Goldman Sachs warned that Democratic wins at the state and local level could trigger temporary data center construction moratoria affecting 45% of planned US data center capacity, according to a summary of its note posted on X on 4 November. We have not read the note.
On the ballot itself, limits on data centers lead in 40 of the 41 local measures on Ballotpedia’s 3 November list, by our recount of unofficial returns at 16:19Z. About 74% of the votes counted went to the limiting side, by our arithmetic.
Business Pill · Grid Cost Allocation
A short Business Engineer explainer on how grid costs are allocated between customers. It teaches the general idea only and does not discuss the elections or these ballot measures.
The key insight: By our recount of unofficial returns, limits on data centers lead in 40 of the 41 local measures on Ballotpedia’s 3 November list, with about 74% of counted votes on the limiting side; Goldman Sachs’s 45% is a different measure, the share of planned US capacity its note says temporary moratoria could affect, as summarised on X.
We re-read Ballotpedia’s list at 16:19Z. Its page still says six earlier 2026 measures were decided, all toward limits, and its scorecard still reads “Six decided • 44 pending”. Of its 41 measures dated 3 November, 38 now show vote counts, up from 35 when we last read it at 09:04Z.
The three new counts are all in Ohio: Oregon’s Issue 25 at 4,994 Yes and 3,235 No, and township zoning measures in Monroe (148 to 100) and Sprigg (330 to 297). A Yes vote limits development in all three.
Ballotpedia still shows Granville’s ban losing and the four Wilmington rezonings passing. Local reports citing the county boards say the opposite: Granville’s ban passed with 2,015 For and 429 Against, and Wilmington voters rejected all four rezonings by about 80%. With those five rows replaced, as in our correction on the first piece, all 38 lean toward limits.
For the three measures with no Ballotpedia count, we used the Data Centers on the Ballot tracker, which cites state and county results. Ottawa, Kansas, backs a ban 2,882 to 1,129, and Boulder City, Nevada, rejects a measure to allow data centers 1,973 to 4,726. Harwood, North Dakota, rejects a charter amendment requiring independent review of large industrial projects, 105 to 325.
That makes 40 of 41 toward limits, with Harwood the one exception. Across the 41, 158,164 of 213,525 votes counted went to the limiting side, about 74%, by our arithmetic. All 21 bans and pauses lead, and all 13 measures to allow, back or uphold development trail.
Share of votes counted for the side that limits data centers, per measure on Ballotpedia’s 3 November list. Data: Ballotpedia (16:19Z); Granville and Wilmington from local reports citing county boards; Ottawa, Boulder City and Harwood from Data Centers on the Ballot. All counts unofficial; grouping and shares are our arithmetic.
What Was Posted on X
GOLDMAN SACHS: DEMOCRATIC WINS COULD THREATEN 45% OF US DATA CENTER EXPANSION
Goldman Sachs warns that Democratic victories at state and local levels could trigger temporary data-center construction moratoria.
🔸 45% of planned US data-center capacity could be affected.
🔸 Texas alone accounts for roughly 25% of planned capacity.
🔸 Texas, Georgia and Michigan could introduce restrictions on tax breaks, electricity, water and environmental permits.
However, Goldman expects any moratoria to be short-lived, with political pressure easing over time.
The Walter Bloomberg account posted this summary of a Goldman Sachs note at 12:02Z on 4 November. It says 45% of planned US data-center capacity could be affected by temporary moratoria; we have not read the note itself.
Last night’s datacenter moratorium/restrictions appear to have gone 43 – 1, with a massive swell of public support averaging 77% to shut it all down. Public sentiment on on AI in America has been going steadily downhill for three years. It is once again at a new low.
Andrew Curran posted on 4 November that the data-center moratorium and restriction votes appear to have gone 43-1, with support averaging 77%. The 43-1 matches the Data Centers on the Ballot tracker’s count; we did not reproduce the 77%, against the tracker’s own mean of 73.4%. His post does not name a source.
Why the Counts Differ
The tallies circulating on 4 November count different things. The Data Centers on the Ballot tracker, prepared by Redbud Consulting, lists 46 measures on 3 November. At 11:00 ET its data showed 44 with results: 43 with the restrict side leading and one, Harwood, with the allow side leading.
That matches the 43-1 tally Andrew Curran posted on X. His average of 77% support is higher than the tracker’s mean of 73.4% at 11:00 ET, and we could not reproduce it. The tracker lists five more 3 November measures than Ballotpedia.
Engadget, citing the same tracker earlier in the morning, reported that “33 measures to restrict the facilities in some form had passed, with another six projected to go the way of anti-data center advocates”.
The Washington Post counts jurisdictions, not measures. Techmeme summarised its 4 November story as “US voters in 29+ jurisdictions overwhelmingly pass data center bans or restrictions”. Some places held several votes; Wilmington alone had four. The Post page returned an access error to our fetch, so we rely on the summary.
Jim Bianco wrote on X that “30 ballot measures sought to restrict data center construction. Restrict is leading 29 to 1”.
What Goldman Said
The summary posted by the Walter Bloomberg account (@DeItaone) at 12:02Z says “45% of planned US data-center capacity could be affected” and “Texas alone accounts for roughly 25% of planned capacity”.
It adds that “Texas, Georgia and Michigan could introduce restrictions on tax breaks, electricity, water and environmental permits”, and that “Goldman expects any moratoria to be short-lived, with political pressure easing over time”.
The 45% is a share of planned capacity at risk, not a count of votes, so it does not compare with the ballot tallies. Reuters reported on 5 October that Goldman then said political opposition had limited near-term impact on the expansion, and raised its forecast for US data center capacity at the end of 2026 to 64 GW.
What Else Was Reported
The same Techmeme summary of the Post story says “S&P says 145+ US projects were delayed or canceled through early August”. Capacity reported that 451 Research by S&P Global “tracked 3,173 operational US data centres with about 56.6GW of IT power as of Q2 2026, plus 1,140 planned facilities”.
On policy, Techmeme summarised Politico as reporting that Democrats “running on AI regulation and anti-data center platforms won key races, including Sherrod Brown in Ohio, as leadership signals aggressive AI oversight”. Techmeme summarised a Transformer piece by Shakeel Hashim as saying the sweep “raises the likelihood of strong frontier AI legislation passing in early 2027, including mandatory testing and developer liability”.
As we read the tallies, the gap between 29+, 40 of 41 and 43-1 comes from what is counted: jurisdictions or measures, Ballotpedia’s list or the tracker’s, and passed or leading.
As we read the Goldman summary, the risk it describes runs through state and local permits, tax breaks, electricity and water, the same levers the local measures use, but it is measured in planned capacity rather than in votes.
Goldman Sachs note, as summarised by @DeItaone on X, 4 Nov 2026
“Goldman expects any moratoria to be short-lived, with political pressure easing over time”
Three Implications
FOR DEVELOPERS All 21 bans and pauses on Ballotpedia’s 3 November list lead on unofficial counts, and all 13 measures to allow, back or uphold development trail, by our arithmetic.
FOR CAPACITY PLANNING The Goldman summary says 45% of planned US data center capacity could be affected and that Texas alone accounts for roughly 25% of planned capacity.
FOR THE TALLIES The tracker’s 43-1 at 11:00 ET and our 40 of 41 differ because the tracker lists 46 measures on 3 November and Ballotpedia lists 41.
The framework says: “The supercycle has a ceiling now, and the ceiling is political.”
As we read it, the local ballot results and the Goldman summary are two readings of that ceiling, one in votes and one in planned capacity. The framework predates these results and does not discuss the 2026 ballot measures.
All counts here are unofficial; no county has certified a result in our data. We did not re-check each Ballotpedia row against its county board, and the tracker’s three rows are its figures, not ours. Some measures are advisory. We have not read Goldman’s note or the S&P data behind the 145 figure, and we make no forecast about moratoria, capacity or legislation.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
By our recount of unofficial returns, limits on data centers lead in 40 of the 41 local measures on Ballotpedia’s 3 November list, with about 74% of counted votes on the limiting side. Goldman’s 45% is a separate measure: the share of planned US capacity its note says temporary moratoria could affect, as relayed on X.
A note on sourcing. Vote counts are unofficial, from Ballotpedia (read 16:19Z on 4 November 2026), local reports citing county boards of elections and the Data Centers on the Ballot tracker (11:00 ET). Goldman Sachs figures are as summarised on X; we have not read the note. The Washington Post figures are as summarised by Techmeme. Groupings, shares and totals are our arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
