Three data center numbers went viral on the night of November 3, 2026: 13.3 GW of Texas projects "at risk," a 56% exit-poll figure, and ten local ballot votes that all went toward restriction. We could trace the context behind each, but not the exact figure behind any of them. This post separates what is sourced from what is not, so you can use the numbers with the right caution.
Our earlier guides cover the setup: the Ohio ballot measures explained and where ten midterm races stand on a pause. This post checks the claims that came after the polls closed. Results were still partial at the time of writing, so we do not state any final outcome.
TL;DR: the three claims at a glance
| Claim | Who said it | What we verified | Status |
|---|---|---|---|
| Jefferies says 13.3 GW of announced Texas projects are at risk if Gina Hinojosa wins | Relayed on X by market accounts and a reporter | Hinojosa's moratorium call is real. We found no Jefferies note and no outlet quoting it | Unconfirmed |
| NBC exit poll: 56% say data centers were a crucial or major factor, 23% crucial | Relayed on X by several accounts | No NBC release found. Pre-election polls show wide opposition | Unconfirmed |
| First 10 reported ballot measures all favored restricting data centers | Relayed on X from early returns | Ballotpedia confirms the four earlier measures all backed limits. The election-night count is not final | Partly confirmed |
Claim 1: "13.3 GW of Texas data center projects at risk"
The claim, as it spread on X, says Jefferies estimates that 13.3 GW of announced data center projects in Texas could be at risk if Democrat Gina Hinojosa wins the governor's race, because she has called for a broader moratorium. Accounts relaying it also tied it to a drop in tech stock futures. We cannot confirm either the Jefferies attribution or the market move, and we do not treat share prices as evidence of anything about policy.
What is verified about Hinojosa's position
The policy part is documented. KFOX reported that on Thursday, July 23, 2026, Hinojosa urged Gov. Greg Abbott to sign an emergency executive order imposing a pause on data centers until the Legislature can act. She said Abbott "has the power to do it today," and, in the same report, "I'm calling for a moratorium until a legislative session that puts Texans first."
She also proposed priorities for the next session. They include giving communities a role in deciding whether and how data centers are built, barring nondisclosure agreements, making sure households do not pay higher electricity or water costs, ending subsidies so data centers "pay for themselves," and making water contracts unenforceable if they threaten a community's supply.
What is verified about Abbott
The same report says Abbott called on June 30 for a ban on data centers in rural Texas. Weeks before that, he wrote to the Public Utility Commission of Texas and the state grid operator ERCOT, asking them to stop data centers from draining water and to keep infrastructure costs from landing on communities. So this is not a clean "pro-build versus anti-build" race. Both candidates have moved toward tighter rules. Our earlier post on Abbott's closed-door data center roundtable shows the state was also discussing a fast track for some projects.
Where the 13.3 figure is shaky
Searching for the number turned up a different story. One search summary described a 13.3 GW figure as a utility's projection of peak data center demand years from now, not a count of Texas projects at risk. We could not open that source, so we do not rely on it, but it shows how easily a gigawatt number can attach itself to the wrong claim.
The scale that is sourced is much larger. Search summaries of industry reporting put ERCOT's large-load queue at more than 1,800 projects and more than 474 GW of requested demand, with roughly 90% of it tied to data centers. A 13.3 GW slice of a queue that size would be under 3%. That is a useful sanity check: even if the Jefferies number is real, it describes announced projects, not the whole pipeline, and many queue entries never get built.
How to check it yourself: look for the Jefferies research note title and date, a named analyst, and a definition of "announced." If none of those appear in a post that cites the figure, it is a rumor with a brand name attached.
Claim 2: "56% of voters say data centers were a key factor"
The second claim says NBC News exit polling found that about 56% of voters called data centers a "crucial or major" factor in their vote for Congress, and 23% called them crucial. It arrived in a handful of X posts within minutes of the polls closing.
We could not find the NBC release or any outlet reproducing the question wording. That matters because exit-poll items vary a lot. "AI and data centers" combined in one question can produce a very different number than data centers alone. Treat 56% as a claim about one unseen question, not a general measure of how voters feel about data centers.
What earlier polls actually show
Rounded houses with a blank sign, showing local opposition to data centers measured in pre-election polls
There is solid pre-election data that points the same way on opposition, though not on voting priority:
- A Fox News poll of 1,211 registered voters, fielded September 11 to 14 and published September 22, found 71% opposed to building AI data centers in their area, including 46% strongly opposed. Margin of error was plus or minus 3 points. It also found 52% favor building them in remote locations.
- In the same poll, 60% viewed AI unfavorably overall, and 63% said AI's effect on jobs is bad.
- Search summaries of an NYT/Siena poll from mid-September reported 61% of likely voters opposed building data centers to support AI. A search summary of a Reuters/Ipsos poll put the cost of living as the top vote factor for 47% of potential voters.
The gap is the point. People say they dislike data centers near them in large majorities, yet in other surveys data centers rank below cost of living as a reason to vote. A 56% "crucial or major" result would be a jump, which is a reason to wait for the actual exit-poll tables. Our posts on the ABC/Washington Post poll on AI and data centers and the Maine data center revolt cover the same split between opposition and priority.
Claim 3: "The first 10 ballot measures all favored restrictions"
A ballot box taking a check card, standing for early data center election results that are still partial
The third claim says that on election night, the first ten reported local data center measures all favored limiting the facilities, out of a slate of 41 measures in 10 states, 21 of them in Ohio.
What Ballotpedia confirms
Ballotpedia's September count listed 38 local measures across 35 jurisdictions in 30 counties and 8 states. Ohio had 20 of them in 17 jurisdictions, more than half. Thirty-four of the 38 (89%) were on the November 3 ballot, and the other four had been held earlier in the year. In all four earlier votes, residents backed tighter limits, either by approving a restriction or by rejecting a measure that would have allowed development.
The numbers differ from the election-night claim (38 versus 41 measures, 8 versus 10 states) because counts kept changing as courts added and removed questions and as Ballotpedia updated its tally. That is the same count problem we flagged in the Ohio ballot guide.
Why "10 of 10" is weaker than it sounds
Three cautions apply before you read a trend into early returns:
- Order of reporting. Small precincts report first. Whether those are typical of the later ones is unknown.
- Mixed measure types. Ballotpedia says 9 of the 38 are advisory only. The other 29 enact, amend, approve or reject charters, zoning decisions, laws or development agreements. A yes on an advisory question and a yes on a charter ban are not the same outcome.
- Who writes the question. Ballotpedia found 31 of 38 (82%) came from citizen petitions and 7 from local governing bodies. Petition measures are written by the people who want a restriction, so a pass rate for restrictions is partly a selection effect.
Final totals will come from county election boards and Ballotpedia, not from a screenshot. Wait for those before quoting any win rate.
What builders should take from this
None of these three claims changes a model price or a GPU lease today. The useful reading is slower and more practical.
Siting risk is now political risk. If a state or a county adds a pause, a cap, or a public-vote rule, every project in that area slows. For API users, the effect arrives as capacity limits and higher prices over quarters, not days. Our coverage of the New York moratorium debate and the San Francisco 45-day moratorium shows the same pattern in other places.
Watch the mechanics, not the headline. The key variables are the size threshold (a 25 MW cap blocks most AI campuses), whether projects already permitted are protected, and whether costs are pushed onto ratepayers. The Senate investigation into hyperscaler cost claims and the permitting bill that would make data centers pay transmission costs are where those details get written into law.
Be skeptical of gigawatt headlines. A queue number is a request, not a build. A moratorium number is a scenario, not a forecast. Ask for the source note, the date, and the definition.
A short verification checklist
Use this for the next viral number about AI infrastructure:
| Question | Why it matters |
|---|---|
| Is there a named document (note, poll release, filing)? | A named document can be opened and read. A relayed number cannot |
| What exactly was measured? | "Crucial or major," "announced," and "queued" all mean different things |
| Who gains from the claim spreading? | A research desk, a campaign, or an account selling attention |
| Is the sample or base stated? | Exit polls and early returns are partial by design |
| Do two independent sources agree? | One source repeated by ten accounts is still one source |
What we will update
This post is a snapshot from the morning after. When the Jefferies note, the NBC tables or the county results surface, we will add an update section here with the numbers and links. If a claim turns out to be wrong, we will say so at the top.
Details are accurate as of November 4, 2026. Election results and poll figures can change as counts are finalized.
Related reading
- Ohio data center ballot measures on election day
- Data center backlash and the midterms: Texas, Pennsylvania, Ohio, Michigan
- Abbott staff roundtable on data centers and ERCOT fast track
- ABC/Washington Post/Ipsos poll on AI and data centers
- Maine data center revolt and the Senate race
- New York data center moratorium explained
- Senate investigation into hyperscaler data center cost claims