Labelbox Raises $100M+ Series E to Power Reinforcement Learning and Enterprise AI Agents

The infrastructure behind AI that learns on the job

Labelbox, the San Francisco AI platform powering reinforcement learning for frontier models and managed agents for enterprises, has raised more than $100 million in Series E financing, including equity and a credit facility. The round was led by Infinitum Partners with participation from Wipro Ventures, Andreessen Horowitz (a16z), B Capital, Tau Ventures, and Snowpoint Ventures.

The raise values the company at the center of AI's next shift: from models that answer questions to systems that learn from doing real work.

Two products, one bet

- Horizon: reinforcement learning environments, expert feedback, and evaluation infrastructure that help frontier AI labs train and improve models. Labelbox works with over 90% of leading U.S. AI labs. - Recursion Managed Agents: lets enterprises deploy fleets of AI agents inside existing business applications — agents that coordinate across tools and improve with every run.

The business fundamentals are striking: more than 400 customers, profitability reached in 2026, and revenue growing more than fivefold year over year.

"Every company will soon run on AI systems that learn from their own work," said CEO Manu Sharma. The funding accelerates R&D across the full stack — from sandbox infrastructure and the agent harness to reinforcement learning post-training.

What this means for your business

1. The agent layer is where the money is going. When a16z and Wipro Ventures back an agent-harness platform at this scale, the "AI pilots" conversation is over. The winners are moving to managed agent fleets with feedback loops. Audit where your company stands.

2. Reinforcement learning is the differentiator. Generic fine-tuning is commoditizing; RL environments that teach agents from real task outcomes are the scarce resource. If you're building custom agents, prioritize feedback-loop infrastructure over bigger models.

3. "Improves with every run" is the new vendor promise. Evaluate AI vendors on this metric: does the system get better from your usage, or does it stay static? Recursion's pitch makes this table stakes.

4. Profitability in AI infra is possible. Labelbox's profitable, 5x-growth story is a datapoint for founders: pick a real bottleneck (evaluation and RL infra), serve the labs and the Fortune 500, and the unit economics work.

The takeaway for leaders: the companies winning the AI buildout aren't just buying compute — they're buying the machinery that makes AI systems teachable.