The Small Town That Banned a Data Center – and Got Sued for $300 Million

The Fight That Looks Decisive Isn't

Read enough of these stories in a row and a pattern emerges that the headlines miss: the town rarely wins. Hill County folded in a week. Monroe Township's own council reportedly acknowledged in a June statement that an outright ban "would lead to litigation" – they passed it anyway, and got exactly that. That is not the end of it: in late July Monroe moved to have the suit dismissed, its attorney arguing that a zoning ordinance is not a speech restriction and that the developer never held a binding contract. The motion is unresolved, so the one case here with a live docket has not actually gone the developer's way yet. Local opposition is real, it is expensive, and it occasionally kills a specific site. But as a brake on the industry, it looks more like a speed bump that developers are pricing in and paying through than a wall anyone expects to hold.

So if the political fight isn't what actually decides which data centers get built first, what does? This desk has already argued that the honest answer starts with the grid: the interconnection queue that ERCOT and other operators are sitting on is mostly a filing list, not a construction schedule, and utilities themselves plan for far less demand than the announced gigawatts imply.

That reading got a fresh confirmation this month – Wood Mackenzie now projects that US utilities will likely commit to only about 28% of the 1,066 gigawatts requested for data centers nationally, and Texas Governor Greg Abbott ordered a full audit of every data center seeking grid access on August 3, pausing approvals and putting $13 billion in industry revenues at risk, by BloombergNEF's estimate. The known workaround for that queue is to skip the grid and generate power on site.

What hasn't been widely reported is that the workaround has itself split into a race between two different technologies – and the fight between them is being decided by a second permitting queue almost nobody outside the power industry is watching.

Why One Kind of Generator Skips the Line

A gas turbine generates electricity by burning fuel, and burning fuel makes nitrogen oxide. Above a fairly low threshold, that triggers a Clean Air Act review that can take years to clear – the same kind of multi-year review that already makes new grid interconnections such a bottleneck. A fuel cell generates electricity electrochemically, with no combustion at all, and in many configurations it falls under the threshold that would classify a site as a regulated "major source".

That single difference is why a fuel cell can often be permitted and installed on a timeline a combustion turbine cannot match, and it is the actual mechanism behind a sentence that otherwise sounds like a stray earnings-call remark. On Bloom Energy's Q2 2026 call, CEO KR Sridhar said customers who had already ordered combustion turbines and reciprocating engines were canceling those orders and switching to Bloom instead – and named Nebius as one of them.

That is a company changing its equipment order after the fact, for a reason that has nothing to do with a town council. It also lines up with an unglamorous but important fact: turbines themselves are now capacity-constrained. BloombergNEF puts the cost to build a new combined-cycle gas turbine plant at $2,157 per kilowatt in 2025, up 66% from under $1,500 in 2023, with manufacturer backlogs stretching toward the early 2030s.

A buyer facing a multi-year wait for the turbine itself, on top of a possible multi-year air-permit review once it arrives, has a real reason to look elsewhere – even before comparing the electricity bill.

The Named Deal Behind the Number

The concrete version of that switch became public the week of August 13, 2026, when Nebius confirmed it had chosen Bloom Energy's solid-oxide fuel cells to power a 300-megawatt AI data center in Vineland, New Jersey, on site and behind the meter. Nebius described the choice as delivering power "quietly and ultra-low emissions," with "no significant impact expected on the project timeline". Bloom Energy's stock rose about 13% on the news, to roughly $238, and FuelCell Energy – a smaller rival chasing the same substitution – rose about 11%.

Bloom's underlying business is growing fast enough to support that reaction on its own: Q2 2026 revenue hit $1,065.4 million, up 165.5% from $401.2 million a year earlier, and the company raised full-year guidance to $3.9-4.2 billion from the $3.4-3.8 billion it had set in April.

None of that means the electricity itself got cheaper. Bloom has published exactly one per-kWh figure in its corporate history: $0.099, posted in July 2023 with the Series 10 managed service – five-year term, maintenance and 24/7 monitoring included. US industrial customers actually paid 8.71 cents per kWh in May 2026. One side of that comparison is three years stale, and it is the only side Bloom's own disclosures offer. Lined up against comparable published rates, the actual trade looks less like a discount and more like a bounded toll:

Bloom's own number sits roughly 14% above what industrial customers actually paid this May, not below it. What buyers are purchasing with that premium is a technology that clears a permitting review the alternative cannot, on a timeline a compute buildout running an 18-month model cadence can actually use.

That is the trade this story is actually about, and it is a narrower, more specific claim than "fuel cells are cheaper" – a claim several industry blogs made this year with cost figures that could not be traced to any named source and that contradicted each other across the same sites' own other posts. Those figures are not used here.