Despite growing political resistance, investment in data centers isn't slowing. Ariana Salvatore explains why supply constraints may actually accelerate AI capital spending.

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, I'll be talking about why we still expect robust AI capital spending in spite of some rising political pushback.

It's Thursday, July 23rd at 10am in New York.

It should be no surprise to our listeners that data center pushback, a topic that we've been following for some time, has been growing louder. But in 2026, it's accelerated meaningfully. Data we track suggests that an estimated $156 billion of projects were canceled or delayed in 2025. This year alone, in just the first quarter, we've seen almost that same exact number.

The opposition is coming from several directions.

Communities are raising concerns about rising electricity bills, environmental pressures related to water use, and the local quality of life effects of large-scale construction. But it's also coming from lawmakers across the aisle. State legislatures with both Democratic and Republican lawmakers have been advancing this type of policy.

At the same time, we're forecasting a little less than a trillion dollars of AI CapEx this year alone, and we think it's an increasingly important component of the macroeconomic growth outlook.

So how do we square that circle?

First, and most importantly, we think this is primarily a supply-side risk rather than a demand-side one. We don't expect the backlash to materially reduce projections for compute demand. Instead, it could widen the gap between that demand and the industry's ability to bring new capacity online through things like permitting delays, grid interconnection constraints, and local opposition.

Despite that more difficult political and infrastructure environment, our internet team, led by Brian Nowak, remain constructive on AI capital spending. Our broader thematic estimate for total AI CapEX, including the neo cloud providers, stands at approximately $870 billion in 2026, and we actually see risks skewed even higher from here.

So why is spending still increasing as the environment for building data centers becomes more challenging? There are a few reasons.

First, the AI ecosystem remains compute constrained. The urgency to invest has not diminished. In fact, growing social opposition and political uncertainty ahead of the 2028 presidential election may actually be encouraging hyperscalers to begin projects earlier, which our credit strategists outline as a potential scenario here. A pull forward of demand before the political and execution risk grows even louder.

Second, the timelines associated with data center construction have become longer. From groundbreaking to operational launch, projects can now take as long as three years or even more. That gives companies a strong incentive to begin developing future capacity well in advance, even if the political pushback is strong.

And third, the underlying demand signal is not slowing. Global weekly token usage, which our analysts view as an important proxy for compute demand, has increased since early January. It's rising and continues to do so throughout the course of this year.

So, in short, the pushback is real, but it appears to be reshaping the build-out rather than stopping it.

That's why our base case is for a conditional build-out. We think projects are likely to face greater scrutiny, we think projects are likely to face greater scrutiny, longer delays, and more requirements related to environmental impact and community benefits.

But ultimately, we still think they cross the finish line. That could mean higher costs, it could mean longer development timelines, and greater geographic dispersion of projects away from the largest existing data center markets.

It could also accelerate the shift toward on-site and behind-the-meter power generation. Fuel cells, turbines, and energy storage are becoming increasingly important as operators look for ways to reduce their reliance on these lengthy grid interconnection processes, and that can benefit companies that are able to bring those solutions to the forefront.

Meanwhile, our U.S. equity strategy team maintains a relative preference for hyperscalers over semiconductors over the next several months. As you heard our CIO and Chief Equity Strategist Mike Wilson explain yesterday, that's because the team sees the hyperscalers as early in discounting the market's renewed focus on CapEx discipline.

Putting it all together, we see the growing pushback against data centers as representing a genuine risk to the pace, cost, and geography of the AI infrastructure build-out.

But again, this isn't just a demand story, it's a supply story. And somewhat paradoxically, the scarcity and the uncertainty created by these constraints could actually end up pulling capital spend forward rather than reducing it.

Thanks for listening. If you enjoy the show, please leave us a review wherever you listen and share thoughts on the market with a friend or colleague today.