“Super Intelligence”: what the great AI renaming actually changes
On September 22, 2026, from the UN podium, Trump announced that US government documents will now say “Super Intelligence” (SI) instead of “Artificial Intelligence”. Behind the PR stunt lies a very real race for means: 26 GW of silicon contracts, $500B committed, a Congress trying to ban ASI. We broke down the facts, the players, and what it changes for an SME. No promises, just numbers.
A speech at the UN, an official renaming
This is not just another tweet. It is a statement before the UN General Assembly, relayed by Reuters: the official vocabulary of the US government has shifted.
“From now on, the United States will call AI ‘Super Intelligence’ — AI or, SI.”
“The only guardrail AI needs is a strong and intelligent president.”
“A race without guardrails risks a gigantic catastrophe.”
Sixteen days that produced “SI”
Congress fights back
Sanders and Casar introduce the “Ban Artificial Superintelligence Act”: ban systems equal to or greater than humans. A direct response to executives declaring ASI “imminent”.
The founding poll
On Truth Social: “Super Intelligence vs. Superior Intelligence — which name for the new phenomenon?” followed by “Whoever wins AI, WINS!”
Jensen Huang: “AGI has arrived”
NVIDIA’s CEO gives the industrial signal that legitimizes the vocabulary shift.
“Hoax buster”
AI warnings become a hoax “crazier than the Russia hoax or global warming”. Public attack on Dario Amodei (Anthropic), who had been arguing for slowing down.
Altman: 10% extinction risk
And OpenAI’s IPO postponed to prioritize safety. The debate leaves tech circles and becomes frontal.
AI Force + AI Czar
Announcement of an “AI Force” modeled on the Space Force, and of an “AI Czar” to be named. No budget, no name. Same day: antitrust complaint against OpenAI.
Second round of the poll
“Super Intelligence vs. Superior Intelligence” — first appearance of the “SI” acronym.
Speech at the UN
Renaming made official in government documents, warnings reclassified as “hoax”, call to accelerate against China. US-China talks on AI on the sidelines of the summit.
Three federal layers, one legislative counterweight
The rebranding covers a governance built in superimposed layers — with no unified budget and no “AI Manhattan Project” funded as such.
Genesis Mission — the science
Federal datasets and compute platform run by the DOE and the National Labs. More than $5B committed, framework voted in late 2025, currently being executed.
EO 14409 — the security
Executive order “Advanced AI Innovation & Security” signed on 02/06/2026: government model sharing 30 days before publication, voluntary pre-disclosure on the industry side.
AI Force + AI Czar — the military
Announced Sept 19 on Truth Social, Space Force model. Today: an intention. Zero published budget, no czar named.
Providers & compute: the numbers that matter
The dominant design of this race: labs do not own their assets — they pre-purchase silicon capacity through multi-year contracts or rent from hyperscalers. The record for the largest AI datacenter doubles roughly every 7 months (Epoch AI, 86 sites tracked).
| Initiative | Players | Commitment | Status |
|---|---|---|---|
| Stargate | OpenAI · Oracle · SoftBank | $500B by 2029 · 10 GW target | Under construction — Abilene TX (1.2 GW), TX, NM, WI, MI, OH |
| OpenAI silicon contracts | NVIDIA ≥10 GW · Broadcom ~10 GW · AMD 6 GW | ~26 GW cumulative | LOIs and agreements 09-10/2025 |
| Colossus 1+2 | xAI | ~$18B · 946 MW in operation | Operational — Memphis TN, largest site in the world |
| Anthropic compute | Google Cloud (1M TPU) · AWS | ~$65B in commitments | Underway — 3.5 GW from 2027 |
| Genesis Mission | DOE · National Labs | > $5B | Underway — federal platform |
What it actually changes for you
The vocabulary will seep in everywhere
Government documents, tenders, certifications: “SI” will work its way into official US formulations, then by contamination into RFPs and maturity grids. Nothing to do right away — but keep an eye on your glossaries and your “AI” pages.
Deregulation on the US side, regulation on the EU side
The two blocs are diverging: acceleration and rebranding in the US, AI Act and labeling in Europe. A company operating on both sides must document its AI uses for the EU without being hit by American FUD.
Compute remains the bottleneck
26 GW of contracts for a single player means: capacity prioritized for the giants, stable or rising prices for everyone else. An SME does not buy GW — it arbitrates between managed APIs, open-weights models and custom agents.
The signal not to follow
“AGI has arrived” (Huang) and “Super Intelligence” (Trump) are positioning statements, not audits. What matters for you: your measured use cases, your costs per task, your data governance. The rest is background noise.
The automation window is opening
While the majors commit to a race for means, usable models (reasoning, vision, voice) remain affordable. The value for an SME is in orchestration — workflows, agents, integrations — not in the model race.
Geopolitical vigilance on data
The race rhetoric (US vs China) increases the risk of technical fragmentation: diverging standards, silicon export restrictions, residency clauses. If your IT depends on a single US provider, document your plan B. In Europe, the echo is already here: Belgium reports 1,000–1,500 security investigations opened per year (VSSE, ministerial answer 03/2026) against hybrid threats — espionage, interference, cyberattacks, disinformation. The technical fragmentation your IT fears is an active intelligence dossier, not a hypothesis.
What is established, what remains open
Main references
Want an AI watch that only brings you what matters?
This white paper is how we work: sourced facts, verified numbers, zero hype. We apply the same rigor to the automations we deploy.
Talk about your AI project →© 2026 Iamin.Digital OÜ · Watch white paper — written with AI assistance, verified and edited by the iamin team (EU AI Act art. 50 disclosure: AI-generated content). Full internal research file: September 22, 2026, multi-source collection (press, technical, social media). Metadata: Rank Math · AI-generated disclosure meta.