Index Ventures and Benchmark just priced a free app at $2.5 billion, and the founder's best public evidence of traction is that people used it to buy groceries.
Instinct's valuation rose 5x in three weeks with zero disclosed revenue, users, or pricing, a number that at a generous 25x multiple implies $100 million a year that nobody has claimed.
“Instinct, a personal AI assistant startup, is worth $2.5 billion after raising a $250 million Series B just three weeks after a $500 million Series A mark.” [SOURCE ↗]
Instinct, the viral personal AI assistant, raised a $250 million Series B co-led by Index Ventures and Benchmark at a $2.5 billion valuation, three weeks after a round that marked it at $500 million. The money is real and the coverage confirms every dollar of it. What the coverage contains nowhere is a revenue figure, a user count, or a price: the product is free, invite-only, launched in February 2026, and has never named a business model. At a generous 25x forward-revenue multiple, $2.5 billion implies roughly $100 million a year in revenue that nobody, including the founder, has claimed. The founder's offered evidence of traction is that users planned road trips, bought groceries and concert tickets, and cancelled some subscriptions.
On August 26, 2026, TechCrunch reported that Instinct, the viral personal AI assistant, had raised a $250 million Series B co-led by Index Ventures and Benchmark at a $2.5 billion post-money valuation, bringing its total raised to $350 million. Three weeks earlier the company carried a $500 million mark. That is a 5x repricing in three weeks for a product that is free, invite-only, and about six months old.
We are not calling the raise fake. Every dollar checks out. We are calling the number unaccompanied.
The evidence trail, receipt by receipt
Receipt 1: the TechCrunch funding story (August 26). It confirms the $250 million, the $2.5 billion, the $350 million total, and the Index plus Benchmark co-lead. Then it reaches for traction and produces this: users have "planned cross-country road trips, bought weekly groceries and concert tickets, and cancelled hundreds of dollars of subscriptions." That is the founder's own offered evidence. Three anecdotes, zero numbers. The largest quantity in the entire traction case is "hundreds of dollars," and it describes money users stopped spending.
Receipt 2: SiliconANGLE (August 27). Independently confirms the shape of the story and adds the prior round: $100 million from a consortium including Conviction Partners and Greenoaks, and a market entry in February 2026 as an invite-only private beta. It also carries no revenue or user disclosure. Two outlets, one silence.
Receipt 3: the business model analysis. "The product is free, invite-only, and has never named a business model." That line is reported. The $100 million a year figure built on it is not: it is this outlet's own illustrative math, a generous 25x multiple applied to the valuation, offered here to size the silence rather than as evidence of what Instinct actually earns. Nobody, including this analysis, has claimed a real revenue figure.
Receipt 4: the terms of service (TechCrunch, August 24, two days before the raise). Instinct's terms grant it a "perpetual and irrevocable" license to "access, use, host, cache, store, reproduce, transmit, display, publish, distribute, and modify" what flows through it. That is a separate story about data rights. We flag it here as an open question worth watching, not as proof of what the $2.5 billion is actually pricing.
Receipt 5: the $500 million mark. The prior Series A comes from one outlet: "Kleiner Perkins led a $75 million Series A at a reported $500 million valuation in early August." It is not independently confirmed elsewhere in this evidence set. The pace (weeks, not months) and direction are consistent with everything else here, so we keep the figure, but treat the exact 5x multiple as approximate rather than confirmed.
The steelman, taken seriously
The best case for $2.5 billion goes like this. Index Ventures and Benchmark are not tourists. They saw the private dashboard: retention curves, invite-list length, growth rates the press never gets. Series B consumer pricing has never been about the current P&L. It prices the chance that this becomes the default way people buy things, and if agent-executed commerce becomes a category, the toll booth on it is worth far more than $2.5 billion. All of that is fair, and it is why the verdict is needs context rather than false.
But notice what the steelman concedes: the case for the number lives entirely on evidence nobody outside the room can inspect. The public case, the one the founder chose to make, is groceries and concert tickets. Companies do withhold metrics for competitive reasons. They rarely withhold good ones for six months while raising $350 million.
Who benefits from an unaccompanied number
A valuation with no disclosed fundamentals is not a measurement, it is a press release with a dollar sign.
And it works for everyone in the room. The startup gets recruiting gravity and free distribution ("the $2.5 billion assistant") without publishing a single metric a competitor or journalist could check. The earlier investors, Conviction Partners and Greenoaks among them, watch a three-week-old $500 million mark become $2.5 billion, a 5x paper markup they can show their own backers. The new leads get entry into a hot asset plus a headline that makes the price look like consensus rather than a bilateral negotiation. The only party with no seat at the table is the reader, who receives the number as if it described the business.
Five questions before you believe the next billion
1. Can anyone pay this company money today? If there is no price, there is no revenue, and the valuation is pricing something else. Find out what. 2. Divide the valuation by 25. That is roughly the annual revenue a generous multiple implies. Has anyone, anywhere, claimed a number in that neighborhood? 3. Count the traction evidence. Anecdotes have no denominator. "Users bought groceries" is not a metric, it is a vibe with a shopping cart. 4. Ask who held shares at the old mark. A markup is also a favor to everyone already on the cap table. That does not make it wrong. It makes it non-neutral. 5. Read the terms of service. In a free product, the business model is usually in there, in the license grant. Instinct's is perpetual and irrevocable.
None of this means Instinct fails. It means that today, August 31, 2026, the only verified number in the story is the one investors paid, and the number they paid it for does not exist yet.
Hot-round valuations price the option on a category, not the business that exists, and that is a legitimate bet for the people writing the checks. The problem starts when the number leaves the room: a $2.5 billion headline does recruiting, press, and next-round work with nothing on the other side of the ledger, and every uncontested billion makes the next one cheaper to print. When the product is free and the terms of service claim broad rights over user data, the thing being priced may be the data, not the software.
Dated August 31, 2026: Instinct takes its next markup before it publishes a single revenue, pricing, or user number. If a disclosed figure beats the next round to print, we were wrong.
Instinct or its lead investors disclose a revenue run rate, paying-user count, or product pricing that makes the $2.5 billion legible (on the order of $100 million a year at a normal 25x multiple), or they publish the business model with numbers attached. A credible report of internal ARR at that scale, from an outlet holding the documents, also flips this from needs context to holds.
RECEIPTS (5) · CONFIDENCE HIGH
every URL below answered a live HTTP check before publish · sweep 2026-08-17
- ▲ techcrunch.com ⧉ · “They've planned cross-country road trips, bought weekly groceries and concert tickets, and cancelled hundreds of dollars of subscriptions.”
- ● siliconangle.com ⧉ · “previously raised $100 million from a consortium that included Conviction Partners and Greenoaks”
- ▼ businessmodelanalyst.com ⧉ · “The product is free, invite-only, and has never named a business model.”
- ● techcrunch.com ⧉ · “perpetual and irrevocable" license to "access, use, host, cache, store, reproduce, transmit, display, publish, distribute, and modify”
- ● businessmodelanalyst.com ⧉ · “Kleiner Perkins led a $75 million Series A at a reported $500 million valuation in early August.”
