Nvidia Nears $500m Deal to Buy AI Startup Inception

Nvidia is in advanced talks to acquire Inception, a two-year-old startup building a different kind of large language model, for at least $500 million, according to a report by The Information.
The report, cited by TipRanks and The Decoder on 20 November, says the deal would also include performance-based payments for Inception’s more than 30 employees. Three people familiar with the talks were cited as the source. Nothing has been signed or announced, and neither company has commented, so the deal is reported rather than confirmed and the terms could still change.
What Inception builds
Inception develops diffusion-based large language models, which work differently from the conventional models behind products such as ChatGPT and Claude. Conventional models generate text one token at a time, in sequence. Inception’s models refine a whole response step by step, allowing multiple tokens to be generated in parallel.
Its Mercury family of models is built on this approach, and its most recent release is Mercury 3. Inception says its models can generate text up to 10 times faster than comparable models. The company has also developed diffusion-based coding models and published research on multimodal understanding and reasoning.
Who would join Nvidia
All three of Inception’s co-founders are expected to join Nvidia full time if the deal goes through. They are Stefano Ermon, the startup’s chief executive and a professor at Stanford University; Aditya Grover, a professor at the University of California, Los Angeles; and Volodymyr Kuleshov, a professor at Cornell University.
Inception is based in the San Francisco Bay Area and has more than 30 employees. It raised a $50 million seed round in November 2025 led by Menlo Ventures, with investors including Microsoft’s M12 fund, Mayfield and Nvidia’s own venture arm, NVentures.
Why Nvidia wants it
The acquisition would aim to strengthen Nvidia’s open-source Nemotron models and its physical AI models for robots, helping it compete with OpenAI and Anthropic as both build their own inference chips. Nemotron models are already used by Booz Allen Hamilton, Palantir and several US government agencies, which can customise them for specific applications.
The talks began before Jensen Huang’s departure as Nvidia’s chief executive, and Microsoft had also shown interest in a similar deal, according to the reports. That timing matters: the deal would be one of the first major moves under Nvidia’s interim leadership, and it would extend the company further beyond its core business of supplying AI chips into owning the models that run on them.
What happens next
There is no public timetable, and advanced talks can still fall through or change in price and structure. The reported figure of at least $500 million is a floor rather than a final price, with the earnout payments tied to performance targets for the employees who move across.
If completed, the deal would bring a small but specialised research team, and a technology designed for speed, into a company that has invested heavily in its Nemotron line. For now, both sides have stayed silent, and the only public account of the talks comes from the people familiar with them cited by The Information.
Sources
TipRanks: Nvidia Nears $500M Deal for AI Startup Inception to Strengthen AI Models
The Decoder: Nvidia bets at least $500 million on Inception’s diffusion models to strengthen its open-source Nemotron lineup