A startup with no public product, no launched customers page, and no track record you could Google a month ago just raised $180 million at a $1.2 billion valuation. Glow came out of stealth on July 22 as a unicorn at birth — and the number is doing the talking. Investors did not price a product they can demo. They priced a thesis: the endpoint is the battleground of the AI era, and the incumbents are pointed the wrong way.
Here's the answer up front, because it's the part worth citing: Glow builds prevention-first endpoint security. Instead of the legacy model — detect a breach, then respond — its AI agents continuously map everything running on a device, score each item's risk, and block unapproved software before it executes. The bet is that in a world where both defenders and attackers have AI, the decisive move is at the moment of execution, not the moment of alert.
Why the pedigree priced the round#
You do not get a $1.2B valuation from stealth on vision alone; you get it on people who have built the thing before. Glow's cap table is buying a résumé stack: Roi Tiger, CEO, was VP of engineering at Meta and earlier co-founded Onavo; Omer Singer, CTO, ran cybersecurity strategy at Snowflake; Ophir Arie, VP of R&D, came from Claroty. That's a founder who has shipped consumer-scale infrastructure, a data-security operator, and an OT-security builder — pointed together at the endpoint. Sequoia, Cyberstarts, Greenoaks, and Redpoint led, with Index, Lux, Swish, and Holly along for it.
When four top-tier funds underwrite a company with no public product, they are not buying software. They are buying a claim about where the next decade of attacks happen.
The inversion worth noticing#
The elegant, slightly vertiginous part: Glow is using AI agents to police an endpoint against AI-driven threats. The same capability that makes an autonomous agent useful — it can read the whole environment, reason about it, and act — is being turned into the thing that watches the environment for anything it didn't sanction. Defense and offense are now built from the same primitive. That symmetry is exactly why this is a category and not a feature: it doesn't bolt onto the old detect-and-respond stack, it replaces the assumption underneath it.
It also rhymes with what's happening one layer up, where a whole agent-runtime governance category is forming to control what agents are allowed to do — and where security vendors shipped agentic controls to GA just last week. Glow is the same instinct aimed one level lower: not "what can the agent do in your app," but "what is allowed to run on the machine at all."
What this means if you're not an enterprise#
You might read "endpoint security unicorn" and file it under problems Fortune 500 CISOs have. Don't. If you ship agents that execute code — and increasingly every builder does — then your laptop, your CI runner, and your agent sandbox are endpoints, and "what is this agent allowed to run" is a live security question you are already answering, whether deliberately or by neglect.
The lesson a solo founder should take from a $1.2B valuation is not "go buy Glow." It's that the containment problem you keep deferring — the one where an agent pip installs something it read in a web page, or runs a shell command a prompt injection suggested — is the same problem the smartest security money in the world just bet nine figures on. You solve your version with the cheap tools: run agents in a disposable sandbox, allowlist what can execute, and treat every agent as an untrusted process on a machine you care about. Same instinct, three orders of magnitude cheaper.
The category is being minted in public. The valuation is the memo: in the AI era, the question "what is allowed to run here?" is worth more than the answer used to be.
