Data Center Moratoriums 2026: Which Counties and Towns Passed Them
NationGraph / Blog / Land Use and Permitting
Every jurisdiction with a data center moratorium or equivalent instrument on the record, with dates, durations, stated reasons and a link to the minute or the video timestamp.
Published Sep 14, 2026
Quick Answer
NationGraph's census of local government records identifies at least 41 US counties, cities, townships and special districts with a data center moratorium or an equivalent land use instrument on the record, 10 of which adopted one in the last 18 months. Knox County, Tennessee approved resolution R-26-6-904 on June 22, 2026. Ypsilanti, Michigan adopted a 365 day moratorium on March 3, 2026. The reason given most often is not water and not power. It is that the local code has no category for a data center, which leaves one permitted as a generic industrial use, so a temporary pause is the only instrument available until the code is rewritten.
This article is built on a NationGraph census of 88 records from 84 distinct county, city, township and special district bodies across 33 states, captured between July 2024 and August 2026. Every record carries a working source link: 49 come from meeting video with second accurate timestamps, 26 from board minutes, 12 from agenda packets and one from a public notice. Almost none of it appears in a bid feed, because a moratorium is not a procurement. It is a governing body changing what it is allowed to approve, which is the same class of upstream event that surfaces a purchase long before the RFP.
Which counties and towns passed data center moratoriums?
Ten in the last 18 months, three more before that window, across 10 states. The instruments are not identical and the differences matter more than the count.
Clarke County, Iowa has the cleanest resolution text in the file. Its June 2025 resolution establishes a temporary moratorium "on the construction of data centers in the unincorporated areas of Clarke County," and states the purpose in the same passage: to write appropriate regulation. Scope, instrument and reason in one sentence, with nothing inferred.
Comstock Charter Township, Michigan shows why the word moratorium hides real variation. Its resolution pauses the processing of applications, not construction. That is a narrower legal instrument than Clarke County's and it does different work. Fluvanna County, Virginia went narrower still and paused only new applications, with a defined end month. Kootenai County, Idaho used an emergency moratorium, the fastest instrument in the census.
Stanley, North Dakota is the only jurisdiction with a complete legislative trail on the record: two readings in November and December 2024, then adoption on January 14, 2025. Marshall County, Indiana is the only one with a vote tally, "Motion carried 3-0." Most records show that a moratorium exists without showing how close the vote was.
The best available count of how far this has spread does not come from us. It comes from a council reading its own resolution into the record. Ypsilanti stated on March 3, 2026 that "at least 19 Michigan communities um have passed data center moratoriums finding these facilities require specific regulatory controls not present." That is Ypsilanti's figure, phrased as a minimum, and this census does not independently prove it. Our own records confirm four Michigan jurisdictions with an instrument, plus Ann Arbor studying and Auburn Hills ruling itself out. Treat 19 as a credible number attributed to Ypsilanti, not as a finding.
The working title for this piece said counties. The data says otherwise. Of the 84 entities in the census, 46 are municipalities, 33 are counties, 6 are townships and 3 are special districts, including a water district and a hospital district. Counties are not even the largest group.
Why do local governments pass data center moratoriums?
Because the code has no provision for a data center. That reason appears in 11 records, more than water, power or taxes.
Lee County, North Carolina put it in one line in February 2026: "Currently, our Udo has no provision for data centers. A moratorium would give time for an in-depth study." Putnam County, Florida supplied the legal consequence in August 2026. Absent a regulation saying otherwise, "a data center is a generic industrial use, a use by right."
A use by right. If a county's code has no data center category, a data center is just an industrial building, and the county cannot say no to one. Denial is not on the menu. A moratorium is not a rejection of the industry. It is a local government buying time to acquire the power to decide at all. Acme Township, Michigan recorded the mechanics of that purchase in its planning commission packet: a moratorium "may require both a zoning ordinance and police-power," meaning two instruments, not one.
Water is second at 10 records and it is real. Leavenworth County, Kansas heard that data centers "use huge amounts of water. Water we do not have to spare." Electricity and grid capacity come third at 6, tax abatements fourth at 6, noise once, neighborhood effects once. Only 35 of the 88 records state a reason at all. In the rest, the instrument is recorded and the motivation is not.
How long does a data center moratorium last?
Anywhere from six months to three years, and the record almost never explains the choice.
Two things are worth noticing. First, the durations are not converted to a common unit anywhere in this dataset, and they should not be. Ypsilanti wrote 365 days. Red Willow County wrote 12 months. Those are the same length of time and two different drafting decisions, and flattening them loses the distinction.
Second, the range is argued inside single meetings. The Leavenworth County session that produced the two year proposal also recorded a call for a three year one. Lackawanna County, Pennsylvania records the process failing outright: a letter dated February 17 "calling for a three-year moratorium on data center development. I have not gotten a response."
There is no record in this census of any of the ten adopted moratoriums expiring, being extended, or being followed by the ordinance it was meant to buy time for. The file captures decisions. It does not capture consequences.
What size data center triggers a local rule?
Between 10,000 square feet and 25 megawatts, depending on who is drafting. The national average build is roughly twenty times larger than the largest of those triggers.
Palo Alto's utilities commission put the scale benchmark on the record in February 2026: the "average size of a new data center in the US is 500 megawatts." Ann Arbor's planning commission put the local reality next to it six months later. The data centers already in the city are "less than 10 megawatts," small enough to sit on ordinary distribution.
That gap is the drafting problem, and it shows up as inconsistency. Conneaut, Ohio wrote a 25 megawatt trigger into a ballot measure. Champaign County, Illinois used floor area instead, at 10,000 square feet. Wiscasset, Maine went the other way and deleted its thresholds: the square foot area and the "20 megawatt uh capacity has been removed. So, we've got a generic" definition. Yankton County, South Dakota debated the tiers out loud, weighing "25 megawatt versus the 300 plus."
A threshold set at 25 megawatts catches a building an order of magnitude smaller than the thing everyone is arguing about, and a generic definition with no threshold catches the server room in the back of an office park. Both failure modes are in this file.
How much does a data center cost to build?
Roughly $11.3 million per megawatt for a standard shell in 2026, with AI fit out adding as much as $25 million more per megawatt on top.
This is the number that turns an abstract argument about megawatts into a dollar figure, and not one of the 88 local government records in this census contains it. Local bodies are writing thresholds in megawatts and square feet while the published cost benchmarks that price those thresholds sit on real estate advisory sites their staff are not reading.
Two jurisdictions did their own version of the math. Citrus County, Florida worked through "current data center market capex research" and Florida tax valuations for a 75 megawatt facility. Park City, Kansas produced the only record in the census with both sides of the ledger in one line: a capital investment of $18M against "a 10-year projection of property taxes for all taxing entities of $2,918,700." That is the calculation every other jurisdiction in this file is arguing about without performing.
We did not multiply the published benchmark against any local threshold. JLL's figure describes a single tenant, air cooled, 50 megawatt shell. Applying it to a 25 megawatt ordinance trigger or a 300 megawatt campus is a scope error rather than a calculation, and this article does not make it.
Do data centers get tax breaks from local governments?
Yes, and the terms are generous where they are stated: 15 and 20 year abatements, 70% to 100% on equipment, against investments of $750 million and $6 billion.
The largest figure in the census belongs to a hospital district. Reeves County Hospital District in Texas noticed a hearing on an abatement where "the estimate of capital investment of all such improvements subject to the requested tax abatement is $6 billion." The applicant named on the notice is Energy Forge One LLC. Public reporting since identifies Energy Forge One as a Chevron subsidiary building a gas plant near Pecos to power a data center, with Microsoft named as a potential offtaker. None of that appears in the local record. What the hospital district saw was an LLC.
That is the pattern, not the exception. Across 88 records, only three companies are named anywhere: Energy Forge One in Texas, Vantage in Indiana and UR Energy in Iowa. Vantage is a real and identifiable developer, with a 618 acre rezoning in Pittsboro proposing 256, 96 and 48 megawatt buildings. It appears in the Pittsboro record only because an existing industrial employer raised a grid capacity objection. No hyperscaler appears by name in this file at all.
The abatement decisions split inside single states and single months. Michigan City, Indiana passed a data center abatement resolution in September 2025. Merrillville, Indiana heard in February 2026 that Indiana practice is to "grant 100% property tax abatement on equipment and partial exemptions on land buildings," with shifted tax burdens named as the consequence for homeowners. Massillon, Ohio heard a resident ask what a twenty year abatement at 70% actually returns to the city. Wildwood, Missouri is drafting the opposite instrument: a code provision saying the city "will not give tax abatements or chapter 100 to data centers." Auburn Hills, Michigan told its council staff is "not aware of any data centers receiving tax abatements" there.
How much water and power does a data center use?
The honest answer from this census is that nobody publishes the number for the building actually in front of them.
The hardest water figure in the file comes from Parker Water and Sanitation District in Colorado, and the conditional is part of it: a data center "uses 5 million gallons of water a day if it's if it's evaporation." Strip the cooling condition and the number means something else entirely.
Independence, Missouri is the only jurisdiction in the census that measured its own operating data center and published the result. A council member noted that people keep raising excessive water use, and that the city has already produced the numbers, making the facility "the 51st uh large water user here." Fifty first. Not first. That single record is the strongest check on the water argument in the whole file, and any piece that leaves it out is not reporting the data.
The closed loop claim gets examined in exactly two places. St Louis planning commission testimony described loop systems that "actually use less water on site, but end up using more power and more water overall." A Sedgwick County, Kansas commissioner pointed out that commissioning a closed loop system still requires a large one time volume, though the figure itself is cut off in the source.
Everywhere else the number is simply absent. Yukon, Oklahoma: "I've asked before, how much water is our center going to use?" San Angelo, Texas asked the enforcement version of the question, whether a facility that "claims to use only x amount of water" can be capped at it. Matanuska-Susitna Borough in Alaska asked the assembly directly. Whitman, Massachusetts recorded the moment the issue landed on a committee that had not considered it: "with data centers though is their water use. >> Their what? >> Water use."
What do local governments do instead of a moratorium?
Five jurisdictions in this census skipped the pause entirely and wrote a permanent conditional use permit requirement instead. None of them shows up in a moratorium count.
Maumee, Ohio is the clearest version. Its council "moved to change the language and make the data center conditional use" in November 2025, before any application arrived, so the rule was teed up for if or when a developer came looking. That is the mature form of the same instinct a moratorium expresses, and it is invisible to anyone counting pauses.
The variants are worth knowing individually. San Marcos, Texas built a two permit structure, with a separate conditional use permit for the major utility use. Monterey Park, California paired a moratorium with a mandatory environmental impact review. Conneaut, Ohio sent the question to voters. De Forest, Wisconsin took the most constructive route in the file and wrote a "reasonable maximum daily water use to serve as a guardrail," which answers the water objection with a condition rather than a delay. Two weeks later the same village board heard a demand for a two year moratorium anyway.
Have any data center moratoriums been lifted?
Two jurisdictions in this census are moving the other way, and both are in Iowa.
Jackson County lifted its data mining moratorium in October 2024 with the developer in the room. The minutes record the zoning administrator and a representative of UR Energy discussing the lifting together. Story County was recorded in March 2026 moving to rescind its moratorium on data centers and crypto data centers.
That is the whole finding, and its size is the point. Ten adopted, two reversing, five permitting instead, seven handling abatements, all inside the same 18 months. The frame that survives contact with this data is not that local government is banning data centers. It is that local governments discovered their codes had no word for this, and are now split between pausing, permitting, subsidizing and refusing to subsidize, often within the same state and the same quarter.
Where are local governments writing data center rules right now?
Sixty seven records in this census are pre decision. These 16 are the closest to a vote, with a named instrument and a date.
For anyone selling into these bodies, the stage column matters more than the jurisdiction. A county whose attorney has been contacted about moratorium process, as in Nodaway County, Missouri, is earlier and cheaper to reach than one with a packet already drafted. Six jurisdictions in the census name outside help on the record: Nodaway County's county attorney, Colliers on a Lansing, New York zoning rewrite, Oregon's SIP program in Morrow County, Acme Township's dual ordinance legal advice, Pennsylvania American Water in Scranton and Vantage in Pittsboro.
Be clear about who that reaches. The buyers visible in this data are land use counsel, planning and zoning consultants, water and utility engineering, and economic development advisory. It is not an IT buying signal. A jurisdiction writing a data center ordinance is not shopping for software, and treating a moratorium as a technology purchase signal is exactly the proxy metric error that makes a pipeline look larger than it is.
Methodology
Every figure comes from a county, city, township or special district record with a working source link, and zero records lack one: 49 meeting videos with second accurate timestamps, 26 sets of board minutes, 12 agenda packets and one public notice. The census holds 88 records from 84 distinct entities across 33 states, 81 of them within the last 18 months. Both underlying searches returned capped result sets, so 84 is a floor, not a total.
No figure here was computed. Durations, thresholds, abatement terms and dollar amounts are printed as stated. Fourteen records carry a figure cut off mid passage, including Michigan City's cost cap and Sedgwick County's commissioning volume; each was left blank rather than estimated. The build cost table is external benchmark data, labeled as such, and was not multiplied against any local threshold.
Excluded: a district's or city's own IT migration, cloud migration and server rooms. Counts follow which bodies publish searchable minutes and video, which over represents the Midwest and small jurisdictions. This describes our corpus, not the market.
Related reading
- How to Track Government Contracts Before the RFP in 2026
- 8 Key Data Sources for Pre-RFP Government Signals
- Why EV Charger Deals Never Show Up in Purchase Orders
- Why Dangerous Intersections Still Don't Get Traffic Lights
- How to Monitor School Board and City Council Spending
- America's AI Action Plan Explained: Your SLED Vendor Playbook
Frequently Asked Questions
Which counties have passed data center moratoriums?
At least 41 US jurisdictions have a data center moratorium or equivalent land use instrument on the record, though most are not counties. Ten adopted one in the last 18 months, including Knox County TN, Greenback TN, Ypsilanti MI, Comstock Charter Township MI, Barberton OH, Fluvanna County VA, Red Willow County NE, Clarke County IA and Pennington County SD. Of the 84 entities in the census, 46 are municipalities and 33 are counties.
Why do local governments pass data center moratoriums?
Most often because the local code has no provision for data centers, which appears in 11 records, ahead of water at 10 and electricity at 6. Putnam County FL stated the legal consequence: without a regulation saying otherwise, "a data center is a generic industrial use, a use by right." A jurisdiction with no data center category in its code cannot deny one, so a pause is the only available instrument.
How long is a data center moratorium?
Durations stated on the record range from six months to three years. Ypsilanti MI adopted 365 days, Red Willow County NE and Champaign County IL used 12 months, Palmetto GA six months, Leavenworth County KS and De Forest WI two years, and Lackawanna County PA and Yukon OK referenced three years. Several adopted moratoriums state no term at all and are recorded simply as temporary.
Can a city stop a data center from being built?
Usually not, without changing its code first. If the zoning ordinance has no data center category, the facility is treated as a generic industrial use permitted by right, and the jurisdiction has no basis to deny a compliant application. That is why a moratorium, which pauses applications or construction while the code is rewritten, is the instrument that appears most often.
How much water does a data center use?
Parker Water and Sanitation District in Colorado stated 5 million gallons a day, conditional on evaporative cooling. The only measured local result in this census points the other way: Independence MO produced its own numbers and found its data center was "the 51st uh large water user here." No record in the census carries a published water figure for a specific proposed facility.
How much electricity does a data center use?
Palo Alto CA's utilities commission stated that the average size of a new US data center is 500 megawatts. Local reality is smaller: Ann Arbor MI reported that the data centers already in the city are "less than 10 megawatts." Columbia County GA described 21 generators on a 48 megawatt facility and a practical ceiling of about 200 megawatts per campus.
Find your first opportunity in minutes
