Island raises $400 million at a $6.4 billion valuation to police AI agents inside the browser

Island, the enterprise-browser startup that positions itself as the control point for how AI agents behave inside a company, has raised $400 million at a $6.4 billion valuation in a Series F led by Evolution Equity Partners, the company announced Sept. 24. Sequoia, Coatue, Cyberstarts, Insight Partners, J.P. Morgan Growth Equity Partners, Georgian, G Squared, Squarepoint and cybersecurity investor Dmitri Alperovitch personally all put in money alongside Evolution, pushing Island's total raised past $900 million since its 2022 launch out of stealth. The company now employs roughly 1,000 people.

The pitch is that the browser -- not the API layer, not the network, not the endpoint -- is the one place a company can actually see and govern what both its human employees and its AI agents are doing with corporate data, because it is the one piece of software both of them are forced to use. "The enterprise browser has become the most critical infrastructure for the modern workforce," the company's messaging around the round argues, and now, per its own framing, "the browser is the only place where you can govern the intersection of humans, agents, and data at scale." CEO Mike Fey put the same idea more simply in Island's own release: "Island created the first Enterprise Browser to protect the primary place people work."

The valuation math, and what's actually disclosed

The new number is not Island's first big re-rating this cycle. The company was valued at $3 billion after a $175 million Series D in 2024, then at $4.8 billion after a Series E in March 2025, and is now at $6.4 billion seventeen months later -- a path of roughly doubling, then re-rating another third higher, in under two years. Island says its annual recurring revenue has doubled every year since its 2022 launch, which is the number underwriting all three jumps. It is also the only one of the underlying figures the company has not put an actual dollar amount on.

"Doubled every year" is a growth rate, not a figure -- and a rate compounding on an undisclosed base can justify almost any valuation a term sheet asks for, which is exactly why it's the kind of number arithmetic-skeptic readers should sit with rather than nod past. Without a disclosed ARR figure in dollars, there is no way to check what revenue multiple $6.4 billion actually implies, whether that multiple is in line with comparable enterprise-security software, or whether "doubled" describes the same growth rate in 2023, 2024 and 2025 or a rate that's already decelerating and being described the same way regardless.

Nobody has agreed on where this control layer should live

Island's $6.4 billion is also a bet that the enterprise-browser approach wins a category that, on the evidence available, has not settled on one architecture. Reporting on the round lays out at least three competing bets on where AI-agent governance should actually sit: Dataiku and SAP build it into API-level monitoring, watching what an agent calls rather than what it clicks; NiCE Cognigy builds it into the routing layer that sits between an agent and the systems it's allowed to reach; and newer entrants -- Meta's Sentinel and a startup called GrokBot -- are betting on kernel-level enforcement and persistent cloud virtual machines respectively, control points further from the browser than Island's model and closer to the operating system. None of these has emerged as the standard; Island's valuation prices in a bet that browser-native control wins that argument, not evidence that it already has. The company frames the competition differently -- as validation that the whole category is real, rather than as a fight it might lose -- but a market with four structurally different answers to "where does agent governance actually live" is a market still being defined by argument, not by customer consensus.

Why now

The round lands the same week two of the Trump administration's own cabinet officials were publicly arguing that a company, not its AI agents, bears legal responsibility when those agents misbehave -- a liability framing FTC Chair Andrew Ferguson and Treasury Secretary Scott Bessent each stated on the record days apart. If that framing holds, a product that lets an employer prove, after the fact, exactly what an agent did inside a browser session -- and stop it before it does something costly -- stops being a nice-to-have compliance tool and starts being the thing a general counsel asks for by name. Island's own recognitions this year (a Frost & Sullivan award, a CNBC Disruptor 50 slot, a Gartner Peer Insights customer-choice badge for Security Service Edge) are the kind of validation that moves an enterprise buyer, even though none of them substitute for a disclosed revenue number. Enterprise buyers adopting agentic AI tools this year have had little choice but to bolt governance on after the fact, since most agent platforms shipped the capability well before any standard way to audit what an agent did with it; Island's bet is that whoever owns that audit trail first captures the budget line before a slower-moving standards process defines what it's supposed to look like.

What's actually established, set against what's still Island's own account: a real $400 million landed in the company's bank account from name-brand investors who did their own diligence and were willing to price the company at $6.4 billion. What's not established is whether that price reflects a revenue multiple any outside analyst could independently check, or whether it reflects investors pricing a category -- agent governance -- that is moving fast enough that being first and best-capitalized matters more than the arithmetic working out on paper today.