San Francisco-based Labelbox has raised more than $100 million in Series E financing, including equity and a credit facility, at a $1.25 billion pre-money valuation.
The round was led by Infinitum Partners, with participation from Wipro Ventures, Andreessen Horowitz, B Capital, Tau Ventures and Snowpoint Ventures.
Founded in 2018 by Manu Sharma, Brian Rieger and Dan Rasmuson, Labelbox originally built software for labeling and managing machine-learning training data. The company has since expanded into reinforcement learning infrastructure and managed enterprise AI agents.
Its business now operates across two major products.
Horizon provides reinforcement learning environments for frontier AI laboratories. Instead of relying only on static datasets, models can attempt realistic tasks, receive structured feedback and improve through repeated training. The platform creates simulated work environments containing code, documents, spreadsheets, dashboards and other enterprise data.
Recursion Managed Agents targets enterprises that want AI systems to perform work inside existing business applications. These agents can execute recurring tasks, evaluate their performance and retain information from previous runs.
Labelbox says it serves more than 400 customers and works with over 90% of leading US AI labs, according to its own classification. It also reports reaching profitability in 2026 while growing revenue more than fivefold year over year.
Internally, the company says it operates more than 50 agents across functions including engineering, research, sales and finance. Its broader infrastructure includes reinforcement learning systems, sandbox environments and an expert network of more than four million people.
The funding will accelerate development of Horizon and Recursion, expand engineering and research, and increase access to computing infrastructure.
The strategic shift is significant. Labelbox is moving from helping humans prepare training data toward supplying environments in which AI systems can learn from performing work.
Competition remains substantial. Frontier laboratories can develop training infrastructure internally, while enterprise agent platforms face established cloud providers and emerging startups.
Labelbox's position will depend on whether its environments and evaluation systems improve model performance enough for customers to continue buying them externally, and whether its managed agents can deliver reliable business outcomes beyond internal demonstrations.