The $3.6 Billion Company Reading the Web for the AI Agents (and You've Never Heard of It)

There is a company you have never heard of that knows, minute by minute, what the cheapest flight from New York to Chicago costs, what a GPU sells for in Jakarta, and what thousands of retailers are charging for the same pair of sneakers. Until this week, that company had no group name, no holding structure, and almost no public profile. As of Tuesday, it has all three, plus a $3.6 billion valuation.

The company is Source, a new umbrella group launched October 13 in Vilnius, Lithuania, bringing together five web-intelligence businesses led by Oxylabs. The move follows a $130 million investment from Warburg Pincus that valued the group at $3.6 billion, making it Lithuania's sixth unicorn and one of the highest-valued companies in the country's history. And the timing tells you where the money is going: demand across the group's companies has quadrupled in the past year, driven by AI agents.

What "web intelligence" actually means

Strip away the branding and web intelligence is a straightforward business: reading the internet at machine scale. Source's companies build the plumbing for large-scale data collection, proxies that route millions of requests through rotating IP addresses so collection does not get blocked, and scrapers that extract structured information from websites. Customers range from startups and Fortune 500 companies to governments and nonprofits, who use the data for everything from market intelligence and trading decisions to cybersecurity defense and disinformation research.

The scale is hard to overstate. The group says about 350,000 technology teams use its products, generating an annual recurring revenue run rate of $350 million. Its data-processing capacity grew roughly tenfold in the first half of 2026 alone. The Source group employs 800 people worldwide, including more than 350 developers and engineers, and holds over 180 patents. Half of its revenue comes from the United States, where it plans to open a San Francisco office.

Why AI agents changed the math

For a decade, Oxylabs grew quietly by selling data collection to businesses. Then AI agents arrived, and everything changed. An agent that books travel, comparison-shops, monitors competitors, or researches a market does not query a database: it goes and reads the live web, over and over, at a volume no human could match. "The next generation of AI won't be powered by static indexes that only capture yesterday's internet," CEO Vytautas Savickas has said.

The company claims eight of the ten largest language-model laboratories use its infrastructure. It declined to name them, which is standard in an industry where client lists are shielded by confidentiality agreements. What is not disputed is the direction of travel: Source says demand from next-generation search, AI agents, and frontier AI products drove a four-fold increase in demand across its companies in the last year alone.

AI agents don't read the internet the way you do. They read it constantly, in bulk, and they need someone to keep the pages turning.

The open web is closing just as AI needs it most

Here is the paradox sitting underneath Source's launch. Just as AI agents became the internet's most voracious readers, the internet started locking its doors. Publishers, platforms, and cloud providers have spent the last two years putting up paywalls, bot-detection walls, and licensing gates between their content and automated collection. The company's own OxyCon 2026 conference carried the warning in its headline: "Experts Warn the Open Web Is Closing."

This is the fight Source is positioning itself to have, in public. The new group's stated mission centers on "responsible access to public web data," and Savickas frames it as core infrastructure: "For these systems to work, responsible access to public web data is no longer just a technical tool. It is the core infrastructure for the next wave of progress." The pitch is that agents need the live web the way factories needed electricity: as a utility, governed by rules, not blocked by them. Bloomberg's headline on the Warburg deal put it more bluntly: "AI Turns Little-Known Lithuanian Unicorn Into Data Powerhouse."

What Source wants to be next

Source is not content to be plumbing. According to reporting on the launch, the group is developing a web API backed by its own internet index, a service designed to combine its data-collection tools into one product for AI agents. It is in early beta and expected to launch in the coming weeks. That puts it in direct competition with Israel's Bright Data, the industry's other heavyweight.

The bigger bet is on the shape of the internet to come. If agents really do become the web's primary readers, then the companies that control how agents read it will have leverage that search engines never had. Source, bootstrapped for a decade before taking its first outside dollar in 2026, is now a $3.6 billion wager that the age of agentic AI will be built on the open web: scraped, proxied, indexed, and sold by the millisecond.

Whether the web stays open enough for that wager to pay off is the question everyone, including Source, is watching.