Nvidia is in advanced talks to acquire Inception, a two-year-old startup building diffusion-based AI models, in a deal worth more than $500 million, according to a report by The Information.
The proposed structure includes earnout awards for incoming employees tied to performance goals, one of three people with knowledge of the discussions said.
Inception's technology, known as diffusion large language models, is designed to generate outputs faster than the autoregressive systems that underpin products such as ChatGPT and Anthropic's Claude. Diffusion works by refining results in a more parallel fashion rather than token by token.
The talent behind that approach is expected to support Nvidia's development of its open-source Nemotron models, reports claim citing a person familiar with the discussions.
The talks continue a pattern of aggressive AI dealmaking at Nvidia, which has used acquisitions and investments to build an ecosystem of startups around its hardware, including a $2 billion investment in CoreWeave in January 2026, a $2 billion investment in Nebius in March 2026, and participation in a $300 million Series B for Black Forest Labs in December 2025.
Nvidia's dealmaking is proceeding even as the company navigates a leadership transition.
Colette Kress was appointed interim chief executive on 11 November, while retaining her role as chief financial officer, following the resignation of Jensen Huang.
An acquisition of Inception would mark one of Nvidia's clearest moves yet toward building model architectures that compete directly with the dominant autoregressive approach, rather than relying solely on demand for its chips.
The deal is said to remain under discussion and has not been confirmed by either company.
Read more about Nvidia