Post from Truth Social

A hoax ledger with AI added at the end

The post begins with a list: "RUSSIA, RUSSIA, RUSSIA, UKRAINE, UKRAINE, UKRAINE, Global Warming, Impeachment Hoax #1, Impeachment Hoax #2, Men in Women's Sports, Transgender for Everyone". The entries don't belong to one category. Two federal investigations, a scientific consensus, two House impeachment votes and two cultural disputes are all filed as "Hoaxes, all generated by the Radical Left Dumocrats, for purposes of destroying our Country." Criticism of AI then goes on the end of the list as "the decimation, or destruction, of AI."

This is splitting applied to the whole opposition. Once a concern is traced to "Dumocrats," it counts as a hoax, and its substance never comes up. The post offers no argument about whether any AI criticism is correct. Where the criticism comes from is treated as settling it, which is the closed-loop epistemology the schema calls epistemic closure. The attribution is also inaccurate. Through 2025, resistance to data centers and to federal preemption of state AI laws came from both parties: the Senate voted 99-1 to strip a state-AI-law moratorium from the 2025 reconciliation bill, and Republican governors and senators led part of that fight. "For purposes of destroying our Country" gives the opponents a destructive motive the post never shows. That is projection of intent, stated as fact.

The data center story told as a won battle

The middle section tells a short redemption story: "It all began with an attack on our Data Centers, until people realized how wealthy and prestigious they were... Higher Salaries, Lower Taxes, and Safer Streets, was the result, and the crazed Data Center attack has largely failed." Disputed local benefits are presented as a finished result, and "Safer Streets" has no plausible link to data center construction. The story is rationalization. It explains why critics have moved to a new target without granting that the old target (power prices, water, the small number of permanent jobs) had any merit.

He then explains the shift with an analogy: opponents "changed the term 'Global Warming' to 'Climate Change'" to widen the "territory" of doubt, and are now "going straight at AI." The analogy fails on its own facts, since the best-documented push for "climate change" over "global warming" was Frank Luntz's 2002 memo to Republicans. It also sits awkwardly next to his own post from earlier the same day, a poll asking followers to rename AI "Superior Intelligence," "Extreme Intelligence" or "Supreme Intelligence." Hours after proposing a rebrand himself, he accuses opponents of winning ground by rebranding.

Nurture language, then a noun in capitals

The tone turns warm: "we will cherish it, help it, and watch over it, as it grows!" These are the verbs of raising a child, used about an industry. The post's only communion-register language goes to AI as an object of national prestige. Every human group in the post is either an enemy or unmentioned.

"However, we will also be looking for BAD, and we can do that, very easily, with our already existing Criminal and Civil Justice System." Read as policy, this means current law covers AI harms and no new regulation is needed, which matches the administration's position against state AI rules. But "BAD" is capitalized and never defined, and it comes right after a paragraph about enemies trying to destroy the country. A reader can fill it with the critics as easily as with bad AI actors. The sentence names no person and gives no instruction, so it adds ambiguity without mobilizing anyone.

Grandiosity follows in two familiar forms: "I am forming the AI Force, much like I did Space Force, which has been a tremendous SUCCESS," and "Only High I.Q. individuals need apply!" The new agency is justified by the earlier one, and IQ serves as a status gate, a long-running motif in his posts. Presenting the "AI 'Czar'" as a new post leaves out that the title existed in 2025, when David Sacks held it. Whether it is vacant by September 2026 isn't established in the context provided. The closing forecast, "possibly as much as 25% of our Country's GDP," is far above mainstream estimates, and "We are leading China... and I intend to keep it that way!" looks ahead to the Xi state visit on September 24, where AI is on the agenda.

Language: self-repair and a backwards gloss

Two phrases show small production slips. "The decimation, or destruction" corrects itself as it goes, a gloss he often uses. "AI, commonly known as Artificial Intelligence" reverses the usual order: the abbreviation is the common term, and he glosses it as though the full name were. His morning post loops the same way ("the words 'Artificial Intelligence' are inaccurate... relative to AI, or Artificial Intelligence"). The same circular gloss twice in one day on one topic looks more like topic fixation than word-finding failure. It is a mild circumlocution to track, and on its own it doesn't mark decline.

The long sentences stay on track. The post moves from hoaxes to data centers to rebranding to the AI Force to China, and each step follows from the last. Syntax is at his usual level: comma splices ("Higher Salaries, Lower Taxes, and Safer Streets, was the result"), random capitalization, and em-dash joins.

Who wrote it

The time was 17:12 UTC, 1:12 pm Eastern on a Saturday, if he was in Washington or Bedminster. That is office hours and would ordinarily point toward staff. The text points the other way: the coined "Dumocrats," the self-correction "decimation, or destruction," the reversed acronym gloss, the subject-verb mismatch, and a hoax list that wanders across categories. The announcement itself (an AI Force, a czar) may have started with staff, but the wording reads as his own dictation or typing. It also matches the voice of his morning renaming poll and his same-day Supreme Court complaint.

Context for the day

This is one of several posts on September 19, along with the AI renaming poll and an attack on the Supreme Court over tariffs and birthright citizenship ("lacked the Courage"). The court loss is the one clear injury of the day. This post doesn't mention it, but it has the same structure: an institution or movement blocks what he wants, and its opposition is recast as bad faith. The state is grandiose-dominant, with the country's enemies briefly cast as the victim's persecutors before he pivots to protection and command. The archetype is the Guardian-King. He stands over a precious national asset ("watch over it"), with a Warrior frame aimed at a "crazed" opposition.

Fact Verification

Opponents are winning local fights more often than before. Data Center Watch's Q1 2026 report is its latest (no Q2 2026 edition is posted). It counted at least 75 projects worth about $130 billion blocked or delayed from January through March 2026. That is the largest single quarter since tracking began in 2023, and about equal to all of 2025. Active opposition groups more than doubled, from 396 at the end of 2025 to 833 across 49 states. More than 300 state bills were filed in the first six weeks of 2026, and 14 states saw moratorium proposals. Heatmap Pro's review of public records (published July 30, 2026) found:

• more than 500 counties or municipalities restricting or blocking new data centers, nearly 190 of them added since June 1
• more than 50 projects canceled after local pushback in 2026, more than double the roughly 25 in all of 2025
• about a 50% chance that a project challenged by local activists gets canceled (71% in Michigan, 56% in Indiana, 17% in Texas)

This summer the fight reached state governments:

• New York: Gov. Kathy Hochul's Executive Order 62 (July 14, 2026) is the first statewide moratorium. It pauses discretionary environmental permits for facilities of 50 MW or more for one year.
• Texas: Republican Gov. Greg Abbott ordered a PUCT/ERCOT audit on Aug. 3, 2026. It froze approvals for the roughly 1,800 projects (474 GW) in ERCOT's large-load queue, which BNEF puts at about 20% of the U.S. data center pipeline. On Aug. 18 the Data Center Coalition agreed to his conditions: pay for their own grid infrastructure, reuse their own water, avoid disrupting neighborhoods, and give up taxpayer incentives. Those are largely the opposition's own demands.
• Pennsylvania: Gov. Josh Shapiro issued an executive order tightening permits.
• Maine: the legislature passed a statewide moratorium in April 2026, which Gov. Janet Mills vetoed.

The biggest single fight also went to the opponents. Virginia's Prince William Digital Gateway (23 million square feet on 2,100 acres near Manassas battlefield) died on July 2, 2026, when QTS dropped its last appeal after residents' lawsuit voided the rezonings.

Polls show opinion turning against data centers, the opposite of what the post says happened:

• Gallup (March 2-18, 2026): 71% oppose AI data centers in their area, 25% favor them.
• Reuters/Ipsos (June 3-8, 2026): 57% would oppose one in their community, and just 14% would be fine with one nearby.
• Heatmap Pro/Embold (Aug. 8-13, 2026): 75% oppose one near them, up from 42% in August 2025, when support and opposition were split evenly. Republicans are net 43 points against.
• Fox News: 70% opposed, including 60% of Republicans.

The kernel of truth is that opposition hasn't stopped the national buildout:

• Census data show data center construction spending up about 57% year over year in July 2026.
• ConstructConnect counts $81.5 billion in starts through June 2026, already above all of 2025 ($72.5 billion).
• The four largest hyperscalers planned about $630 billion in 2026 capital spending.
• Many projects still win approval. Weld County, Colorado approved what would be the state's largest on Sept. 9, 2026.

That spending mostly reflects sites approved earlier, though. The losses are piling up among new proposals. "Largely failed" holds only in the narrow sense that construction continues at record levels. Every tracker and poll found contradicts the claim that communities came around and the opposition faded. | | "AI could reach as much as 25% of US GDP" | Mostly False | The post says AI will be bigger than the Internet, "possibly as much as 25% of our Country's GDP." It gives no timeframe, and "possibly as much as" makes the number a ceiling rather than a projection. That means it can't be checked against outcomes yet. It can be checked against published forecasts, and it sits far outside the mainstream.

Read literally, the claim is that AI could make up a quarter of the economy. No forecast found supports that:

• AI computing infrastructure was about 1.5% of U.S. GDP in Q1 2026 (Epoch AI).
• Goldman Sachs projects AI capital spending at 1.8% of GDP in 2026, 2.5% in 2027 and 2.8% in 2028. That fits the 2-5% peaks of earlier technology buildouts.
• IDC's estimate of AI's total economic impact is 3.5% of global GDP in 2030.
• The whole digital economy, three decades into the Internet era, was 10.3% of GDP in 2021 by BEA's measure ($2.41 trillion), or 18% in 2024 by IAB's broader internet-economy count.
• AI's big share of recent growth doesn't change this. The St. Louis Fed puts it at about 39% of GDP growth in the first nine months of 2025, and David Sacks said AI was 75% of Q1 2026 growth. Those numbers measure growth, not the size of the economy, and a Federal Reserve note says imported equipment offsets much of that investment.

Read generously, as AI raising GDP by up to 25%, the figure sits at the very top of published scenarios. The White House's Council of Economic Advisers (2026 Economic Report of the President, chapter 5, table 5-1) lists estimates from about 1% to more than 45%:

• It calls Oxford Economics (1.8-4% after 8 years), McKinsey (2.4-4.1%) and Goldman Sachs (7% over 10 years, global) midrange.
• It calls PwC (up to 15% by 2035, global) and a BIS paper by Aldasoro and colleagues (20-45%) high. In the BIS model, U.S. GDP ends almost 30% above a no-AI path after a decade.

Other estimates run far lower. Acemoglu estimates 0.9-1.6% over 10 years, and the Penn Wharton Budget Model 1.5% by 2035 and 3.7% by 2075. Anthropic's September 2026 scenario tool gives +1.6%, +8.3% and +32.4% by 2030, and says the scenarios aren't predictions. The Forecasting Research Institute's 2026 survey found economists' median forecast was about 2.5% annual growth in 2030 and 2050, close to the historical trend. Its rapid-AI scenario would leave 2050 GDP about 25% larger, but economists gave that scenario only a 10-14% chance.

Trump's own number also moves. On Dec. 11, 2025 he said AI "could be 50, 60 percent of our economy."

As a boost to GDP, 25% isn't impossible, since a few aggressive scenarios reach or pass it. But it is several times the central estimates of 1-7%. As a share of the economy, which is how both the post and his earlier remarks frame it, nothing supports it. The hedge "possibly" keeps it from being flatly false. | | "The United States is leading China in AI" | Mostly True | The Stanford AI Index 2025 counted 40 notable US models against 15 from China in 2024, and the US leads in advanced compute. The same report found the performance gap between top US and Chinese models had narrowed sharply. |

Overall Veracity: 37%