Costs: What do we really get?

The big promise

Burkhalter is pitched as up to $1.2 billion in investment by the developer's fiscal impact study, with only about 12 projected permanent jobs. (LOCI study, May 2026 ↗)

A fiscal impact study built with a model called LOCI® says the project would bring about $76.6 million to the City, the County and the schools over 10 years (study ↗). Nearly all of that is property tax. Sales tax and business fees barely show up. The study counts almost nothing for franchise fees: about $1,900 for the City over 10 years. The City Manager's slides from the September 1, 2026 council meeting say franchise fees would add about $1 million a year (slide deck ↗). When we asked how the City got that number, it said it has no records showing where it came from.

No customer picked. The numbers are exact anyway.

The developer says no end user, meaning the company that would actually run the data center, has been picked. The City says the permit application is incomplete until one is named. But the study the City is leaning on needs exact numbers for that customer's equipment. Here is what each document in our files says:

Nobody has named the customer, yet the equipment, the cost and the jobs are all spelled out to the dollar.

So which is it? If the developer knows enough about the customer to say it will buy $325 million of equipment and replace it every five years, it knows something about who the customer is. If it doesn't know, those numbers are guesses, and a guess can't be the basis for what the City, the County and the schools are told to expect. The City can't treat the customer as unknown when it comes to approving the permit, and as known enough when it comes to believing the money. Council should wait for a named operator and an independent fiscal study before any vote.

What the study counts, in plain words

A fiscal impact study adds up two things: the money a project brings in, and what it costs local government to serve the people who come with it.

Money in: the study says about $76.6 million over 10 years after costs. Nearly all of it is property tax on the building and the computer equipment.

Costs: the study counts people, not buildings. The company says 12 people would work there. The study turns that into new households and new students, then charges the City, County and schools an average cost for each:

That is why the money looks huge and the cost looks tiny. The study isn't saying the project costs the community almost nothing. It is saying its method only counts people. It has no line for what a 99 megawatt building adds on its own: fire and emergency response for a site with batteries, generators and fuel, road wear from construction and trucks, or the power lines and substations. Water and sewer are left out on purpose, and so are the electricity sales tax and the franchise fees tied to the project.

New schools? The study assumes three new students, so it assumes no new classroom costs. We can't prove that either way. The school board and county planners can say.

Would anyone's property tax go down? The study assumes tax rates stay the same ("the millage rate is held constant"). So the $76.6 million is extra money for government, not a promised cut to anyone's bill. A cut would be a separate decision by the commissioners and the council, and the study gives no number. We will ask the County Tax Assessor for an estimate.

Whose taxes, where do they go, and who checks them?

• Whose taxes? The owner of the building ($524.9 million) and of the computer equipment ($325 million). The study doesn't say who that is. The developer builds, but whoever moves in owns the servers, and no one has been named.
• What do they cover? The study doesn't say. It compares the money to the average cost of serving a few new residents. There is no spending plan and no promise that the money goes to anything in particular.
• Where is it used? Not stated anywhere in the records we have.
• How can it be checked? The study's own words: its numbers come "per the information provided by the client." You can compare them to the County Tax Assessor's value of the building and equipment, the tax rates, how fast the equipment loses value, and whether the Authority offers any tax break. We searched and could not find any case where a LOCI projection was compared with what was later collected. The numbers rest on a promise that can only be checked years from now.

Why the equipment tax fades

Most of the property tax comes from the computer equipment, not the building. The study counts $325 million of servers and chips, and it assumes the company buys another $325 million every five years.

Property tax follows what the equipment is worth, and computer equipment loses value fast. A study from Utah's Gardner Policy Institute (read it ↗) followed $1 billion of chips and servers:

The tax bill drops the same way. The study here keeps the money coming by assuming new equipment arrives on schedule: every five years for Burkhalter, every four years in the Candler study. Nothing in the records requires the company to buy anything. The Utah study says replacement schedules affect how much tax comes in, and that early relief for property owners "may be followed by future tax increases."

What if demand slows, or the operator runs the machines longer, or moves them to another site? Then the equipment tax falls, and nothing in the study says who covers the gap. That is why we are asking the City whether it will require a clawback if the promised revenue doesn't arrive.

Who paid for the study?

The study says it was done "at the request of the Development Authority of Bulloch County," about a "hypothetical data center." It never names the developer. But the Authority's own records, which we got through an open records request on October 5, 2026 (available on request), tell a different story:

• April 9, 2026: the Authority told the economist, "the developer has agreed to cover the costs of the work."
• April 13: the economist asked to bill the Authority instead of the developer: "That way, I am technically working for you and not the developer. Numbers from a developer's 'hired gun' sometimes aren't treated quite the same as those done for the development authority."
• May 13: the Authority told him where the numbers came from: "We received this data from the developer."
• June 25: the Authority paid the economist $11,500.
• July 8: the developer paid the Authority back $11,500, with "LOCI Study" written on the check.

Both images are from the Authority's open records response (October 5, 2026). We blacked out bank account numbers and signatures.

Developer-paid studies are common, and nothing here shows anything illegal. But the public was shown this as the Development Authority's own study about a "hypothetical" project. We found out later that it was about this specific project, with this developer's numbers, paid for with this developer's money. We respectfully ask the Development Authority to get a second opinion, or to explain why Economic Impact Group's method can be trusted when the numbers come from the developer. Why should the City rely on one study paid for by the company that stands to profit, when so many questions about this project are still unanswered?

The numbers moved

The developer's first data sheet said the building alone would cost $855 million, with no computer servers until year five. The economist wrote back: "it is the craziest data center I have ever seen."

After a phone call that same day, the developer sent new numbers: a $524.9 million building, plus $325 million of servers in the first year, replaced every five years. Same project, very different math.

The Authority also wrote: "We're not planning to offer any incentives for this project." That's good to know. Council should get that in writing before any vote.

Same economist, very different math

The same firm, Economic Impact Group, also ran the numbers for a "100MW" data center in Candler County, next door, in February 2026. Burkhalter asked for 99 megawatts. Almost the same size. Look at the difference:

• Burkhalter (99 MW): up to $1.2 billion invested, about 12 projected permanent jobs, about $76.6 million over 10 years for the City, County and schools.
• Candler (100 MW) (study ↗): $31.6 billion invested over 20 years (with the servers replaced every four years), 40 jobs, and "just over $1 billion over 20 years."

Per year, Candler's study promises about 7 times as much money. Why? Both studies run on numbers the customer hands over: "per the information provided by the client." The model only knows what it's told.

A fiscal impact study is only as good as the numbers someone feeds it.

Read the Burkhalter one-page summaries: City, County, Schools. And the Candler study.

The tax break

Georgia gives data centers a break on sales tax for the computers and equipment they buy (Georgia Rule 560-12-2-.117 ↗, Department of Revenue overview ↗). It's not a property tax break.

Georgia's own state auditor checked this tax break. It found the numbers used to sell it to the public were inflated by more than two-thirds, and that 70% of data center projects would have come here anyway, tax break or not (The Current ↗, Good Jobs First ↗). A University of Georgia study called the tax break a net loss for the state overall, even when the local town gets a property tax bump.

In 2025 alone, Georgia gave up about $474 million in state revenue on this tax break, against about $41 million in tax revenue from data centers.

State projections show it getting bigger: about $2.5 billion lost in fiscal year 2026, and nearly $3 billion in 2027. About $1.1 billion of the 2026 loss is local sales tax (Good Jobs First, citing the state's projections ↗).

Will the money last?

What if the company leaves early? PSC Commissioner Peter Hubbard has said even the biggest power contracts have a loophole: "with four years of payments, you can get out of paying the full 15" (Statesboro Herald, April 22, 2026 ↗). Then "someone has to pay for the infrastructure that was built."

A brief filed with the PSC says the same: on a 15-year contract, "more than a decade's worth of bills could go unpaid." (PSC Dockets 56298 and 56310, December 16, 2025.)

Whether this deal covers that is in a private contract the public can't see.

See the four-year loophole and what Council should require.

A study from Utah's Gardner Policy Institute (read it ↗) found data center equipment loses value fast, so early property tax relief can fade and the tax burden can shift back to other taxpayers. In its example, $1 billion of computer chips and servers was worth about $70 million after five years. That's a Utah finding, not something proven in Statesboro. But it's a real risk to weigh before assuming the money is permanent.

What happens when the company can't deliver the city revenue it promised? Will the City ask for a clawback to cover the losses?

Statesboro's ordinance says the builder must "demonstrate positive economic impacts to the City of Statesboro" (Ordinance 2026-01 ↗). Right now, the only proof is a study the developer paid for, using the developer's numbers.