A farmer left land for a park. The city is paving it for a data center.

What happened

In 1953, a Peoria-area farmer named Carl Wilkinson deeded farmland to the city of Peoria, Illinois with a written purpose: it should become a public park. For seventy-odd years it sat mostly undeveloped — fields, treelines, the kind of low-utility green space that survives only because no one has bothered to monetize it. According to 404 Media's reporting, that calculus just changed. The Peoria City Council voted to rezone and lease the parcel as part of a roughly 400-acre site for a hyperscale data center, with a developer pursuing what local filings describe as a multi-building campus capable of drawing hundreds of megawatts.

The city's argument is straightforward: the deed restriction is old, the donor is long dead, the heirs were notified, and the tax base needs the win. The data center developer — whose ultimate tenant is, per the 404 Media piece, still concealed behind an LLC and an NDA — is offering the kind of capex number that makes a mid-sized Midwestern city's annual budget look like a rounding error. Residents who showed up to the council meeting cited the deed, the trees, and the principle. The council cited the assessment roll.

This is what land-use politics looks like in the AI buildout era: a 73-year-old handshake getting voided in a single council session because a hyperscaler's site-selection team put a pin on a map. The 404 Media reporters obtained the rezoning packet and found what's now a familiar template — anonymous LLC, generic "technology campus" language, power and water commitments that dwarf anything else the municipality has ever signed.

Why it matters

The Peoria story is not really about Peoria. It's about the operating model that AI infrastructure has imported into municipal politics, and how badly that model overruns the slow-moving instruments — deed restrictions, conservation easements, comprehensive plans — that small cities relied on to keep promises across generations.

The site-selection playbook is now ruthless and fast. A hyperscaler or a colo developer working on their behalf identifies candidate parcels filtered by three things: substation proximity, water rights, and a municipality desperate enough to move quickly. The pitch arrives as an LLC with no public parent. Cities are told the deal evaporates if the tenant is named, if the timeline slips past 90 days, or if the rezoning goes to a full public referendum. The NDA isn't incidental; it's the mechanism. It lets the developer extract concessions from the city before voters can organize around the actual buyer's identity.

Compare this to the previous generation of large industrial site fights — a Toyota plant, a Foxconn campus, an Amazon HQ2 bid. Those took years and were fought in the open, with named counterparties and visible incentives. The AI data center cycle has compressed that timeline by an order of magnitude while simultaneously hiding the tenant. By the time a community knows whether the building will house Microsoft, Meta, xAI, or a CoreWeave-style neocloud, the slab is poured.

The deed-restriction question is the genuinely interesting legal wrinkle. Courts have historically been willing to lift old donor restrictions under the doctrine of *cy pres* — "as near as possible" — when the original purpose becomes impractical. But "impractical" used to mean the land had been swallowed by a highway or the neighborhood had changed beyond recognition. Peoria is arguing, in effect, that a 400-acre park is impractical because a 400-acre data center is more valuable, which is a definition of impracticality that swallows every charitable land bequest in American history. If that argument holds, every donor-restricted parcel near a substation is now in play.

The community reaction in the HN thread (327 points, ~600 comments) split along predictable but useful lines. One camp argued the donor's intent should be sacred — that breaking a 1953 promise teaches every future donor not to bother. Another camp pointed out the obvious: the land had been sitting unused for decades, and a tax-generating facility serves the public more than fallow fields. The most interesting comments came from people who've worked municipal site selection. Their read: the city probably had no real leverage. Once the developer threatens to walk to the next county, the council's choice is between the data center and nothing — because the park was always going to be "nothing" on a P&L.

What this means for your stack

If you ship anything that runs on a hyperscaler's infrastructure — and you do — the Peoria pattern is now a load-bearing assumption in your cost structure. The AI capacity buildout is being subsidized by exactly these kinds of fights: tax abatements measured in decades, power rates negotiated below industrial tariff, water commitments that quietly preempt residential growth. Your $0.0001-per-token inference cost is, at the margin, paid for by a dead farmer's broken bequest. That's not a moral argument; it's a supply-chain fact. When the political backlash to these deals finally catches up — and it will, because the deed-restriction stories are unusually sympathetic — capacity costs will reprice.

The practical implication for engineering leaders: stop treating compute capacity as a commodity with a smooth supply curve. The next two years of GPU and inference availability depend on roughly two thousand site fights like Peoria's playing out in favor of the developers. Some will lose. Northern Virginia is already pushing back. Memphis has organized opposition to xAI's gas turbines. Ireland and the Netherlands have moratoria. The cities that say yes are increasingly the ones with the weakest civic infrastructure to say no — which is also where your latency and redundancy story gets worse, not better.

For anyone building on top of this layer: pay attention to which regions your providers are quietly de-emphasizing in their capacity announcements. That's the leading indicator. The lagging indicator is your AWS bill.

Looking ahead

The Peoria deed will probably end up in court, and the city will probably win on cy pres grounds, because American property law has a strong bias toward letting current owners do what they want with land regardless of what previous owners intended. But the precedent matters less than the pattern. We are in the early innings of a multi-year cycle in which donor restrictions, conservation easements, agricultural preservation districts, and historic designations all get tested against the largest infrastructure buildout since the interstate highway system. The deals will keep getting done. The promises will keep getting broken. And somewhere a developer is already drafting the LLC paperwork for the next 400 acres.