The Big Data Center Debate
Monday's $500 billion Nvidia deal thrusts data centers further into the spotlight, just a few months before the midterm elections.
On Monday, Nvidia signed MOUs with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish independent compute financing platforms, aiming to mobilize more than $500 billion of third-party capital for AI infrastructure over time. Special-purpose entities will hold the GPUs, issue privately placed bonds and credit notes against them, lease the compute to Nvidia’s customers, and service the debt out of usage revenue.
Jensen Huang called it “a new class of productive, investable infrastructure.” David Solomon called it a new credit market. Larry Fink pitched it as yield for investors who are overweight equities – and a huge job creator. Jim Zelter suggested that this deal signifies American Exceptionalism and will “benefit the U.S. economy as a competitive tool for the advancement for the next decade.”
All this roughly ninety days before a midterm election in which data centers have become a major hot-button issue across the entire political spectrum.
Fortune’s headline the next day read: “Nvidia found a new way to keep the AI boom funded: your retirement money.” Whether that’s a fair characterization is a completely separate question from whether it works as a message.
The Data Center Backlash
The scale of local opposition is easy to underestimate from inside the tech industry. A March Gallup poll found 71% of American adults oppose an AI data center being built in their local area – higher opposition than Gallup has ever recorded for a nearby nuclear plant. More than 100 local moratoriums and 300+ state bills were filed in the first six weeks of 2026. At least 75 projects worth roughly $130 billion were stalled or blocked in Q1. Maine banned construction outright. Arizona, Illinois and Ohio paused their tax incentives.
This backlash is now showing up in the supply data. Capacity under construction fell at year-end 2025 for the first time since 2020. Wood Mackenzie found new pipeline additions dropped roughly 50% in Q4 2025. Sightline Climate expects 30-50% of the 2026 pipeline to slip. Policy has become a gating factor on delivery.
The Data on Pros/Cons is Inconclusive
The empirical record on the benefits and drawbacks of data centers in local communities is truly mixed – which means the bulls and the bears both have their talking points.
On jobs, two independent studies landed in roughly the same place. A Brookings study built a synthetic control across ~770 facilities and found total private employment up 4-5% over five to six years in treated counties, with information-sector employment up 22%. A Georgia Tech study found employment up ~0.9% in the first three years, growing to ~3.5% long term. Both surfaced the same asymmetry: counties with a single facility see essentially no ecosystem effect, while counties with four or more see 23% information-sector growth.
On electricity prices, the conflict is sharper. PJM – the largest power grid operator in the United States – ran its own independent market monitor that attributed 63% of the 2025/26 capacity cost increase to data centers; capacity costs went from $2.2B to $14.7B in a single year. Meanwhile, The Electric Power Research Institute (EPRI) (an American independent, nonprofit organization) published a working paper in June that found the opposite – data centers lowered average retail rates from 2015 to 2024, with every 10% increase in capacity reducing residential prices about 0.4%.
Both can be true. The EPRI paper measures average embedded cost over a period that predates AI-scale load; PJM’s capacity market measures marginal cost under a binding supply constraint. The authors themselves note the effect reverses if supply falls behind demand.
If we take these studies at face value, that creates quite the paradox – if we as a country build enough supply, retail energy prices will likely go down. But if we fight supply growth because a supply constrained grid increased retail energy prices, the retail prices will go up further. The antidote for data centers’ impact on local retail energy prices may be… more data centers.
The Arguments For and Against
There are a lot of valid reasons to be both pro- and anti-data center build outs.
For
Data centers provide a lot of macroeconomic benefit. Take Virginia, for example, which has been amongst the most aggressive states in building data centers the past few years. Their government ran a study that credits the industry with roughly 74,000 jobs, $5.5 billion in labor income, and $9.1 billion in state GDP; data centers paid $733 million in Loudoun County property taxes in 2024, about a third of the county’s total. Georgia found similar results in their state.
Additionally, the AI buildout contributed roughly 0.8% of U.S. GDP by early 2026, and one economist calculated that information-processing investment drove about 92% of GDP growth in the first half of 2025.
Also keep in mind that China and other authoritarian regimes aren’t going to wait around as they plan their own data center build outs. Compute capacity is quickly becoming foundational to defense, intelligence, and frontier research – falling behind on the data center build out is a risk to American hegemony.
Against
On the other hand, the cost story I alluded to above is still contested – and American consumers are very sensitive to our cost of living.
While data centers can have really positive local economic impact, most states subsidize data center builds with an equipment sales-tax exemption that can cost billions annually – it cost Virginia $1.6 billion last year.
And lastly, there are a few local externalities of data centers – water consumption, noise, transmission corridors, farmland conversion – all variables that give locals cause for concern.
So that’s the objective overview of how folks on both sides of this issue are assessing it. Now for a little editorializing.
The “water consumption” argument is a little silly – most of the water in a data center is used in a closed loop for cooling racks, and data center water usage is very small in comparison to other industries that get no pushback on this issue. I also find it very telling that more Americans oppose data centers than have ever opposed nuclear plants – in a world where the Chernobyl and Fukushima Daiichi disasters both happened.
I believe the data center debate is a product of populism more than anything else. Whether energy prices go up or down from here is going to be irrelevant for years – both sides already have their data talking points ready. And it’ll be 5-10 years (or more) until we have conclusive data that overcomes the noise from both sides.
But the average American isn’t going to like that data centers help the hyperscalers. They’re not going to like that Nvidia has a $500 billion agreement in place with the biggest PE firms in the world to fund AI infrastructure. The faces of the data center buildout are Sam Altman, Mark Zuckerberg, Elon Musk, Dario Amodei, and so on. The rise of populism on both sides of the political spectrum suggests that Americans are more concerned with what they have in relation to others, instead of independently assessing how their own quality of life changes.
Where We Go From Here
Hyperscalers and the various layers of government (federal, state, local) have different views on data center implementation. However wide the gap is between them, hyperscalers are going to be responsible for closing 80%+ of it. Governments aren’t going to move that quickly, and if their constituents are opposed to data centers in the first place, that puts the onus on the hyperscalers that much more.
Texas Governor Greg Abbott tweeted last week that data center builds are being paused in his state. Texas is a notably business friendly climate, and Abbott’s list of demands for new data center builds is pretty reasonable:
“Data centers must disclose their plans to:
Pay their own way
Provide their own power
Reuse their own water
Reduce the cost of electricity
& avoid disturbing neighborhoods”
This is a good framework that data center advocates will need to solve for.
They need to answer these questions, and they will. There’s too much money at stake for them not to.