Hook
Sept. 26. The President of the United States said he wants to stop calling it artificial intelligence. He prefers "Super Intelligence." His stated reasoning: "artificial" implies something fake — and what he is looking at, in his framing, is not fake but extremely powerful.
No executive order. No NIST framework revision. No Federal Register entry. No budget line. One sentence, and it has already carried more narrative weight than a dozen semiconductor earnings calls.
I pulled the tape before I pulled the transcript. That order matters more than the quote itself. Within hours of the remark circulating through a Web3 news feed, the ASI-labelled token complex — the merged Fetch.ai, SingularityNET and Ocean Protocol stack that branded itself the Artificial Superintelligence Alliance in 2024 — was the fastest-moving AI basket on my watchlist. Not the largest. Not the most liquid. The fastest.
Alpha detected. Position established.
Context
The term is not Trump's invention, and that is the first thing to understand. "Superintelligence" entered policy-adjacent vocabulary through Nick Bostrom's 2014 book of the same name. In the safety literature, ASI denotes a system that exceeds human performance across essentially every domain — a hypothetical, formally distinct from AGI (human-level generality) and from "frontier model" (a present-tense engineering category). OpenAI's charter, by contrast, is written around AGI. The distinction is not pedantic. It is the boundary that separates a product roadmap from a speculative future, and every risk-management framework in the field is built on that boundary.
Now layer the policy environment. The current White House rescinded the prior administration's AI executive order in January 2025, replaced it with a "Winning the AI Race" posture, and tied frontier compute to energy permitting and grid buildout. Export controls on accelerators were reframed as a competitiveness instrument rather than a safety one. In that environment, a presidential preference for the stronger noun is not a slip of the tongue. It is ideological consistency.
Then layer crypto on top. The ASI Alliance merger folded three tokens into one. Announced in March 2024 and executed through a sequence of migrations that retired two symbols into a third, the deal concentrated the entire "artificial general intelligence" trade into a single liquid ticker. That structure is unusual. Most narratives in this market are spread across a dozen assets and a dozen whitepapers. This one has a face, and it makes the basket trivially easy to bid and trivially easy to exit.
And layer the regime on top of that: we are sideways. No directional macro catalyst. Chop is for positioning, not for conviction.
Core
Here is the mechanism nobody is writing down. In a rangebound market, narrative is the only variable that can move a cluster's correlation structure. When a head of state renames a category, market makers do not reprice a company. They reprice the cluster.
I have run this exercise before. In 2017 I read whitepapers as a student in Madrid, hunting consensus-mechanism flaws rather than investment theses. The projects that raised fastest were rarely the strongest technically — they were the ones with the cleanest noun in the abstract. A crisp category name is a fundraising primitive. That lesson never expired; it just moved upstream. In 2017 the noun was minted in a whitepaper. In 2025 the noun is being minted at the level of the American presidency, and the distribution channel is faster than any Telegram group ever was.
The transmission chain has four links, and only the last one is real:
- Headline — the quote. Free, instant, unpriced.
- Funding rate — perp longs pile into the labelled basket.
- Open interest — leverage builds faster than spot liquidity.
- Spot bid — the only link that touches an actual balance sheet.
Link one happened on Sept. 26. Links two and three are mechanical, and they are already visible. I pulled the funding curve on the ASI complex and on the broader AI-token basket the same afternoon. Positive flip, thin order books, and a spot depth profile that could not absorb the open interest it was being asked to carry. That configuration has a name, and everyone who has run a liquidation bot knows it.
Liquidation pending. Don't be the exit liquidity.
The link everyone skips is the fourth. Nothing in this event changed inference cost. No FLOPs were added. No revenue line moved. The chips are the same chips. The data centers are the same data centers. What changed is the label on the box — and in a narrative-driven asset class, the label is the product.
That is also why the "who wins" question is being framed wrong. The contest is not about which architecture deserves the name. It is about which protocol captures the naming tax first. I have watched the Layer 2 wars up close, and the honest read there is that OP Stack versus ZK Stack was never settled by proving rounds. It was settled by whoever convinced more teams to deploy chains under their banner. Same structure here. The first project to put "Super Intelligence" on a landing page and behind a perp ticker collects the narrative rent, regardless of what actually runs underneath it.
I ran a version of this playbook in early 2024, when the spot Bitcoin ETFs cleared and every European desk suddenly needed a translation layer for American flow. The lesson then was that institutional narratives do not move price because they are true — they move price because they change who is allowed to bid. A naming decision is exactly that kind of event. It adds no buyer. It removes a reason not to buy.
For my own diligence I track one hard signal: does the term survive contact with official text? The Federal Register, an OMB memo, a revision to the NIST AI Risk Management Framework, procurement language inside a defense appropriation. Terminology in procurement documents converts into budget lines. Terminology in a press gaggle converts into nothing. Right now we have the gaggle.
There is a crypto-native second-order effect worth flagging. If "artificial" gets reframed as a pejorative — fake, imitation, ersatz — then the regulatory tone around autonomous agents transacting on-chain softens by default. Agentic payment rails, machine-to-machine settlement, autonomous treasury management: all of it gets easier to pitch when the underlying software is officially described as "extremely powerful" rather than "simulated." That is not a technical unlock. It is a permission unlock. Different thing, same chart.
Contrarian
The unreported angle is semantic inflation, and it runs in the opposite direction from the price.
In the safety literature, ASI is a hypothetical system. Slapping that label onto a 2026-era language model collapses the distance between what exists and what is imagined. That distance is where every serious AI risk argument lives. Erase it, and you do not merely lose precision — you lose the basis for compute thresholds, disclosure requirements, red-team mandates, and the whole scaffolding that hangs on the 10^26 FLOP trigger. You deregulate a category by renaming it. Vocabulary as policy, executed without a signature and without a hearing.
Note also the source of record. This reached me through a Web3 feed, not a policy desk. That is information in itself. AI narrative is now distributed through crypto channels because crypto is the only venue where a narrative can be monetized within minutes of being spoken. The wrapper tells you what the content is for.
Which brings me to the rebranding tell I keep coming back to. I have spent two years arguing that most "Bitcoin Layer 2s" are Ethereum projects wearing a new logo, and the test is simple: strip the name and ask what actually changed. Apply it here. When the first "Super Intelligence" token prints, check whether it shipped a model, a cluster, or a paragraph. My base case is a paragraph.
Takeaway
Watch four things. The Federal Register — any administrative document adopting "Super Intelligence" as a defined term, which is the moment a phrase becomes a procurement line. NIST's next revision cycle, where vocabulary hardens into compliance language. The funding-rate curve on ASI-labelled perps, which separates positioning from an outright squeeze. And the spread between open interest and spot depth, which tells you when the squeeze ends and who is left holding it.
Arbitrage window closing in 10 minutes.
The question is not whether the technology is super. It is whether you are long the capability or long the word — and only one of those still has a bid when the sentence is forgotten.