The conventional read on AI companies flooding the 2026 midterms with $265 million is that Big Tech is flexing its new political muscle, buying the regulatory environment it needs to keep building. That reading isn’t wrong — but it misses the more uncomfortable story underneath it.
The money is a symptom, not a strategy. And it may be making the politics worse.
The Reading
What Actually Happened
According to a Wall Street Journal analysis, AI companies have committed roughly $265 million to super PACs and political groups ahead of the November 2026 midterm elections — making AI, alongside cryptocurrency and sports betting, the leading industry by political spending this cycle. By June 2026, corporate midterm spending overall was already running 40 percent ahead of the entire 2024 election cycle, per reporting from OpenSecrets.
The headline donors are familiar names. The super PAC Leading the Future has raised $140 million, including $50 million from venture firm Andreessen Horowitz and a combined $25 million from OpenAI president Greg Brockman and his wife Anna, according to Federal Election Commission filings. Anthropic is linked — through a dark money nonprofit called Public First Action — to at least $40 million in contributions from a reported $100 million war chest. Elon Musk’s super PAC is operating at a similar scale, and Meta is tied to multiple PACs with tens of millions ready to deploy.
OpenAI, for its part, issued a statement clarifying that the company itself has not donated to candidates, campaigns, or PACs, and that the Brockmans’ involvement with Leading the Future was in a personal capacity. The distinction matters legally, even if the line between a company’s founders and the company’s interests is blurry to most voters.
The trigger for all of this is straightforward: the five biggest hyperscale tech companies — Amazon, Microsoft, Google, Meta, and Oracle — are on track to spend more than $560 billion on AI infrastructure in 2026. Data center construction in July was up roughly 60 percent year-over-year, per U.S. Census Bureau data. By 2027, U.S. data center power demand is expected to double. Communities near these facilities are noticing — and reacting.
The Overlooked Angle: Money Can’t Buy What Trust Already Lost
Here’s what the spending figures don’t capture: the communities pushing back aren’t waiting for election results to act. New York issued the country’s first statewide moratorium on large data center construction in July, effective for one year. Red and blue states alike are considering similar bans. A recent Economist/YouGov poll found that only 31 percent of Republicans — and just 14 percent of Democrats — think new data centers are good for America.
President Trump stepped into this debate this week, posting on Truth Social that communities opposing data centers would end up “backwards and poor,” calling the facilities a “Golden Goose” and suggesting critics were acting on behalf of China. He had reportedly seen internal polling showing broad public opposition — and chose to attack it rather than address it. Commerce Secretary Howard Lutnick compounded the problem by claiming data centers “don’t use water,” a statement that contradicts both the scientific record and his own prior public comments acknowledging the facilities’ water consumption.
These aren’t the talking points of an industry winning a public argument. They’re the talking points of one that has stopped trying to win it.
The concerns driving the backlash are concrete and locally felt. An estimated two-thirds of data centers built since 2022 are located in areas of high water stress, many in the western United States. Power demand from these facilities is already straining regional grids, and electricity price anxiety in host communities is real — a dynamic we’ve examined in detail in our coverage of the environmental impact of data centers and the broader question of what AI infrastructure actually costs the planet in electricity, water, and land.
The Evidence: A Race That Said It All
The New York 12th congressional district Democratic primary is the clearest case study available. AI-industry money poured in — an estimated $27 million in AI-related political funds in a single House primary race. A PAC tied to Anthropic’s Public First Action spent more than $13 million supporting Alex Bores, a state assemblyman and former Palantir employee who authored a state AI safety law. A super PAC linked to Leading the Future spent more than $8 million opposing him.
Bores lost, narrowly, to fellow assemblyman Micah Lasher. Lasher was a co-sponsor of Bores’ AI safety bill and supports a temporary national moratorium on data center construction. In his victory speech, Lasher said he would not be “taking my cues” from the companies that intervened in the race.
Brendan Glavin, director of insights at the campaign finance transparency group OpenSecrets, told The Independent that AI companies are running the same playbook crypto used in 2024 — targeting specific races rather than donating to party-wide victory funds, and often running ads that don’t mention AI or data centers at all. “Crypto did this in 2024,” Glavin said. “It was seen as successful. Now you’re seeing it repeated and AI has taken on the same playbook.”
But the crypto analogy has limits. Cryptocurrency’s political fights were largely about regulatory classification — abstract enough that most voters didn’t have a direct stake. Data centers are physical infrastructure. They go up in people’s backyards, draw from the same aquifers, and load the same power grids. The opposition is tactile in a way that crypto deregulation simply wasn’t.
There’s a structural irony here that the spending numbers obscure: the same capital intensity that makes data centers politically valuable — the jobs, the tax revenue, the infrastructure investment — is precisely what makes them politically threatening. The scale of a hyperscale facility means it can’t be quietly tucked away; it reshapes whatever community hosts it. AI companies appear to have assumed the economic benefits would sell themselves. The backlash suggests communities want a seat at the table before the groundbreaking, not a check afterward. As we’ve reported separately, even Amazon’s own engineers have pushed back on the data center spending surge — indicating that the political problem isn’t purely external.
What This Changes
The political implications run in both directions. For AI companies, the 2026 cycle is a live test of whether electoral spending can shift legislative outcomes on infrastructure permitting, environmental review timelines, and state-level moratorium legislation. If it can’t — and early results are mixed at best — the industry may need to reconsider whether community engagement is cheaper than political combat.
For candidates, data center opposition is becoming a viable electoral position in districts that host or are slated to host facilities. That’s new. For most of the last decade, opposing a major tech investment in your district was political risk. That calculus is shifting, and candidates in tight races are watching Lasher’s win carefully.
For the broader AI buildout, the political environment matters practically. Permitting delays, state moratoriums, and utility commission pushback can slow construction timelines significantly — and the hyperscalers’ $560 billion capital commitment assumes the regulatory path stays reasonably clear. It may not. The Federal Reserve has already flagged data center construction as a contributor to inflationary pressure, adding a macroeconomic dimension to what started as a local land-use argument.
The industry’s preferred framing — jobs, economic growth, national competitiveness — has genuine substance. Data center construction has created a meaningful blue-collar jobs wave in many host communities. But framing alone doesn’t address the electricity and water concerns that are driving the opposition. And when officials make factually incorrect claims about water usage, it actively erodes the credibility of the legitimate economic case.
The Strongest Counterargument
The steel-man case for the AI industry’s political strategy goes like this: democratic systems respond to organized interests, and an industry that fails to organize politically gets regulated by people who don’t understand it. The crypto sector learned this the hard way in 2022 and 2023, when it was largely absent from electoral politics and watched hostile legislation advance almost unopposed. By building political infrastructure now — even imperfectly — AI companies are ensuring that their perspective is represented when consequential decisions are made about permitting, energy policy, and environmental review standards.
This argument is made, in varying forms, by figures like Andreessen Horowitz partners who have publicly argued that AI’s geopolitical importance to the United States justifies aggressive political engagement. The underlying premise — that the U.S. needs to build AI infrastructure at scale to remain competitive — is genuinely contested but not frivolous. Federal officials including those at the Department of Energy have acknowledged the strategic dimension of the buildout.
The counterargument’s weakness, however, is that it addresses the wrong problem. The political opposition to data centers isn’t primarily ideological — it’s not an anti-AI movement in any coherent sense. It’s a land-use and infrastructure conflict playing out community by community. Winning a congressional seat doesn’t resolve a water rights dispute. Electing a pro-AI assemblyman doesn’t accelerate a utility interconnection queue. The mismatch between the tool (electoral spending) and the problem (local permitting and resource conflicts) is significant, and the NY-12 outcome illustrates it plainly.
What I Expect Next
My expectation is that the $265 million won’t buy the industry what it’s hoping for — at least not in the form of a cleared political path for unconstrained data center construction. The opposition is too locally rooted and too bipartisan for a PAC strategy designed in San Francisco to neutralize it from Washington. What I think is more likely: the industry’s political spending produces incremental wins on federal permitting reform and grid interconnection timelines, while state-level moratoriums and local opposition continue to grow in parallel. The net result is a slower, more contested buildout than the capital commitments assume — which eventually shows up in delayed revenue projections and renegotiated capacity timelines for the hyperscalers.
The signal that would prove me wrong is simple: if Lasher or similarly positioned candidates who rode data center opposition into office subsequently vote to ease construction restrictions — because the jobs and tax revenue arguments prove more compelling in practice than in campaign season — then the industry’s long-game bet on economic pragmatism will have paid off. Watch how newly elected “anti-data-center” candidates actually vote on state-level permitting bills in 2027. That’s the real scorecard, not the midterm results themselves.
