NVIDIA Said to Be in Talks to Acquire Inception for at Least $500 Million to Boost Development of Its Nemotron Open-Source Model

Deep News07:20

People familiar with the matter said NVIDIA (NASDAQ: NVDA) is in advanced talks to acquire the startup Inception. Inception was founded only two years ago, and the artificial intelligence technology it has developed is designed to run faster than the AI used by products such as ChatGPT and Claude.

One of the people said the proposed deal amount is at least $500 million, and it also includes "earnout awards" for new employees — that is, additional rewards paid depending on whether they meet specific targets. These talks indicate that even though NVIDIA (NASDAQ: NVDA) is facing a leadership crisis after Jensen Huang stepped down as CEO, the company's efforts in acquiring AI startups, as well as its push to develop its own AI technology to counterbalance Anthropic, are still continuing.

The people said Inception's so-called "diffusion large language models" and the technical talent behind them are expected to help NVIDIA (NASDAQ: NVDA) develop its Nemotron open-source model. Talks between NVIDIA (NASDAQ: NVDA) and San Francisco Bay Area-based Inception began before Huang's departure; previously, Microsoft had also shown interest in reaching a similar deal. However, given that NVIDIA (NASDAQ: NVDA) has already invested tens of billions of dollars in the Nemotron model, there is currently little indication that the company would choose to walk away. These investments include spending billions of dollars to acquire, or obtain in a deal resembling an acquisition, related technology and AI researchers, with the aim of expanding its Nemotron-related team, which currently has hundreds of employees.

NVIDIA (NASDAQ: NVDA)'s Nemotron model is intended to provide enterprises and governments with an alternative so they do not have to rely on closed-source AI provided by companies such as Anthropic and OpenAI; at the same time, these models also help advance NVIDIA (NASDAQ: NVDA)'s work in robotics AI. Nemotron users include Booz Allen, Palantir, and multiple U.S. government agencies. Compared with closed-source models, open-source or open-weight models are easier for various organizations to customize according to their own specific needs, but this process requires a certain level of technical expertise.

NVIDIA (NASDAQ: NVDA) is working to compete with Anthropic and OpenAI; these two leading AI developers are currently designing their own AI server chips in an effort to reduce their dependence on NVIDIA (NASDAQ: NVDA) graphics processing units (GPUs). According to people familiar with the matter, as part of a recent acquisition aimed at strengthening the Nemotron model, Inception co-founders Stefano Ermon, Aditya Grover, and Volodymyr Kuleshov are expected to join NVIDIA (NASDAQ: NVDA) full time. All three are well-known machine learning professors, teaching at Stanford University, the University of California, Los Angeles, and Cornell University, respectively. Inception previously raised $50 million in financing from investors including Menlo Ventures, M12, Mayfield, and NVIDIA (NASDAQ: NVDA).

One of the biggest advantages of diffusion models is that they generate much faster than autoregressive models, especially when generating longer responses. Inception is a company with more than 30 employees. In early 2025, it became widely known for releasing a "diffusion" large language model. This type of model is completely different from the "autoregressive" language models used by products such as ChatGPT and Claude, which generate content word by word from left to right. Diffusion language models work by first generating a complete piece of content that initially looks chaotic and seemingly random, and then repeatedly refining it until it is turned into a formal response.

Diffusion models have traditionally been used in image generation models, but recently, startups such as Inception and large developers such as Google DeepMind have applied them to language generative AI. In May of this year, NVIDIA (NASDAQ: NVDA) released a version of its Nemotron language model that combined diffusion mechanisms with autoregressive and other more traditional forms of language AI. A notable advantage of diffusion models is that they generate much faster than autoregressive models, especially when generating longer responses.

According to people familiar with the matter, researchers within NVIDIA (NASDAQ: NVDA) responsible for the Inception project also plan to apply the related results to so-called "physical models" — models designed to help robots perceive and handle the world around them.

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Comments

We need your insight to fill this gap
Leave a comment