Nvidia is considering expanding its investment in AI developer Reflection AI. Providing additional computing power or acquiring employees and technology are also possibilities. A full acquisition has not been ruled out.
Bloomberg reports this based on information from a source familiar with the matter. Discussions are still in the early stages, and no concrete agreements have been reached. According to the Financial Times, one option being discussed is a so-called “acqui-hire,” in which Nvidia would take on staff and obtain a license for the technology.
This interest aligns with Nvidia’s strategy to increase its influence on the development of AI models. In doing so, the chipmaker is explicitly focusing on American alternatives to Chinese open-source models.
Reflection aims to compete with Chinese AI
Reflection AI unveiled its first model, Beam, in early October. The startup, founded by two former researchers at Google DeepMind, focuses on programming tasks and applications involving AI agents.
In open-weight models, the model weights are publicly available, allowing developers to use and customize the technology independently. Reflection claims that Beam delivers performance comparable to that of GLM-5.2, a model developed by the Chinese company Z.AI, on certain benchmarks, at a relatively low cost.
These ambitions require significant investment. According to the Financial Times, Reflection has been in talks regarding a $2.5 billion funding round, at a valuation of $25 billion. In addition, the company has signed billion-dollar deals with SpaceX and cloud provider Nebius to secure sufficient computing power.
Nvidia expands its influence
Nvidia has closed several major deals in recent months. In September, the company reached a $13 billion agreement to acquire Hugging Face. It also signed a licensing deal worth approximately $20 billion with chip developer Groq.
Closer collaboration with Reflection would give Nvidia access to expertise in the field of open-weight models. Reflection, in turn, could benefit from Nvidia’s financial resources and AI chips.
Geopolitical considerations also play a role. Nvidia CEO Jensen Huang (photo) has long advocated for American open-weight models as an alternative to Chinese technology. Those Chinese models are often cheaper to use than American alternatives, while their performance is becoming increasingly comparable.
The U.S. government supports this development. President Donald Trump’s AI action plan from July 2025 explicitly calls for U.S. open models that can serve as an international standard.