The pitch sounds almost too straightforward for a late-stage venture round: send AI agents to browse the actual internet—the messy, dynamic, endlessly shifting web that confounds most automated systems—and return something enterprises can actually trust.
Yet that's precisely the promise that just netted Nimble, a New York-based data infrastructure startup, $47 million in Series B funding. Norwest Venture Partners led the round, announced February 24, bringing the company's total raise to $75 million since its 2021 founding. Databricks Ventures joined as a new strategic investor, alongside a roster of returning backers that includes Target Global, Square Peg, and Slow Ventures.
For a company that's been operating somewhat quietly—fewer than 120 employees split between Manhattan and an R&D center in Israel—the valuation whispers are striking. Israeli business publication Globes pegs Nimble's post-money value somewhere between $400 million and $450 million, though the company declined to confirm figures. What is confirmed: more than 100 enterprise customers, including multiple Fortune 10 companies, are now routing real-time web data through Nimble's platform before feeding it to AI systems that make pricing, compliance, and competitive intelligence decisions.
"Most production AI fails aren't because the models are not good enough—it's because of a data failure," CEO Uri Knorovich told TechCrunch in a recent interview. It's a diagnosis that resonates louder as enterprises race to deploy agents that need more than static training data—they need live context from a web that wasn't built to be machine-readable.
The Deceptively Hard Problem
What Nimble built isn't technically novel in isolation. Web scraping has existed for decades. What co-founders Knorovich and Menachem Salinas recognized was that enterprise AI demands something web scraping has never reliably delivered: verifiable, structured, governed data that can flow into Snowflake or Databricks without setting off alarm bells in compliance departments.
Their approach involves deploying what the company calls "Web Search Agents"—essentially automated browsers that navigate sites the way humans do, handling JavaScript-heavy pages and dynamic content that traditional scrapers choke on. But the harder engineering challenge, according to the company, comes after extraction: validation layers that deduplicate records, flag anomalies, assign confidence scores, mask personally identifiable information, and maintain lineage tracking.
In other words, turning the chaos of the live web into something an enterprise data warehouse won't reject.
The platform integrates directly with the usual suspects—Databricks, Snowflake, BigQuery, AWS—positioning itself less as a standalone tool and more as infrastructure in the emerging stack for what venture capitalists have started calling "agentic AI." Databricks Ventures' participation underscores that positioning. The two companies already collaborate on e-commerce use cases, where AI agents need real-time product pricing and availability data to make split-second recommendations.
Why Norwest Wrote the Check
Assaf Harel, the Norwest partner who took a board seat with the investment, framed the thesis in terms that echo across dozens of AI infrastructure pitches lately: trusted data as the bottleneck. "Trusted live web data is increasingly becoming a prerequisite for AI agents performing critical business decisions," he said in a statement that accompanied the funding announcement.
Norwest published a detailed diligence memo surfacing specific customer wins—a global food-delivery marketplace using Nimble for dynamic pricing intelligence, a Fortune 50 beverage company monitoring brand presence across thousands of retail sites. The use cases cluster around competitive intelligence (tracking rival pricing in real-time), digital shelf analytics (ensuring product listings appear correctly), brand monitoring, KYC processes, and what the company calls "deep search for AI agents."
Israeli media reports have circulated a customer list that includes LG, Deloitte, Uber, L'Oréal, Coca-Cola, and Tripadvisor, though Nimble doesn't display those logos publicly—a caution that may reflect enterprise sensitivities around disclosing reliance on external data vendors for competitive work.
A Crowded, Complicated Market
Nimble enters a space where competition exists but fragmentation remains high. Apify, Bright Data, and Oxylabs all offer web data extraction at scale. What Nimble claims as differentiation hinges on enterprise-grade governance—the schema enforcement, reproducibility guarantees, and audit trails that IT organizations demand before routing external data into production AI systems.
Whether that positioning justifies a valuation nearing half a billion dollars depends partly on timing. If agentic AI becomes the next wave—agents autonomously booking travel, negotiating supplier contracts, managing supply chains—then real-time web context stops being a nice-to-have and becomes foundational. If that wave stalls or takes longer to materialize than venture timelines allow, Nimble's pitch becomes harder.
The company maintains it's been working with Deloitte since 2023, a partnership that presumably lends credibility in enterprise sales cycles. And Knorovich's framing of AI failures as data failures, rather than model failures, aligns with a broader reckoning in the industry: throwing better models at bad data rarely fixes anything.
What Happens Next
The Series B capital will fund R&D into multi-agent web search—presumably systems where multiple AI agents coordinate to gather, cross-reference, and validate information from disparate sources. Nimble is also building what it describes as a "governed data layer," which sounds like middleware designed to sit between raw web extraction and enterprise systems, handling the messy transformation work that most companies would rather not build internally.
Platform development and partner ecosystem expansion round out the roadmap. The integrations with Microsoft, AWS, Databricks, and Snowflake suggest Nimble is angling to be a standard component in modern data stacks, the way Fivetran became synonymous with data pipeline integration or dbt with transformation.
Whether the market Nimble is betting on fully materializes remains to be seen. But at $75 million raised and a rumored valuation that would make early employees very comfortable, the company has bought itself runway to find out—assuming the web data it's structuring proves as mission-critical to AI agents as its investors believe.
For now, perhaps the strongest signal is who's writing checks: Databricks Ventures doesn't typically invest in companies unless there's a clear strategic play. That suggests at least one major platform provider sees live, governed web data as a layer worth owning a piece of. Whether others follow will tell us a lot about where enterprise AI infrastructure is actually headed.
