AI training environment company Fleet's revenue has increased 60 times over several months, currently in talks for a $750 million valuation funding.

ME News Report, April 14 (UTC+8), according to 1M AI News monitoring, reinforcement learning training environment startup Fleet is negotiating a new round of financing, at least $50 million, with a post-investment valuation of about $750 million, more than seven times the valuation of less than $100 million at seed stage. Bain Capital Ventures, a subsidiary of Bain Capital, is negotiating to lead the investment, with Sequoia Capital, Menlo Ventures, and SV Angel planning to co-invest. Fleet was founded in 2024, with its core business building reinforcement learning training environments (RL gym) for AI labs: developing replicas of popular applications like Salesforce and Excel, enabling models to learn how to operate these software programs. The company initially focused on training data for browser and computer operation scenarios, later expanding to more types of applications. Founder Nic Ouporov was previously the lead engineer at no-code automation startup Respell, which was acquired by Salesforce in January 2024. The key driver for this funding round is explosive revenue growth: Fleet’s annualized revenue increased from about $1 million at the end of last year to over $60 million in recent weeks (calculated based on the most recent quarterly revenue multiplied by four). However, the sustainability of this growth is questionable. In recent months, AI labs’ demand for training data has surged dramatically, with sales of multiple data and RL environment startups soaring, reflecting a rising tide lifting all boats. Over time, labs may concentrate their data spending on a few suppliers, potentially leading to the elimination of some startups. Fleet’s competitors include not only other RL environment providers but also traditional data annotation companies like Mercor, Surge, and Turing, which are layering RL environment services on top of their existing businesses. (Source: BlockBeats)