The AI industry has decided to be a factory — and the land is closing

"Factory" is the word the industry has settled on. DigitalBridge's chief executive describes a $900 billion-a-year build-out of what he calls "AI factories," and Jensen Huang has adopted the same term ^[1]. The word is not a metaphor. A factory needs land, a grid connection, water, and a tax abatement the way a steel mill needs a rail spur — and the AI industry has now decided that is what it is. The open question is whether it can build one. The decision was forced. MIT's Project NANDA found 95% of enterprise generative-AI pilots have produced no measurable P&L impact, and ServiceNow's index shows spending up 110% while maturity sits at 51 out of 100. OpenAI missed its internal revenue targets and its one-billion-user goal, and its CFO warned the company may struggle to fund $600 billion in compute commitments without faster growth ^[2]. Sam Altman has said the quiet part himself.

When bubbles happen, smart people get overexcited about a kernel of truth. — Sam Altman

OpenAI projects a burn rate of up to $140 billion a year by 2029 and does not expect to be cash-flow positive until 2030 ^[3]. The industry's answer to that gap arrived on a single day. On April 24, OpenAI, Microsoft, Google, and Anthropic all launched enterprise agent platforms at once — a coordinated pivot from selling model intelligence to selling the deployment layer that sits between the model and the company ^[4]. OpenAI cut its Luna model prices 80% to chase enterprise customers ^[5], created a deployment subsidiary, and filed confidentially for an IPO; Anthropic, at a $965 billion valuation, filed too ^[6]. The industry is acting as one organism, and the organism has decided what it is: a factory. A factory needs a physical home, and the homes are closing. Between January and March, at least 75 data center projects worth nearly $130 billion were blocked or delayed — the highest quarterly obstruction rate since 2023 ^[7]. New York imposed the first statewide moratorium on large data centers in July ^[8]. Indianapolis froze new projects through 2027, and Calvert County passed a moratorium after Amazon withdrew a $22 billion proposal ^[9]. The reversal is bipartisan, which is what makes it durable — no party is running against the backlash. In Kentucky, a state that expanded data center incentives to all 120 counties just a year earlier, Democratic Governor Andy Beshear is now calling for their repeal.

I don't think we should be providing them, and so I would certainly support rolling back any of that. — Andy Beshear

His co-sponsors include Republicans ^[10]. In Texas, the most data-center-friendly state in the country, Republican Governor Greg Abbott is proposing to repeal the incentives and add water rules.

Data centers must operate in ways that reduce costs for residential electricity customers, do not drain water needed for our communities, and take into consideration the needs of our neighborhoods. — Greg Abbott

^[7] The public is ahead of the politicians. Gallup found 71% of Americans oppose AI data centers in their neighborhoods — higher than opposition to nuclear plants ^[11]. A polling firm tracking the shift called its speed rare.

It’s rare to see political attitudes change this fast in an era where they are often quite static and partisan. — Public First

^[12] Now the market itself is pricing the friction. ERCOT power futures signal a slower build-out than expected — even in Texas ^[13]. JPMorgan, Morgan Stanley, and Bank of America have folded community sentiment and permitting readiness into their credit checks on data center loans. Bank of America's definition of readiness is telling.

Readiness means all the permitting and approvals that are required, and the community support from the people who are going to live around it. — Bank of America

^[14] To be clear, the machine is not stopped. Nearly three-quarters of states still offer incentives, and some counties are still competing for the investment ^[15]. But the ERCOT signal is the tell: if the market is pricing friction in Texas, it is pricing it everywhere. The contradiction is concrete, not rhetorical. The industry is filing IPOs, cutting prices, and embedding engineers to become a factory ^[6] — and the land, power, and tax breaks a factory needs are being pulled back by the communities it would be built in. One side has decided what it is. The other is deciding, county by county, whether to let it exist.