In Port Washington, Vantage's $15B OpenAI–Oracle Stargate campus turned farmland into a lottery ticket — some sellers cleared millions, some wish they'd held out, and the neighbors who didn't sell face eminent domain. It's also a preview: the same power-first, low-opposition playbook is now spreading to 10 emerging markets, from Reno to Tulsa to New Mexico's $165B Stargate site.
An Inc. story this month put a human face on a number we track constantly: what a megawatt of AI infrastructure is actually worth to the developer — and how little of that value the first sellers tend to capture.
The setting is Port Washington, Wisconsin, a town of about 12,000 on Lake Michigan in Ozaukee County. There, Vantage Data Centers is building the "Lighthouse" campus — a $15 billion-plus development that is part of the OpenAI–Oracle Stargate buildout. It sits on roughly 672 acres of former farmland about a mile inland from the lake, and is slated for completion around 2028.
For the farmers who sold, the offer was life-changing. For the neighbors who didn't, the same project is arriving as a condemnation notice.
The land didn't change hands in one splashy announcement. Developers assembled it the way they usually do — quietly, parcel by parcel, often through blandly-named LLCs so no single seller could see the full picture. By the time the scale was public, buyers had reportedly spent $125 million-plus assembling more than 1,500 acres in and around Port Washington. One parcel alone sold for $9.3 million.
The per-acre math is what turned heads. Prime Wisconsin cropland typically trades around $5,000–$10,000 an acre. Sellers here were reportedly offered as much as $120,000 an acre — more than ten times agricultural value. That's how a working farm becomes a seven-figure check.
Here's the part the headline captures: some sellers wish they'd held out. That regret isn't irrational — it's the predictable result of an information asymmetry.
When you're the first to sign, you're negotiating against a developer who knows exactly how many acres it needs, how the parcels fit together, and what the whole campus is worth. You know none of that. Early sellers priced their land against farmland comps. The developer priced it against a $15 billion campus that couldn't be built without it. Sellers who held their ground longer — or whose parcel turned out to be a must-have for the site plan — generally did better than the neighbors who took the first "generous" offer.
The lesson isn't "everyone should have held out forever." It's that the developer's walk-away price and yours are wildly different numbers, and only one side knew both.
Selling your land was the good outcome. The harder story belongs to the residents who didn't sell — and are being pulled into the project anyway.
Powering the campus requires up to ~900 MW of electricity, which means new high-voltage transmission lines and substations. Those lines have to cross private property that was never for sale. The American Transmission Company can acquire the easements it needs through condemnation — eminent domain — the government's power to force a sale for "public use." Some property owners, including Wisconsin artist Tom Uttech, are working with legal organizations to fight it.
Sit with the asymmetry: the farmer who sold got $120,000 an acre. The neighbor whose land is taken for the wires that feed the same campus gets an easement valuation — and no windfall. Same project. Opposite ends of the leverage curve.
Vantage's public case for the project is the familiar mix of jobs and growth: 1,000+ long-term jobs, an estimated $2.7 billion contribution to Wisconsin's GDP, and roughly $175 million toward local power, water, and transport upgrades. The company also says a majority of the campus's power will come from zero-emission resources.
Treat those as the developer's projections, not settled facts — the jobs figure in particular tends to shrink between the press release and the finished, largely-automated building (a 200 MW data center typically employs 50–150 people once operational, as we detail in the ERCOT and moratorium pieces below). The point isn't that the numbers are worthless; it's that they're the opening bid in a negotiation, not the final accounting.
The reason this story matters beyond one Wisconsin town: the same playbook is being run in dozens of places at once. A CommercialSearch survey of emerging data-center markets (July 2026) makes the pattern explicit — site selection has flipped from network connectivity toward power availability, cheap land, and "minimal permitting obstacles." In plain terms, developers are hunting for exactly what Port Washington offered: large rural parcels, an accommodating grid, and neighbors who haven't organized yet.
Two of these campuses are literally the same OpenAI–Oracle Stargate program as Wisconsin's — Port Washington in the Midwest, and Project Jupiter, a $165 billion hyperscale facility with Oracle as anchor tenant, in New Mexico's borderplex. This isn't ten unrelated projects. It's one buildout, spreading.
Here are the ten markets on the watch list — and the friction already surfacing in them:
| Market | Why it's on the list | Marquee project(s) | Community friction so far |
|---|---|---|---|
| San Antonio / Austin, TX | Fastest-growing US hub, ~50 facilities | CloudBurst 1.2 GW ($14.5B); Microsoft $1.5B | Water-supply scrutiny |
| West Texas | Off-grid gas + solar, cheap land | GW Ranch 7.65 GW (largest permitted); Meta El Paso $10B | Air permits for on-site gas |
| Columbus, OH | Sales-tax exemption, fast permits | Meta & EdgeConneX, New Albany | Rate-impact questions |
| Reno, NV | Silicon Valley overflow, no corp. tax | Vantage $3B / 224 MW; Switch 650 MW | Growth-pace concerns |
| Kansas City, MO | Fiber crossroads | Google "Project Kestrel" $100B; Meta $1B | Moratorium debates; proof-of-power zoning |
| Salt Lake City, UT | "Silicon Slopes," cold-air cooling | Stratos 7.5 GW | Pushback cut Stratos by 19,000+ acres; Great Salt Lake water |
| Memphis, TN | "Digital Delta" fiber + aquifer | xAI Colossus ~2 GW; Google $1B (W. Memphis) | Water + air-quality fights |
| Omaha, NE | Tier-2 as Northern Virginia maxes out | Google; Meta Sarpy County | Incentive scrutiny |
| Albuquerque, NM | Phoenix / Dallas alternative | Project Jupiter $165B (OpenAI Stargate) | Neighboring-county moratoriums; groundwater |
| Tulsa, OK | Cheap renewable power, big parcels | Meta "Anthem" $1B; Beale "Clydesdale" $3B | Moratorium (Meta exempted) |
Look at the last column. In at least half of these "emerging" markets — Kansas City, Salt Lake, Albuquerque, Tulsa, Omaha — organized opposition, moratoriums, or new zoning guardrails already exist, sometimes before ground is broken. The pattern we flagged in our Morgan Stanley analysis holds: the capital follows the path of least resistance — and the resistance is learning to get there first.
If your metro is on this list, Port Washington is your preview. The window to negotiate opens before the quietly-named LLCs have bought the first 500 acres — not after.
Port Washington is a preview of what a Stargate-scale campus does to a rural land market — and a checklist of what to get right before the offers start:
1. Assume the developer knows more than you do. They've mapped every parcel and priced the whole campus. If offers are arriving quietly through LLCs, that's a signal the assembly is bigger than any one seller is being told. Our Data Centers tab lists the LLCs hyperscalers use so you can connect a local filing to its parent.
2. Neighbors have more leverage together than apart. Piecemeal selling is exactly the dynamic that leaves early sellers with regret. Landowners who compare notes — or negotiate as a bloc — close the information gap the developer relies on.
3. The windfall and the burden land on different people. Sellers get millions; neighbors get transmission easements and eminent domain. A community benefit agreement negotiated before the permit vote is how a town spreads the upside and cushions the people carrying the cost. Model one on the Negotiation Toolkit tab.
4. Fight the fights that actually convert. As we've documented, the data-center defeats that stick often turn on procedure and grid/transmission terms, not vibes — the environmental review, the routing of the lines, the ratepayer impact of ~900 MW of new load. Put those objections on the record with the people who vote. Use the States & Officials tab.
Diagnosing the asymmetry only helps if it points to action. Here is the toolkit communities use to turn a data center from a windfall-for-a-few into shared, durable local value — roughly in order of leverage:
1. Negotiate the land as a bloc, not parcel by parcel. The single biggest value leak in Port Washington was sequential selling. Landowners who form a negotiating group — or grant a shared option to one broker or attorney — deny the developer its favorite tactic: picking people off one at a time at farmland prices. If the campus can't be built without your collective acreage, price it against the campus, not the crop. It works: in Salem Township, Pennsylvania, 96 landowners pooled ~1,700 acres and sold together to QTS (a Blackstone company) for $586 million — about $330,000 an acre — and a second neighborhood bloc has since lined up a ~$1.2 billion follow-on. Same AI land rush as Wisconsin; opposite outcome, because they moved as one. (The Toolkit tab now has a "Negotiate as a bloc" playbook, a model no-individual-deals clause, and a downloadable checklist.)
2. Put the money in a Community Benefit Agreement — before the permit vote. A CBA is a legally binding contract, separate from zoning, in which the developer commits cash and concessions in exchange for community support. Leverage is highest before the vote and near-zero after. New York's EO 62 set a public benchmark: ~$1 million per megawatt as a starting point — a ~900 MW campus like Port Washington's would open negotiations near $900 million in community benefits. Model your own on the 🛡️ Negotiation Toolkit tab.
3. Replace the tax abatement with a host fee or PILOT. The standard deal — a 10–20 year property-tax abatement — starves exactly the schools and roads that absorb the impact. Counter with a Payment In Lieu Of Taxes or an annual host fee (per MW or per acre) that escalates with inflation and starts on day one. A megawatt of always-on load can support a recurring payment, not just a one-time check.
4. Set up a data dividend — the Alaska model. Alaska pays every resident an annual dividend from oil revenue. A community can structure the same thing: route a slice of the host fee or a per-MWh levy into a permanent local fund that pays residents or underwrites property-tax relief for the life of the facility. That's the difference between a few farmers getting rich once and the whole town getting a raise for 30 years. The Data Dividend calculator on the Negotiation Toolkit tab sizes it.
5. Make the data center pay for its own grid. The ~900 MW is the biggest hidden cost: if its new generation and transmission get socialized, everyone's electric bill rises (see the capacity-charge story below). Demand a large-load tariff that assigns the full cost of new generation and transmission to the facility, plus closed-loop cooling and water-replenishment commitments in writing.
6. Get the transmission-corridor neighbors paid — recurring, not one-time. For the residents facing eminent domain, push for above-market easements structured as annual line-rental payments rather than a single condemnation check, plus routing that avoids homes. The people carrying the wires should share the upside, not just the burden.
7. Bond the exit. Data centers become obsolete. Require a decommissioning bond up front so the town isn't left with a stranded concrete shell and a cleanup bill.
The through-line: value you capture contractually, before the vote, and structured as recurring revenue beats a one-time land check every time. A land sale pays the person who holds the deed; a CBA, a host fee, and a data dividend pay the whole community — including the neighbors who never got an offer.
Every mechanism above depends on knowing what your land and your megawatts are actually worth. Most of that price signal is public — communities just don't know where to look. Here's where it lives:
We wired the first two into the 🛡️ Negotiation Toolkit tab: a Land price-discovery tool now shows the USDA cropland baseline for your state, converts any per-acre offer into a multiple over farmland value, and links straight to your county's deed records and the live USDA and Good Jobs First databases. Diagnose the asymmetry, then close it.
The Port Washington story gets told as a feel-good windfall — farmers made millionaires overnight. That's true, and it's also the smaller half of the story. The full version is about who had the information and who didn't: a developer that knew the campus was worth $15 billion, sellers who priced against soybeans, and neighbors who get the wires but not the check.
Communities can't stop the AI land rush from coming. But they can refuse to negotiate in the dark — which is the one condition under which everybody but the developer loses.
Model the numbers on the 🛡️ Negotiation Toolkit tab, trace the operators and their LLCs on the 🏢 Data Centers tab, and put your objection on the record via the 🗂️ States & Officials tab.