Report: Nvidia to splash out over $500 million to acquire Inception, the “AI faster than ChatGPT” company
IT之家 November 20 news: Tech media outlet The Information today (November 20) published a blog post reporting that Nvidia plans to acquire AI startup Inception. The two sides are already in advanced talks, with the deal valuation at at least over $500 million (IT之家 note: about RMB 3.346 billion at the current exchange rate), and may include additional performance-linked incentives.

According to public information reviewed by IT之家, Inception was founded in 2024 and is headquartered in Palo Alto, California, USA. It was co-founded by Stanford University professor Stefano Ermon, UCLA professor Aditya Grover, and Cornell University professor Volodymyr Kuleshov.
The company mainly develops “diffusion large language models” (dLLM). In February 2025, it released Mercury, which it claims is the world’s first commercial-scale diffusion large language model. Its products are already available to developers through its own API, Azure AI Foundry, OpenRouter, Baseten, and other channels.
The biggest highlight of the Mercury model is parallel generation. Unlike the “autoregressive” approach used by OpenAI’s GPT series and Anthropic’s Claude series, this model adopts a diffusion architecture: it starts with a rough, “noise”-like piece of content, then goes through multiple rounds of denoising and refinement to generate multiple tokens in parallel.
In terms of cost-effectiveness, Inception says Mercury’s time to first token can be below 300 milliseconds, its throughput is 5 to 7 times higher than traditional models, and the cost per task can be reduced by up to 70%. The media outlet calls it “AI faster than ChatGPT.”
The report says that if the deal is completed, Inception’s three co-founders are expected to join Nvidia, and its team may also participate in building Nvidia’s Nemotron open-source model ecosystem and extend diffusion technology to robotics and other “physical AI” fields.
Currently, neither Nvidia nor Inception has publicly commented. Because the negotiations have not yet been finalized, there is still uncertainty about the deal price, closing time, and whether products such as Mercury will operate independently in the future.
Related articles
- The “brakeman” of Jensen Huang’s most aggressive AI bet: Nvidia interim CEO Kress once pushed to cut the $250 billion OpenAI gamble
- Nvidia open-sources strongest transcription AI model Nemotron 3.5 Transcribe: supports 40 languages, can accurately transcribe even when 8 people talk over each other
- Nvidia demonstrates DSX MaxLPS technology, AI factory token throughput per megawatt increased by up to 40%
- Largest AI supply chain leak of 2026: 195TB of data scraped from about 2,500 companies worldwide including Nvidia in 40 minutes
- Nvidia accelerates AI compute supply: cumulative Grace chip deliveries exceed 2.5 million, Vera Rubin racks target daily output of 1,000 units
- Agentic AI test: Nvidia Vera CPU is 2.2x faster than Intel Sapphire Rapids