To The Moon

Data Centers: AI's Golden Goose or the Next Political Lightning Rod?

AI's physical backbone is being poured in concrete and copper at record speed. JPMorgan expects about $900 billion of global AI capital spending in 2026 alone, and $5.5 trillion by 2030. Nvidia's own forecast is $3–4 trillion of AI data center spending by 2030. The buildout is enormous, and it is now running into three walls at once: politics, power and investor skepticism.

The power-bill backlash

Virginia leads the country with 663 data centers, with Texas second at 405. Together they hold roughly 26% of all U.S. data centers. Most of the political heat comes from the PJM grid region, which covers Virginia, Ohio and Illinois. Home electricity prices there rose about 13% in Virginia, 16% in Illinois and 12% in Ohio over the period CNBC examined. Texas, which runs on its own ERCOT grid and connects new supply faster, saw a much smaller rise of around 3.8%.

The data is messy, though. The Institute for Energy Research, an industry-aligned think tank, found that the ten states with the most data centers averaged 14.46 cents per kWh in 2025, versus 14.39 cents in all other states. An EIA-based analysis found that the biggest home price increases since 2022 were mostly elsewhere, in DC, Maryland and New Jersey. It also found that business power prices in Ohio and Virginia jumped about 20% in the past year, and Virginia's commercial electricity use grew 62.8% from 2019 to 2026. Higher prices showing up in the same places as data centers does not by itself prove the data centers caused them, but the PJM states are where the argument is hottest.

Politics turns it into a ballot-box issue

What began as zoning disputes is now a campaign theme. Democrats won in Virginia, New Jersey and Georgia in 2025 on platforms promising lower power bills and making data centers pay more. One report puts public opposition to data center construction at 71% of Americans, and NPR found over $45 million in midterm ad spending mentioning data centers between January and August 2026. Opposition is not purely partisan: local fights cross party lines, and some Republicans have criticized data centers as electricity hogs. Senate Democrats blocked a Republican-backed bill on data center electricity costs, arguing it was optional rather than mandatory. Expect more special rate classes, tax-break rollbacks and moratorium proposals. For developers, that is a real risk to timelines.

Market sentiment: from "build at any cost" to "show me the returns"

Investor mood has shifted. For the first time since ChatGPT launched in 2022, commentators note that investors are weighing the cost of capital against the return on capital employed. A Motley Fool survey found 41% of investors suspect an AI valuation bubble, while only 26% see current valuations as sustainable. Bank of America has described the risk as an "air pocket" driven by mounting data center debt rather than a classic bubble. Meanwhile the five largest hyperscalers are projected to spend more than $700 billion on AI data center capex in 2026, and recent selloffs have shown how quickly the tape reacts to capex guidance and debt-funded builds.

The AI playbook from here

My read: the playbook is not dead, but it is maturing. The easy phase where everything with an AI label rallied looks over. The winners from here will need three things: real revenue, financing discipline and access to power. The bottlenecks are moving beyond chips to energy, materials, data center capacity and memory, which creates a second tier of beneficiaries. The tail risks are also clear: export controls on advanced chips, a credit freeze on data center projects, or a regulatory shock that slows approvals. The middle case, lumpy hyperscaler spending that still creates long-tail demand for suppliers and software, may be the one the market is underpricing