The data-center giveaway. Billions out, a trickle back.
Georgia's own auditors found the state gave up $474 million in a single year to exempt data centers from sales tax — and, by the state's own model, 70% of that construction would have happened without the break. At least ten states run programs like this. The biggest deals cost nearly $2 million in public money per permanent job.
This is the austerity double standard in one industry. The money that cannot be found for healthcare or child care is found, without a fight, for warehouses full of servers — and the public ledger, where states bother to keep one, comes back deep in the red.
What this page is about
In December 2025, the University of Georgia's Carl Vinson Institute of Government delivered a legally-mandated evaluation of the state's data-center sales-tax exemption to the Georgia Department of Audits and Accounts. It found the exemption cost the state $474.2 million in forgone revenue in fiscal 2025, against roughly $41 million in direct state tax revenue from the industry — and, crucially, that only about 30% of the construction was actually caused by the break. By the state's own model, 70% would have happened anyway.
Georgia is not an outlier; it is the state that bothered to measure. Good Jobs First, which has tracked these subsidies for over a decade, finds at least ten states now forgo more than $100 million a year each, three forgo more than $1 billion, fourteen disclose nothing at all, and the eleven largest deals averaged $1.95 million in public subsidy per permanent job. This page grades the ledger — what the public gives up, what it gets back — and keeps the industry's economic-impact claims in plain view.
The anchor is a state document: the Carl Vinson Institute evaluation prepared for the Georgia Department of Audits and Accounts under the Tax Expenditures Transparency Act of 2024. The multi-state figures are from Good Jobs First's subsidy research and from a 2017 Washington State JLARC audit. Where a number is a watchdog estimate rather than a state audit, it is labeled as such.
We are not claiming data centers produce nothing. The same Georgia report credits the industry with about $1.0 billion in value added and 8,505 jobs, and notes that tax incentives are one of several factors in a location decision. The claim here is about the ratio — dollars forgone versus dollars and permanent jobs returned — not that the activity is zero.
We are not asserting the industry's own economic-impact totals are wrong. We are placing them next to the forgone-revenue and but-for figures the same evaluations contain, which is where the giveaway becomes visible.
We are not treating watchdog estimates as audits. Georgia and Washington are state-auditor findings; the Virginia and national figures are Good Jobs First estimates, and we say so at each.
Books that go deeper on this story. Links are Amazon affiliate searches — buying through them supports the work at no cost to you.
The same investigation, restaged one beat at a time. Step through it here, or present it fullscreen.
The data-center giveaway. Billions out, a trickle back.
Georgia's own auditors found $474M forgone in one year - and, by the state's own model, 70% of the construction would have happened without the break. At least ten states run programs like this.
The break mostly pays for construction that was coming anyway — and the states can prove it on their own paper.
The pitch for a data-center tax exemption is always the same: cut the sales tax on servers and construction, and the industry will come, bringing jobs and investment it would not have brought otherwise. The load-bearing word is “otherwise.” If the data center was going to be built regardless, the exemption is not an investment — it is a discount on a purchase already being made.
Georgia measured exactly that. Its evaluator built an econometric model to estimate the “but for” share — how much activity the exemption actually caused — and put it at 30%. The other 70% would have happened without the break. The state is paying full price to claim credit for construction it was getting for free.
Stack that against what comes back. Nearly half a billion dollars forgone in one year in Georgia; more than a billion in Virginia; in Washington, a state audit found data centers paying $22 million in property tax while the state lost $57 million in sales tax. And the jobs are few: the biggest deals nationally run close to $2 million in public money per permanent position. That is the thesis — not that data centers are worthless, but that the subsidy is a poor trade the states' own numbers expose.
2017 to 2026: from the first audit warnings to Georgia’s reckoning.
- 2017A Washington State JLARC audit finds data centers paying about $22 million in property tax while the state loses roughly $57 million in sales tax to the exemption — an early, clean audit of the trade.
- 2018Georgia enacts its data-center sales-and-use tax exemption (HB 696, O.C.G.A. § 48-8-3(68.1)), exempting servers and construction materials for facilities that clear investment and job thresholds.
- 2024Good Jobs First estimates Virginia's exemption alone cost more than $1 billion in forgone revenue in fiscal 2024, the largest such program in the country.
- Dec 2025Georgia's Carl Vinson Institute evaluation lands: $474.2M forgone in FY2025, and a revised “but for” estimate of 30% — 70% of construction would have happened anyway (down from the 2022 study's 90%-attributable assumption).
- 2026Good Jobs First reports at least ten states forgo more than $100M a year, three exceed $1B, and fourteen disclose nothing; the eleven largest deals averaged $1.95M in subsidy per permanent job.
Who measured this — mostly the states’ own auditors.
Prepared the December 2025 statutory evaluation of Georgia's exemption under the Tax Expenditures Transparency Act of 2024. Its econometric “but for” model produced the 30%-caused / 70%-anyway finding and the $474.2M forgone figure. This is the state's own analysis, not an advocacy group's.
Has tracked data-center subsidies for more than a decade. Source of the national picture — the count of states over $100M and $1B, the fourteen that disclose nothing, and the $1.95M-per-job figure for the largest deals. Its numbers are estimates built from disclosure filings, labeled as such here.
Washington's nonpartisan audit arm, whose 2017 review put the data-center trade in stark terms: $22 million in property tax paid against $57 million in sales tax forgone. An early state-auditor confirmation of the pattern Georgia would later measure.
The beneficiaries. The largest, most profitable technology companies on earth are the primary recipients of these state sales- tax exemptions — firms for which a state's tax break is a rounding error, capturing public money a state says it cannot spare elsewhere.
Seven claims about the ledger, each graded, each traceable.
Georgia forgave $474.2 million in one fiscal year
FACTThe Carl Vinson Institute evaluation for the Georgia Department of Audits and Accounts estimated forgone state revenue from the data-center sales-and-use tax exemption at $474.2 million in fiscal 2025. Against that, direct state tax revenue tied to the industry was on the order of $41 million. This is the state's own statutory evaluation, not an outside estimate.
“The largest component of the total fiscal impact is forgone tax revenue resulting from the direct cost of the exemption.”
By Georgia's own model, 70% of the construction would have happened anyway
FACTThe evaluator built an econometric 'but for' model and estimated that only about 30% of data-center construction activity was actually caused by the exemption - meaning roughly 70% would have occurred without it. That is a sharp revision from the 2022 study, which had assumed 90% of the activity was attributable to the break. The exemption is largely paying for investment the state was getting regardless.
“A 'but for' estimate of 30%, suggesting that, in the absence of the exemption, 70% of data center construction activity in the state would have occurred anyway.”
The same report credits ~$1.0 billion in value added and 8,505 jobs
FACTKept here in the open, because honesty demands it: the Georgia evaluation also credits the exemption's activity with about $1.0 billion in value added to the state economy and roughly 8,505 jobs. This is the industry-favorable side of the same ledger. Placed next to $474.2M forgone and a 70%-anyway baseline, it is what the public bought - and most of it, by the state's own model, it would have had without paying.
Virginia's exemption cost more than $1 billion in a single year
FACTGood Jobs First estimates Virginia's data-center sales-and-use tax exemption cost the state more than $1 billion in forgone revenue in fiscal 2024 - the largest such program in the country, centered on the world's densest data-center cluster in Loudoun County. Graded as a watchdog estimate built from state disclosure, not a state audit.
A Washington audit: $22 million paid in, $57 million lost
FACTWashington State's nonpartisan JLARC found in 2017 that data centers using the exemption paid about $22 million in property taxes while the state forwent roughly $57 million in sales tax - a better than two-to-one loss on the trade, from a state auditor rather than an advocacy group.
At least ten states forgo over $100M a year; three exceed $1B; fourteen disclose nothing
FACTGood Jobs First's national survey finds at least ten states now forgo more than $100 million a year each in data-center sales-tax subsidies, at least three exceed $1 billion, and fourteen states do not disclose the cost at all. The disclosure gap is itself the point: most of this spending is invisible by design.
The biggest deals cost about $1.95 million per permanent job
FACTAcross the eleven largest data-center subsidy deals Good Jobs First profiled, the public paid an average of $1.95 million per permanent job. Data centers are capital-intensive and employ few people once built; the subsidy-per-job math is the reason the ledger comes out where it does.
The strongest version of the industry’s case, and why the ledger still loses.
The industry's case is not nothing, and the Georgia report makes it fairly. Data centers do generate construction activity, some permanent jobs, property-tax base, and, in Georgia's accounting, about $1.0 billion in value added. The report also notes, correctly, that tax incentives are only one factor in a siting decision, alongside power, land, water, and talent, and that pulling an incentive can signal an unwelcoming business climate. A serious critic has to hold all of that.
But the same document is what sinks the deal. Its own “but for” model says 70% of the activity would have come without the break — so most of that $1.0 billion is not bought by the exemption, it is merely claimed by it. And the incentive is a state sales-tax exemption specifically, the costliest and least visible of the subsidies, flowing overwhelmingly to the richest firms in the world. When you subtract the activity that was coming anyway, the public is paying a large, real sum for a small, incremental slice.
The honest uncertainty is in the estimates. A “but for” percentage is a model, not a measurement, and reasonable economists argue over it — Georgia's own figure swung from 10% to 70% between two studies as better data arrived. The national totals from Good Jobs First are disclosure-based estimates, not audits. We grade the state-auditor findings (Georgia, Washington) as fact and label the watchdog estimates as estimates. What is not in doubt is the direction: every serious measurement, including the states' own, lands in the same place.
The austerity double standard, in one line item.
This hub exists to ask what the public gets for its money. The data-center giveaway is the cleanest case on the board, because the states did the accounting themselves. Half a billion dollars a year in Georgia, a billion in Virginia, a two-to-one loss in Washington — and the AI build-out is accelerating, which means the exemptions, uncapped and largely undisclosed, are set to grow. Good Jobs First projects Georgia's annual cost climbing toward the billions.
Set that beside the companion pieces in this cluster. When a universal program is proposed, the reflex is “we can't afford it” — the subject of The Austerity Myth, which documents programs that pay for themselves being killed on cost. Here is the same treasury, in the same years, finding hundreds of millions for server warehouses it did not need to subsidize to get. The money was never the constraint. The choice of who gets it was. That is what makes this a return-on-investment story: not that the spending exists, but that it is spending the public would never have approved if the ledger were kept in the open.
Questions worth taking seriously
Don't data centers create jobs and investment?
Is the '70% would have happened anyway' figure reliable?
Aren't these numbers from advocacy groups?
Why is this a 'return on investment' story?
If you are named on this page
If you are named on this page, or are a party materially affected by the claims made here, and you wish to respond, correct the record, or add context, use the Contact page. Responses are published verbatim alongside the original claim, with the sender identified and the date of receipt. The channel stays open for the life of the page.
This site aggregates and grades a record that other outlets and primary sources have already put on the record. Every FACT-graded claim above is sourced to court filings, government reports, sworn whistleblower disclosures, published investigative journalism, or named-source statements. The citations are the accountability mechanism; this section is how you get on the record too.
The state evaluations and the subsidy tracking behind the ledger.
Every claim on this page grades to one of FACT · PROBABLY TRUE · SOME SMOKE · PURE SPECULATION · FALSE / MISLEADING. State- auditor findings are graded fact; watchdog estimates are labeled as estimates.
- Carl Vinson Institute of Government, “Tax Incentive Evaluation: Georgia Data Center Sales & Use Tax Exemption” (Dec 2025) — for the Georgia Dept. of Audits and Accounts. The $474.2M and 70%-but-for anchors.
- Good Jobs First, “Data Center Subsidies Surge as States Lose Billions” and related reports — the national picture and the Washington JLARC figure.
- Stateline, “Many states don't report losses from data center tax breaks” (Apr 2026) — the fourteen-state disclosure gap.
- WUSA9 — Virginia's exemption cost >$1B in FY2024 (Good Jobs First estimate).
- America's Plan / Good Jobs First — $1.95M in public subsidy per permanent job across the eleven largest deals.
Full method: Methodology. Related hubs: Return on Investment, The Austerity Myth, and The Corporate State.
Published August 2026. State-auditor findings are graded fact; watchdog totals are labeled estimates. If a link 404s or a figure is wrong, tell us and we will fix it publicly.