Nvidia Explores Deal for Reflection AI to Bolster U.S. Open-Weight Models
Nvidia is in early discussions to acquire or expand its investment in Reflection AI, a U.S. developer of open-weight AI models valued at $25 billion. Nvidia has already put $800 million into the startup, and the talks could lead to a full purchase, a larger investment, or an acqui-hire arrangement. The move is seen as an effort to strengthen America's position against Chinese AI competitors such as DeepSeek, though no agreement has been reached.
Nvidia is reportedly in preliminary talks involving Reflection AI, a U.S. open-weight model developer last valued at $25 billion. Nvidia has already backed it with $800 million. Options include buying the startup, increasing its stake, or an acqui-hire that brings researchers and technology access. No deal is assured, and neither company commented.
Reflection AI was founded in 2024 by former Google DeepMind researchers Misha Laskin and Ioannis Antonoglou, who worked on Gemini. Its first public open-weight model, Beam, targets coding and AI-agent tasks and is claimed to match Z.ai’s GLM-5.2, though that claim lacks independent confirmation. Nvidia already offers its Nemotron open-weight models and used an acqui-hire structure in a December Groq agreement.
If completed, a larger Nvidia role in open-weight AI could affect developers, businesses, and researchers by shaping access to customizable models and compute. Open-weight tools may lower costs and give organizations more control over sensitive data, while also raising questions about concentration, security, and licensing. Chinese and U.S. AI competition could influence which models reach global users, but outcomes remain uncertain because talks are early and no agreement is guaranteed.