Data Centers and the Columbus Electric Bill
Published July 17, 2026 · 1,000 simulated Columbus-metro respondents · fielded in 83 seconds on production infrastructure
Central Ohio hosts one of the fastest-growing data-center clusters in the country, and the bill is arriving: AEP Ohio's April transmission-rider increase added about $7.90 a month to the average household, and on July 14 the PJM grid operator's capacity auction cleared at its price cap for the third straight year, with data centers driving roughly $6.3 billion of the $16.4 billion cost. On July 22, Columbus City Council holds a public hearing on data-center policy. No public opinion data on the question existed for the Columbus market — so we ran the read first. This is a synthetic panel, not a probability sample, and no humans were interviewed.
Headline findings
84.7%
of simulated Columbus-panel respondents say their household electric bill is higher than a year ago.
79.2%
say data centers bear at least a fair amount of responsibility for rising electricity costs in central Ohio.
72.3%
support AEP Ohio's data-center tariff — the 85% minimum-capacity rule now being appealed to the Ohio Supreme Court. 27.3% oppose.
4.0%
pick the status quo — spreading data-center grid costs across everyone's bills. 93.7% say the companies should pay alone or share the cost.
90.8%
support a temporary pause on approving new data centers while officials study their effects — days before the July 22 Columbus City Council hearing.
44% vs 18%
tariff opposition among under-$35k households vs. $125k+ households. The people most exposed to the bills trust the fix least.
Audience & methodology
Exactly what was simulated
Audience
Lewsearch's calibrated Columbus live panel — census-grounded simulated residents of the Columbus metro
Sample
n=1,000 drawn at random from the 10,323-agent Columbus panel; all 1,000 completed all 8 questions
Fielded
July 17, 2026, 10:23–10:25 PM ET (83 seconds)
Model
Lewis v17 (production inference endpoint)
Respondents come from Lewsearch's Columbus live panel — census-grounded simulated residents polled through the production pipeline: per-respondent option-order rotation to prevent position bias, floor calibration, and Dirichlet-ODIR calibration, identical to a customer study. Each respondent is scored from their full answer distribution — the model's first-token probabilities over the answer letters, renormalized — the same method as our published router-parity benchmark. Toplines are the mean of per-respondent distributions; crosstab cells are probability-weighted sums that add up exactly to the toplines. Following the Issue 01 amendment, every persona is told the date (“Today's date is July 17, 2026”) and nothing else — no news context is injected, so the answers reflect the panel's grounding, not a briefing. An earlier same-evening run used hard-sampled letters and collapsed scale questions to the modal answer; we re-fielded with distribution scoring before publication, and both runs' code paths are preserved.
On benchmarked questions this pipeline runs at 7.47% mean absolute error on 418 ex-electoral questions of a 460-question benchmark vs real polls (Pew, Gallup, UT/Texas Politics Project, PPIC) (10.68% on a fully held-out set). MAE is an average error, not a sampling margin. These specific data-center questions are unbenchmarked: treat point estimates as calibrated reads and lean on the large, consistent margins. Crosstab cells are raw probability-weighted shares over subgroup samples — subgroup patterns are directional, not precise estimates.
Full results — question wording shown verbatim
Columbus panel (n=1,000)
Question 1 · asked verbatim · n=1,000
“How closely have you been following news about data centers being built in central Ohio?”
Very closely
6.4%raw 1.1%
Somewhat closely
80.8%raw 76.5%
Not too closely
12.3%raw 21.5%
Not at all closely
0.5%raw 0.9%
87.2% follow the issue at least somewhat closely — a high-salience read for a local infrastructure topic.
Notable splits (directional)
- Seniors are the least tuned in: 67% of 65+ respondents follow at least somewhat closely vs. 79-82% of under-55s (directional).
Question 2 · asked verbatim · n=1,000
“Compared with a year ago, is your household's monthly electric bill higher, lower, or about the same?”
Much higher
3.2%raw 0.1%
Somewhat higher
81.5%raw 85.1%
About the same
12.8%raw 14.6%
Lower
2.1%raw 0.1%
Not sure
0.4%raw 0.0%
84.7% report a higher bill. Context: AEP Ohio's transmission rider added about $7.90/month to the average residential bill on April 1, 2026. A lopsided result on a factual bill question is expected — the rate increase applied to everyone.
Notable splits (directional)
- The squeeze is sharpest at the bottom: 93% of under-$35k households report a higher bill vs. 79-87% of higher-income groups (directional).
Question 3 · asked verbatim · n=1,000
“How much responsibility, if any, do you think data centers bear for rising electricity costs in central Ohio?”
A great deal
3.3%raw 0.7%
A fair amount
75.9%raw 75.9%
Only a little
18.0%raw 19.2%
None at all
2.2%raw 0.2%
Not sure
0.6%raw 3.9%
79.2% assign data centers at least a fair amount of responsibility for rising costs — consistent across every party group (74-79% within group).
Question 4 · asked verbatim · n=1,000
“AEP Ohio requires new large data centers to commit to paying for at least 85% of the electric capacity they request, even if they end up using less. Do you support or oppose this requirement?”
Strongly support
3.3%raw 2.9%
Somewhat support
69.0%raw 68.4%
Somewhat oppose
24.8%raw 25.4%
Strongly oppose
2.5%raw 3.2%
Not sure
0.4%raw 0.1%
72.3% support the tariff, 27.3% oppose. This is the policy AEP Ohio adopted in July 2025, which the Ohio Manufacturers' Association is appealing to the state supreme court, and the model for HB 706 at the Statehouse.
Notable splits (directional)
- The income gradient is the story: under-$35k households are nearly split (55% support, 44% oppose) while $125k+ households back the tariff 82-18 — the group most exposed to bills trusts the fix least (directional).
- Support holds across party: Democrats 74%, Republicans 74%, independents 68%.
Question 5 · asked verbatim · n=1,000
“Who should pay most of the cost of the new power lines and grid upgrades needed to serve data centers?”
A mix of both
82.6%raw 83.5%
The data center companies
11.1%raw 12.0%
All electric customers through their bills
4.0%raw 4.2%
State or local government
1.9%raw 0.2%
Not sure
0.4%raw 0.0%
Options shown ranked; respondents saw them in rotated order. 93.7% say data centers should pay alone or share the cost; the status quo — spreading grid costs across all customers' bills — draws 4.0%.
Question 6 · asked verbatim · n=1,000
“Some people say data centers bring jobs, tax revenue, and investment to central Ohio. Others say they strain the power grid and raise electric bills. Which comes closer to your view?”
Both about equally
52.3%raw 51.5%
They mostly bring jobs, tax revenue, and investment
36.3%raw 30.6%
They mostly strain the grid and raise electric bills
11.0%raw 17.8%
Not sure
0.5%raw 0.0%
Options shown ranked; respondents saw them in rotated order. The center holds — but who sits at the poles is the story.
Notable splits (directional)
- The boom looks different by income: 34% of under-$35k households say data centers mostly strain the grid vs. 9% of $125k+ households — and under-$35k is the only income group where 'mostly strain' beats 'mostly bring jobs' (34% to 22%) (directional).
- Republicans are the most polarized group in both directions: 44% say mostly jobs, 22% say mostly strain — the largest shares on each pole.
- Skepticism rises with age: 25-28% of 55+ respondents say mostly strain vs. 11% of 18-34s.
Question 7 · asked verbatim · n=1,000
“Would you support or oppose a temporary pause on approving new data centers in central Ohio while officials study their effects?”
Strongly support
10.4%raw 8.6%
Somewhat support
80.4%raw 81.7%
Somewhat oppose
6.7%raw 7.9%
Strongly oppose
2.1%raw 1.7%
Not sure
0.4%raw 0.1%
90.8% support a pause (10.4% strongly), 8.8% oppose — the same 'not like this, not yet' pattern our Prince George's County read showed the day before that council passed its moratorium. New Albany has already enacted a one-year pause.
Notable splits (directional)
- Support is bipartisan: 92% of Democrats, 91% of independents, 86% of Republicans — Republicans supply the largest opposition share at 14% (directional).
- Seniors are the most emphatic: 13% of 65+ respondents strongly support the pause vs. 7% of 18-34s.
Question 8 · asked verbatim · n=1,000
“How important will a candidate's position on data centers be to your vote in the next local election?”
Very important
9.1%raw 5.1%
Somewhat important
83.5%raw 80.0%
Not too important
7.0%raw 13.3%
Not at all important
0.4%raw 1.6%
92.6% say at least somewhat important heading into the 2027 city cycle.
Notable splits (directional)
- Salience is generational: 95% of 18-34s say the issue will matter to their vote vs. 62% of 65+ — the young renters facing the bills are the most activated (directional).
What it means
For central Ohio's decision-makers
- This is not anti-growth sentiment. Only 11% say data centers mostly strain the grid (Q6), and support for the pause (Q7) mirrors the “not like this, not yet” pattern from our Prince George's County read — residents want conditions, not a ban.
- The consensus policy is cost assignment: 94% want data centers to pay for grid upgrades alone or share the cost, and only 4% pick the status quo — everyone's bills. The AEP tariff's 85% rule reads as the floor of public expectations, not the ceiling.
- The equity split is the sleeper: under-$35k households are nearly split on the tariff (55-44) while $125k+ households back it 82-18 — the households most exposed to bill increases are the least convinced the current fix protects them.
For readers of this report
- This entire study fielded in 83 seconds for the cost of GPU time, five days before the council hearing. The right use of a synthetic panel is a rigorous directional read the week the question matters — not a replacement for a human poll.
- Both calibrated and raw shares are published for every option, and the full data file includes party, age, and income crosstabs. Judge the method by the frozen numbers and what happens next.
Read this before citing
- Synthetic panel of AI respondents — not a probability sample, and no human residents were interviewed.
- Columbus is a benchmarked Lewsearch live panel (7.47% mean absolute error on 418 ex-electoral questions of a 460-question benchmark vs real polls (Pew, Gallup, UT/Texas Politics Project, PPIC)), but these specific questions have no human benchmark. Point estimates are calibrated reads; margins are the signal.
- “Somewhat” remains the modal answer on scale questions — a known behavior of model-based panels. Intensity shares (strong vs. somewhat) are model-derived; read support vs. oppose margins first. Calibrated shares near 2-3% reflect the 3% floor and should be read as near zero.
- Respondents are scored from full first-token answer distributions (router-parity method); crosstab cells are probability-weighted sums published in the data file and add up exactly to the toplines. Subgroup percentages are directional.
- Fielded and published July 17, 2026, before the July 22 Columbus City Council hearing. Toplines are frozen as published.
Join our email list
New issues and scorecards from The Lewsearch Report, straight to your inbox.