Amazon's $17bn Ohio Exposure Puts a 250 MW Permit to the Test
Amazon can move future capital. Prologis already paid for land and began groundworks. A legal right to build is the part that stays put.
Trenton sold Prologis 141 acres for $7.7 million in October 2025. On 3 November 2026, its voters approved a 25 MW size restriction against a planned 250 MW data centre. Prologis had received a zoning certificate in December 2025. I would refuse to finance the full 250 MW as secure capacity until a court protects that approval. Amazon has options for its future spending; the developer has a particular parcel, an unfinished campus and a very expensive legal question.
Entitlement Durability measures how much planned capacity an existing permission protects after voters change the rules. Trenton provides an inconvenient test: the planned 250 MW project has a December 2025 certificate, while the new threshold is 25 MW. A court protecting that certificate preserves the legal path to the full project. Without it, the four-building design faces a fresh permission problem. When you pay for approved land, you are buying relief from having to win the same argument twice.
Prologis sounded unfazed after the vote: "The result does not affect our project." The developer bought the land from the city, received its certificate before the vote and started groundworks. It had invested $42.3 million by November 2026; its first building could launch in 2028. Fortune cited local reporting that the vote was expected to spare this four-building campus. On that 2028 development horizon, defending an existing approval is more attractive than replacing a site and beginning again. You can make a persuasive case for valuing all 250 MW as protected.
Conserve Ohio asked Trenton to revoke the certificate, and Prologis sued in federal court on 12 November 2026. The planned 250 MW is ten times the new 25 MW threshold. If the earlier certificate governs, the $42.3 million supports the intended capacity and its path toward a 2028 first building. If it fails, the 250 MW design runs into a ban on facilities above 25 MW. The court's eventual public ruling decides which version deserves financing.
Two separate developments near Trenton have become easy to confuse. Project Mila is Prologis's four-building site, with Amazon reportedly its customer. Amazon has separately negotiated over an 18-building, 601-acre proposal in two townships south of the city, carrying a $5bn headline in August 2026. You cannot move Project Mila's December 2025 certificate across a property line to rescue Amazon's other project. The first investment has a right to defend; the second still has permission to obtain.
Amazon's $17bn headline invites a misleading subtraction. On 17 November 2026, the Financial Times identified three affected Amazon projects with approximately 1.8 GW of capacity and $17bn of planned spending, excluding chips and other equipment. Amazon's own Ohio announcement envisages investment surpassing $23bn by 2030. In a scenario requiring new sites or approvals for all three affected projects, $17bn over a plan exceeding $23bn is less than 74% of the Ohio ambition. That is a ceiling on the proportion of planned investment exposed to disruption, rather than a booked loss.
On 4 November 2026, AInvest put the capital-allocation argument neatly: "A $2 billion campus is under 1% of the roughly $220 billion Amazon still expects to spend on capital expenditures in 2026." That budget gives Amazon room to redirect the Sunbury project. Its filing for the quarter ended 30 September 2026 also records approximately $578bn of commitments not yet recognized as revenue, giving infrastructure expansion an economic purpose. Protecting the broader business beats litigating every site, while Prologis seeks its own remedy. Amazon's flexibility is genuine, which makes a corporate solvency scare unconvincing.
Using AInvest's same 2026 spending yardstick, $17bn divided by $220bn is 7.7%, against its less than 1% for Sunbury alone. These multiyear project costs against an annual budget measure scale, rather than cash due in 2026. The Amazon filing supplies the cost of capital intensity: trailing free cash flow was negative $42.357bn at 30 September 2026, versus positive $14.788bn a year earlier, while long-term debt stood at $173.125bn against $65.648bn at 31 December 2025. Management expected cash capital expenditure to "meaningfully increase in 2027". Rerouting capital is possible; doing so while funding existing commitments still consumes financing capacity.
Voters have produced the opposite outcome elsewhere. In Harwood, North Dakota, 325 of 430 votes on 3 November 2026 opposed limits on large industrial developments, while Applied Digital was building a 160-acre campus with Oracle as end user. That was the only developer-friendly result among 45 data centre ballot measures. That 75% vote demonstrates political support for rejecting restrictions. Harwood, however, had no 25 MW ceiling testing an earlier certificate. Trenton's 80% vote for restrictions makes that difference financially consequential for a lender holding its particular property.
In Wilmington, Ohio, even a courtroom win failed to settle the politics. On 18 September 2026, the Ohio Supreme Court barred a resident measure targeting a $4bn AWS development. On 3 November, about 80% of voters rejected four rezoning ordinances. The legal victory and ballot defeat concerned different instruments, and Wilmington sought rezoning rather than defending Trenton's existing certificate. You can win one procedural fight and still need another public permission. A loan against a parcel pays for deliverable capacity, not an impressive collection of court victories.
AEP Ohio's approved tariff requires large new customers to pay for at least 85% of subscribed energy across 12 years, including a four-year ramp, under terms approved on 9 July 2025. Amazon's separate proposal near Trenton is close to a Duke Energy substation, so the AEP rules are no proxy for its power contract. The tariff prices unused electricity; local votes expose the value of permission. Whoever finances a parcel faces the possibility of paying for one input while another is tied up in court.
I would pass on financing Project Mila at a price assuming its full 250 MW is presently secure, and I would resist marking Amazon's $17bn of planned projects as a loss. Entitlement Durability has a single public falsifier: a federal order in Prologis v. Trenton published by the end of November 2027 protecting all 250 MW under the December 2025 certificate without redesign or fresh approval would defeat my financing discount. Until then, Prologis's $42.3 million is exposed to a legal question Amazon's ability to redeploy spending will never answer. I would rather miss some upside than pay today for 250 MW whose legal protection is still being contested.