Warburg Pincus writes a $130 million check into the web-scraping infrastructure firm AI labs quietly depend on, even as Reddit brands it a data "bank robber."
Every large language model needs a constant diet of fresh internet data, and increasingly the labs building those models don’t want to do the scraping themselves. That’s the workflow problem Oxylabs UAB has spent a decade quietly solving — and on July 9, 2026, private equity finally noticed, according to SiliconANGLE: Big Data.
the numbers
Warburg Pincus LLC put $130 million into the Lithuanian company, its first outside capital since founding in 2015, pushing Oxylabs’ valuation to $3.6 billion, as SiliconANGLE reported. The money comes out of Warburg’s $4 billion Capital Solutions Founders Fund, which has also backed NordVPN maker Nord Security UAB — giving the firm stakes in what SiliconANGLE called Lithuania’s two biggest technology companies. By the company’s own account, it counts more than 350,000 customers and has grown annual recurring revenue past $350 million, figures that are striking, if hard to independently verify given the company was bootstrapped and has not previously had to disclose financials to outside investors.
from proxies to “web intelligence”
Oxylabs started as a premium web-proxy provider, routing requests through different URLs to scrape prices, listings and security data without tripping IP bans. It now describes itself as a “web intelligence” platform running billions of requests a day, drawing on a pool of more than 175 million what it calls “ethically sourced” consumer device IPs to dodge CAPTCHAs and geo-blocks, per SiliconANGLE. Its product line has grown accordingly, including a fully managed Web Scraper API, an AI-native Web Unblocker, and a stealthy Headless Browser aimed at letting autonomous AI agents extract data from dynamic websites — infrastructure the company argues will matter as much for agentic AI as for model training. “They need a live feed of what is out there, not a stale index,” said co-founder and CEO Vytautas Savickas, adding: “The next generation of AI won’t be powered by static indexes.”
the reddit problem
The funding lands squarely inside an unresolved legal mess. OpenAI Group PBC, Anthropic PBC and other frontier labs have faced repeated lawsuits over scraping practices, and the vendors who enable that scraping haven’t escaped scrutiny either. Reddit Inc. sued Perplexity AI Inc. and three data-scraping providers — including Oxylabs — last year, according to SiliconANGLE, comparing them to “bank robbers,” saying that they will do almost anything to access its data — except pay for it. Oxylabs’ response has been to lean into an ethics framing: compliance with the European Union’s strict data-collection rules, and founding membership in the Ethical Web Data Collection Initiative, an industry consortium sitting under the Internet Infrastructure Coalition that is trying to build standards for data aggregators. Whether that positioning satisfies plaintiffs, regulators, or the publishers whose content underlies the whole business is, notably, not something SiliconANGLE’s report resolves.
what to watch
Warburg Pincus Principal Allison Ross called Oxylabs “a leader in web intelligence through its sophisticated technology and expansive network,” the kind of line private equity says about most things it buys into. The more interesting test is whether a $3.6 billion valuation on a company’s first-ever funding round holds up once outside investors start asking the questions bootstrapped founders never had to answer — about customer concentration, legal exposure from the Reddit suit, and whether “ethically sourced” scraping infrastructure can keep growing at AI-training speed without inviting the next lawsuit. For now, Oxylabs joins a small class of scraping and proxy infrastructure vendors — the unglamorous plumbing beneath the AI agent boom — suddenly commanding unicorn money, and unicorn scrutiny to match.
It famously compared Oxylabs and the others to "bank robbers," saying that they will do almost anything to access its data – except pay for it.
