -----Ariana Salvatore: Welcome to Thoughts on the Market. I'm Ariana Salvatore, Head of Public Policy Research at Morgan Stanley. Today, what the midterms could mean for the AI investment cycle. It's Wednesday, October 7th at 10am in New York. I've spoken on this podcast before about the broader implications of the midterm elections. This time, I want to go a little deeper on what they could mean specifically for AI sentiment. We think about that through two separate channels. The first is data center opposition, where we do think the elections can be an important catalyst. The second is broader AI safety and regulation, where we think the election outcome matters much less. And that distinction is important. It suggests that the midterms are more likely to influence investor confidence in the pace and the durability of AI infrastructure spending, rather than fundamentally rewriting the rules governing AI. Let's start with data centers. Our base case is still a conditional build-out. We still expect substantial AI infrastructure investment, but increasingly subject to conditions around things like power costs, grid investment, siting, permitting, water use, and community impact. We don't expect a broad federal pause on data center development. AI compute is increasingly being viewed in Washington as strategic infrastructure, particularly when you consider competition with China. So, we think that national security framing should remain supportive of the build-out. But here's the important point for these elections. Many of the policy levers that can actually delay a project sit well below the federal level. Think about things like interconnection approvals, large load electricity tariffs, zoning, water permits, and tax abatements. Those are all generally controlled by states, utility commissions, and local governments. So, when it comes to AI infrastructure, we actually think governors, state legislatures, and public utility commissions may ultimately matter more than control of any one Senate seat in particular.And that brings us to sentiment. At the federal level, we think a Republican sweep would likely be the most constructive outcome for AI infrastructure sentiment. Now, that's because investors would likely expect fewer regulatory constraints ahead. As well as a greater likelihood of active support for permitting reform, additional power generation, and development on federal land. A divided government outcome, we think, looks closer to the status quo. It would preserve questions about the durability of the build-out, but the gridlock in D.C. would also leave many of the substantive decisions at the state and local level. And lastly, we think a Democratic sweep would be the least constructive outcome for sentiment. Now, importantly, that doesn't mean that we expect a nationwide data center moratorium, as I said. Rather, investors could interpret Democratic outperformance as increasing the probability of tighter local restrictions in the near term, and potentially much more federal scrutiny after the next 2028 elections. So that's the infrastructure side. What about AI regulation more broadly? Here, we think the midterms are actually much less consequential. Our expectation remains for incremental and fragmented regulation rather than a sweeping new federal regime. AI safety, data governance, and frontier model oversight, we think can certainly see targeted action. But we just don't think congressional composition by itself is likely to trigger comprehensive legislation – unless there's a sufficiently high-profile safety or security incident. Executive agencies are also likely to remain the more important actors on tech restrictions and model access. And that gets us to the main takeaway for investors. The midterms probably won't determine whether the AI investment cycle continues. We think the strategic case for expanding U.S. compute capacity remains intact. What they can influence, however, is the friction around that build-out. Where the projects get built, how quickly they receive approval? Who bears the cost? And ultimately, how confident investors are in the durability of AI CapEx? So, when we think about what happens in November through an AI lens, we'll be watching D.C. But in many cases, the more important signals can come from gove