US elections take down AI data centers, and Brazil still hasn't had this discussion

In the November 4, 2026 election, AI data centers became a voting issue in the United States, with Texas electing a governor who promises a moratorium. Brazil, racing to catch up with the same infrastructure boom, still hasn't had this discussion.

Data centers became a campaign issue, and lost

In the November 4, 2026 US election, the Democratic Party regained control of both houses of the US Congress, flipping at least seven Senate seats and winning districts that the Republican Party had won comfortably in 2024, according to a Bloomberg report. Decision Desk HQ projected the Democratic win in the Senate; other news organizations already considered the House decided.

The economy and the war in Iran weighed on the result. But according to analyst Ben Thompson, of Stratechery, who closely followed political advertising while watching baseball games on TV during the fall, one specific topic dominated campaign ads, at least in Wisconsin, where he lives: AI data centers.

In Texas, the state that holds more than 25% of planned data center capacity in the United States through 2030 (a figure cited by Yahoo Finance), the election produced the state's first Democratic governor since 1990. Gina Hinojosa, the state's first Latina governor, campaigned promising a moratorium on the construction of new data centers. The detail that shows the scale of the problem: Republican governor Greg Abbott, her predecessor, had already decreed his own moratorium on new data center permits in September 2026, before the election. Both sides of the Texas political spectrum arrived at the same position before a single vote was counted.

It wasn't an isolated case. Thompson cites a survey that mapped local votes on data centers across the country: nearly all the measures opposing construction won or were winning the morning after the election, with a single exception, in North Dakota. For anyone following the sector closely, this is the most important data point of the night: it wasn't a one-off result in Texas, it was a national pattern.

Thompson's thesis: the veto became the real bottleneck

Thompson has been documenting resistance to data centers since the beginning of the year, but the election result made him update his thesis. After watching the industry resolve, one by one, the bottlenecks in computing, memory and energy, the AI sector has run into an obstacle that can't be solved with more capital: political permission to build.

First we were short on compute, but that ramped; now we are short on memory, but that will ramp; we were coming close to being short on power, but that will be rectified. The final constraint is political, and that may be the most imposing constraint of all.

Ben Thompson, analyst and founder of Stratechery

The difference from previous bottlenecks is structural: capital solves a chip shortage, a supply agreement solves an energy shortage, but there's no check that buys permission from a community that doesn't want the project across the street. And unlike globalization, when factories closed and entire cities lost their reason for existing without anyone being consulted, building a data center depends on permits, zoning and, increasingly, votes. It's the first time, Thompson writes, that the population affected by a technological wave has any veto power over it before the fact is already settled.

Who wins, who loses on the American chessboard

Thompson lists ten possible implications of this new balance. Three movements that organize the rest are worth highlighting:

In short: whoever already has scale suffers less and may even come out celebrating. Hyperscalers use the opposition as an excuse to moderate investment without losing ground to competitors, since everyone faces the same brake. Frontier labs like Anthropic gain financial breathing room because the push for efficiency reduces the R&D race that threatened to outpace their own revenue. The real losers are the startups that depended on cheap access to surplus computing, which is becoming increasingly scarce.

The Brazilian mirror

The angle that matters to those building a technology business in Brazil isn't rooting for an American political party. It's the question Thompson leaves open: what happens when the same pressure reaches a market that's still racing to catch up on its own AI infrastructure? In recent years, hyperscalers have announced billions of dollars in investments in new data centers in Brazil, attracted by a power grid that's predominantly renewable and hydroelectric, which removes from the Brazilian debate one of the strongest triggers of American outrage: the fight over scarce energy.

But energy isn't the only resource at stake. Other inputs, such as water for cooling, land and the neighborhood's willingness to go along, could become the new trigger for resistance as the data center boom advances across the region. In practice, it's the same mechanism Thompson describes in the United States: a community that had no voice during globalization discovers it has a voice over the data center next door, and may decide to use it.

The counterargument you can't ignore

There's a strong counterargument, and it deserves to be taken seriously before jumping to any hasty conclusion that Brazil will repeat the United States. Brazilian industrial investment policy works differently: states and municipalities historically compete with each other for factories, warehouses and data centers, through tax wars and tax incentives, rather than mobilizing against them. It's a system designed to say yes to investment, not to block it.

Beyond that, much of the American anxiety stems from a specific narrative: that the AI developed there will replace jobs there. Brazil, today, is mostly an importer of AI infrastructure, not home to a frontier lab directly threatening local jobs. That distance may delay, not eliminate, the kind of political reaction the American election showed. It's a difference of degree, not of kind: the resource in short supply here isn't energy, it could be water, land or the neighborhood's patience, and the spark, when it comes, will probably run through state environmental licensing, not through a specific federal data center bill.

What changes for decision-makers

In short: if environmental licensing and municipal zoning are what will decide the pace of AI expansion in Brazil, the same way they already decide the installation of any heavy industry today, that's the point where a founder or an infrastructure leader should be paying attention now, before the issue becomes election headline material the way it did in Texas.

1. Treat the relationship with the local community as a line item in the project's cost, not as an externality to be dealt with after the problem shows up.
2. Closely track the environmental licensing and water consumption rules in the states where the data center will be located, because that's where, not in some eventual federal AI law, Brazil's dispute will likely be decided first.
3. Use the current window, in which Brazil is still negotiating this boom with communities that haven't seen the worst of the American debate, to build a reputation before distrust arrives ready-made from abroad.

Thompson ends his piece saying that AI wasn't technically on the American ballot, but it was implicit in it, and will continue to be. For those deciding where and how to build AI infrastructure in Brazil, the lesson isn't to predict whether the same veto will happen here. It's to assume that, once Brazilian demand for data centers grows enough to become news, someone will already be organizing the opposition, and arriving after that someone costs more than arriving before them.