Instinct's $10bn Series C shows how fast venture prices are moving in AI agents
A 14-person consumer agent startup quadrupled its valuation in 33 days. The round says more about investor fear of missing the category than about the business.
On Monday 28 September, Instinct, a San Francisco start-up that sells a personal AI agent which carries out errands by text message and phone, announced a $1bn Series C at a $10bn valuation. Sequoia Capital, Benchmark Capital and Coatue backed the round. The company was founded by Noah Shinn, a 23-year-old former research scientist at Sierra, and its product is still available only by invitation.
The headline numbers would be remarkable for any young company. What makes them significant for the venture market is the speed. Instinct's previous round, a Series B at a $2.5bn valuation co-led by Index Ventures and Benchmark, was announced on 26 August. In the 33 days that followed, the price of the company rose fourfold, and it did so without the company publishing revenue, a business model or a single growth metric of its own. As the third quarter closed on Wednesday, the deal stands as the clearest expression of how much capital, and how little patience, late-stage investors now bring to the AI agent category.
A valuation ladder climbed in weeks
The sequence of rounds is worth setting out. According to reporting by OfficeChai, an early round valued Instinct at roughly $50m, with Conviction and Greenoaks as backers. Kleiner Perkins then led a $75m Series A at about $500m in early August. The Series B followed at $2.5bn later that month, and the Series C at $10bn arrived at the end of September. Total funding now stands at more than $1.3bn.
That is a 200-fold increase in valuation within a few months, compressed into a period in which most companies would still be negotiating the terms of a single round. The company itself remains tiny: OfficeChai reports 14 employees, and an operating entity, Spear Street Technology, that was registered in California in April.
The product launched on an invitation basis in August. Users message or call the agent, which uses its own phone number and computer to book travel and restaurants, pay bills, cancel subscriptions, order groceries and conduct research. Recent additions include a concierge feature that telephones businesses without online booking and a network that lets one user's agent coordinate plans with the agents of friends. That network, opened to all users on 14 September, has handled more than 300,000 coordinations, according to OfficeChai. Instinct has crossed 100,000 users, a figure first reported by The Information. The service is free, and the company has disclosed no revenue.
What investors are actually buying
Seen in conventional terms, a $10bn price for a free, invite-only service with no disclosed revenue is difficult to defend. Investors are plainly not paying for current economics. They are paying for three things.
The first is the founder. Mr Shinn was first author of Reflexion, a 2023 research paper on language agents that learn from failed attempts, and at Sierra he co-developed a benchmark for agent tool use. In a market where a small number of researchers are seen as capable of building agents that reliably finish real tasks, pedigree has become a priced asset.
The second is position. Consumer agents are widely expected to become a primary interface between people and online services, and whoever owns that relationship could sit above search, commerce and payments. Early user engagement, rather than revenue, is being treated as evidence of that position.
The third is relationships. Tech Funding News notes that Sequoia was an early backer of Sierra, Mr Shinn's former employer, and that Benchmark co-led the Series B. Insiders who keep buying at sharply higher prices send a signal to the market, but they also have an interest in marking their earlier positions upwards.
The competition is large, and free
The risks are not hard to identify. Meta launched its own assistant, Muse, on 8 September. It offers similar errand-running features, integrates with Instagram, Facebook and Marketplace, and has reached the top of the US app stores. Tech Funding News, citing CNN and Sensor Tower data, reports more than 2.5m downloads. Meta can fund a free consumer agent indefinitely from its advertising business; a start-up with 14 staff and no revenue cannot.
Trust is the second exposure. Instinct's original terms of service drew criticism for granting the company a "perpetual and irrevocable" licence over user data, and the policy has since been revised. A product that acts on a person's behalf, sending emails, spending money and making calls, depends on users granting broad access. One reported mistake can do outsized damage in a category where the value proposition is delegation.
The third is cost. Each completed task consumes model inference, telephony and compute. Until Instinct shows how it will charge, whether through subscriptions, commissions or advertising, every new user adds expense rather than income. The $1bn raised buys time, but it does not settle the question.
A wider pattern at the end of the quarter
Instinct was not the only large late-September financing. On Thursday 24 September, Island, the enterprise browser company, announced a $400m Series F at a $6.4bn valuation led by Evolution Equity Partners, with participation from investors including Sequoia, Coatue, Insight Partners and J.P. Morgan Growth Equity Partners. Island, which employs about 1,000 people, says its annual recurring revenue has doubled each fiscal year since it launched in 2022. The comparison is instructive: a company with years of revenue history and a large workforce commands a lower valuation than a consumer agent launched in August.
At the top of the market, Bloomberg reported on Tuesday 29 September that OpenAI was in talks to raise at least $30bn in a pre-IPO round at a valuation of roughly $1.4tn, according to TechCrunch's account of that report. The talks were not final. OpenAI raised $122bn in March at an $852bn valuation, and run-rate revenue reportedly reached $40bn in August. If the new round proceeds, it would act as a bridge to a listing now expected in 2027.
Taken together, these deals show capital concentrating at both ends of the AI market: vast sums for the leading model developers, and rapid, rising prices for a handful of application companies that investors believe could become category owners.
Our view
Instinct's round is less a verdict on one company than a measure of how venture pricing in AI agents now works. Valuations are being set by competitive tension among a few firms with very large funds, not by revenue, and the gap between rounds has shrunk from years to weeks. That rewards founders who can generate visible momentum and gives them leverage over terms. It also leaves later investors holding paper that has never been tested against a business model.
For founders, the lesson is that capital is available for agents that people actually use, but that the bar has moved to usage at scale and to a defensible distribution channel against Meta and the model developers. For investors, the question to watch is monetisation. The next disclosures that matter from Instinct are a pricing decision and retention data. Without them, a fourfold mark-up in 33 days is a statement of conviction, not of value, and the round that follows will reveal which of the two it was.
Sources
- TechCrunch: Viral AI agent Instinct raises $1B Series C at a $10B valuation
- OfficeChai: AI agent startup Instinct is now worth $10 billion with just 14 employees
- Tech Funding News: AI agent Instinct jumps to $10B valuation with $1B led by Sequoia, Benchmark and Coatue
- Verdict: Island raises $400m at $6.4bn valuation
- TechCrunch: OpenAI reportedly in talks to raise $30B round at $1.4T valuation
Filed under Venture Capital. Written by InvestTimes Editorial under our editorial standards. Commentary and analysis for information only, not investment advice.
