On September 28, 2026, Instinct, a U.S. company developing personal AI agents, announced it had raised an additional $1 billion from Sequoia Capital, Benchmark Capital, and Coatue, bringing its valuation to $10 billion. The product is still in early access, but using its own phone number and computer, it handles tasks on behalf of users—from arranging travel to buying groceries to canceling subscriptions.
Why would a company not yet in full-scale release command such a large valuation? Looking at the background, the money appears to be flowing in not because of current revenue, but because of the company's potential to become a gateway connecting consumer requests to payments, its ability to secure the compute power needed to meet demand months ahead, and its capacity to actually carry out transactions on behalf of users.
A 4x Jump in 33 Days, and the Future Baked Into a $10 Billion Valuation
What stands out about this funding round isn't just the amount—it's the speed. In the Series B that TechCrunch reported on August 26, Instinct raised $250 million at a $2.5 billion valuation. Just 33 days later, both the amount raised in a single round and the valuation had quadrupled: from $250 million to $1 billion, and from $2.5 billion to $10 billion.
However, it would be a mistake to interpret this as meaning the company's underlying value objectively quadrupled in 33 days. The $2.5 billion and $10 billion figures come from two separate funding rounds, and public materials don't even specify whether these are pre-money or post-money valuations. It's also unclear what percentage of equity the new investors acquired, or how much existing shareholders' stakes were diluted. What can be confirmed from public information is simply that investors committed larger sums, in a short span of time, on the premise of a higher company valuation.
What investors appear to be valuing highly is Instinct's potential to become a new gateway through which consumers access services in the future. While typical conversational AI stops at answering questions, Instinct uses existing websites, phone calls, and users' own accounts to actually carry out actions. It doesn't just plan a travel itinerary—it books it. It doesn't just search for products—it buys them. If Instinct can capture the point of contact right before a user completes a transaction, it could position itself to funnel customers to travel agencies, retailers, and various service providers.
That said, this remains a future possibility rather than an established business track record. The company cites recent features such as "Concierge," which handles bookings that require phone calls; "Trusted Person Network," through which users' agents coordinate schedules and share files with each other; and location sharing via iMessage. Each of these expands the range of actions the agent can perform on a user's behalf, but unless usage becomes habitual and the company can earn revenue from its business partners, there's no guarantee that the $10 billion valuation will translate into actual business value.
$1 Billion in Transaction Volume, and Revenue That Remains Undisclosed
In an interview published on September 28, Instinct founder Noah Shinn said the user base is growing at a daily rate of 10-11%, without any advertising spend. He explained that growth accelerated after starting with roughly 200 friends and family as testers, once usage examples began spreading among a user base numbering in the thousands.
However, he did not disclose the current number of users, nor over what period this daily growth rate was measured. There's no reason to assume this growth rate can be sustained over the long term.
Another figure Shinn cited is gross transaction volume approaching $1 billion on an annualized basis, with travel accounting for roughly half of that. This represents the total value of goods and services users purchase through Instinct—not Instinct's own revenue. The company's public materials do not disclose user counts, revenue, gross margin, the pre- or post-money breakdown of its valuations, or the extent of equity dilution.
Turning gross transaction volume into the company's own revenue depends on how much of a cut it can take. Shinn has described a vision in which, rather than charging users steep monthly fees or running ads, the company collects transaction fees from the merchants and service providers involved. But he hasn't disclosed specific fee rates, which partners are already under contract, or how much fee revenue the company currently generates.
The $1 billion transaction volume figure suggests a meaningful scale of purchasing activity is flowing through Instinct, but it doesn't reveal how much the company can actually earn from those transactions.
This distinction also matters for assessing how much negotiating leverage Instinct can wield with its partners. If the company can aggregate a large volume of user requests and complete bookings and purchases at a high success rate, it will be in a stronger position to demand referral fees from partners.
Conversely, if travel agencies and retailers refuse to pay fees—or restrict access by AI agents—in order to protect their own sales channels, monetization may not advance even as transaction volume grows. The current valuation likely bakes in the possibility that Instinct will gain this kind of negotiating leverage in the future.
Compute Power to Support Demand Months Ahead
One clue to how the $1 billion will be spent comes from Shinn's statement that he spends about 40% of his time securing compute resources. As the user base grows daily, so does the amount of compute needed to run the AI agents. But the necessary compute capacity can't always be secured the moment it's needed—it has to be reserved months in advance based on demand forecasts.
An AI agent's job doesn't end once it generates a single response. It breaks a request down into multiple tasks, reads websites, operates on-screen interfaces, makes phone calls when necessary, and tries alternative approaches if something fails. Because completing a request requires running the model repeatedly, inference costs balloon as the user base grows.
Moreover, if Instinct wants to grow while keeping costs low for users, it has to absorb the compute expenses itself until it can generate sufficient revenue.
Shinn has said he wants to avoid a scenario where a shortage of compute forces the company to charge users something on the order of $100 a month. This massive funding round creates room to secure future compute capacity in advance, keep usage fees low, and expand the range of transactions the service can handle.
If compute supply fails to keep pace with rising usage, growth opportunities are lost. But securing too much capacity too far in advance means paying for compute that goes unused. Ample funding serves as a buffer to absorb the timing gap between demand and compute procurement.
Still, a large funding round alone doesn't guarantee improved profitability. The company hasn't disclosed per-user inference costs, the actual compute capacity it has secured, or the terms of its supply contracts.
Unless transaction-fee revenue exceeds inference costs and the cost of compensating for errors, the business could end up burning more cash the more users it gains. Gross margin—not just user growth rate—will be crucial in judging whether the $10 billion valuation is justified, but that figure has yet to be made public.
The Trade-off Between Convenience and Expanding Authority
A look at Instinct's terms of service reveals that delivering this convenience requires granting the AI agent fairly broad authority.
Instinct can retrieve, copy, collect, and index data from the services a user connects to it. It can also purchase products or enter into contracts on a user's behalf, and in doing so may agree to third-party services' terms of use. Even after a user disconnects a service, previously indexed data isn't necessarily deleted automatically—a separate deletion process is required.
According to the company's privacy policy, depending on which connected services a user authorizes, Instinct may handle emails, private messages, voice data, location information, payment and authentication credentials, and even health-related information.
This is a very different risk profile from an AI that simply gives a wrong answer. A mistaken purchase or communication can directly cause financial loss or damage relationships. There's also a risk that malicious instructions embedded on the web could hijack the agent into taking actions the user never intended.
The terms of service explicitly acknowledge that the AI's outputs and actions may contain errors, and that some actions it takes cannot be undone. The terms go on to state that users bear responsibility for the consequences—including financial, contractual, legal, and reputational ones—and the company does not guarantee that its safety measures will always prevent unintended actions.
In other words, even as the AI agent takes on more tasks on a user's behalf, the design still leaves much of the burden for failures squarely on the user.
As safety measures, Instinct cites a sandboxed environment that isolates its workspace, short-lived authentication credentials, signed tool execution, and mechanisms that detect errors before carrying out a response or action.
However, the company hasn't published detection rates, false-positive rates, penetration test results, or third-party audit findings for these mechanisms. Having safety measures in place is one thing; how much they actually reduce incidents in real-world use is another matter entirely.
In its own hands-on trial, The Atlantic was able to have Instinct select a book under $25 and arrange in-store pickup, search for lodging, and request quotes from cleaning services.
On the other hand, the same article also documented cases where the agent canceled a flight against the user's wishes, causing a loss of more than $200; booked a restaurant with a $200 cancellation fee without authorization; and triggered an account suspension after repeatedly hitting a booking site.
These are individual incidents and user-reported cases, not statistics representing the service's overall failure rate. Still, they demonstrate that—unlike a simple wrong answer in a chat—there already exists a pathway by which an AI's misjudgment can lead to real financial loss.
Whether Instinct can build a business that justifies its $10 billion valuation won't be determined by user growth alone. The company needs to raise both the completion rate for requested transactions and user retention, cover its inference costs through fees earned from partners, and simultaneously keep the rate of erroneous actions—and the cost of compensating for them—low.
As figures such as revenue, transaction fee rates, per-user compute costs, incident and compensation records, and third-party safety audits become available going forward, it will become possible to judge more concretely whether the future that investors are betting heavily on can be turned into a sustainable business that users can trust.