Labelbox Series E Funding Raises Over $100 Million for Reinforcement Learning and Enterprise AI Agents

Labelbox, a San Francisco-based AI infrastructure company, announced on October 14, 2026, that it raised more than $100 million in Series E financing through equity and a credit facility at a $1.25 billion pre-money valuation. The round brings the company’s total funding to approximately $289 million. Proceeds will support development of the Horizon platform and the Recursion Managed Agents product for reinforcement learning and autonomous enterprise workflows.

Labelbox Series E Funding Round Details and Investor Participation

The Series E round was led by Infinitum Partners with participation from Wipro Ventures, Andreessen Horowitz, B Capital, Tau Ventures, Snowpoint Ventures, and additional investors. The exact split between equity and credit remains undisclosed. Labelbox reported 400 or more customers and stated that it works with more than 90 percent of leading U.S. AI labs. The company also reached profitability in 2026 while recording year-over-year revenue growth above five times.

Infinitum’s involvement provides access to compute resources through its ties to GroqCloud, which the company said will help scale training and evaluation workloads. Labelbox currently operates more than 50 internal agents across its own functions and plans to expand engineering and research teams with the new capital. CEO Manu Sharma stated that the funding will allow the company to push AI systems that learn from their own work further into production environments.

Readers who build or deploy frontier models can expect expanded capacity for simulated training environments and rubric-based agent evaluation. Enterprises evaluating autonomous agents for tools such as GitHub, Slack, and Salesforce now have an additional vendor with demonstrated internal usage and profitability metrics.

Horizon Platform Capabilities and Recursion Managed Agents

The Horizon platform supplies reinforcement learning environments, expert feedback loops, and evaluation infrastructure that frontier labs use to train models on multi-step tasks such as coding and research. It includes simulated WorldSim environments that allow models to interact with complex, dynamic settings before deployment. The platform also maintains an expert network of more than four million people who provide graded feedback on model outputs.

Recursion Managed Agents, opened to enterprise customers earlier in October 2026, lets organizations deploy fleets of autonomous agents inside existing applications. Agents trigger on schedules, alerts, or tickets, coordinate across tools, receive rubric grades, and improve through memory of prior runs. Labelbox reports that these agents operate without requiring separate infrastructure stacks, which reduces integration overhead for teams already using hyperscaler or startup tooling.

Compared with its origins as a data-labeling service founded in 2018, Labelbox now positions itself as a full-stack provider of sandboxes, agent harnesses, and post-training reinforcement learning. Companies that previously relied on separate vendors for labeling, simulation, and agent orchestration can consolidate those functions. The shift aligns with demand from both AI labs needing scalable evaluation and Fortune 500 teams seeking measurable agent performance in production applications.

Company Background and Planned Use of Funds

Labelbox was founded in 2018 by Manu Sharma and others. It began with data-labeling tools and later expanded into reinforcement learning environments and managed agent platforms. Current customers span frontier labs, hyperscalers, startups, and large enterprises. The company emphasizes full-stack control that includes secure sandboxes and an expert feedback network to support both training and ongoing agent improvement.

Funds from the Series E round will accelerate research and engineering on Horizon and Recursion, expand team size, and increase compute access. No specific external product milestones were announced alongside the funding. The company continues to serve customers that require both high-volume evaluation infrastructure and production-grade agent deployments that improve over time through recorded memory and graded feedback.