Nvidia Reportedly Plans to Spend More Than $500 Million to Acquire Diffusion-Based Large Language Model Company Inception
Nvidia is reportedly in advanced talks to acquire AI startup Inception in a deal valuing the company at more than $500 million, potentially with additional performance-based incentives. Inception's Mercury uses a diffusion architecture to generate model tokens in parallel. The company says it has a time to first token of under 300 milliseconds, throughput five to seven times higher than conventional models, and can reduce the cost per task by up to 70%. Neither party has commented publicly, and the deal price, closing date, and whether Mercury would continue to operate independently remain undecided.
Why read this
AI score 55
The potential acquisition would bring diffusion-based large language models and their parallel generation approach into Nvidia's portfolio, offering clues to understanding its model and physical AI strategy.
Topics
Body · AI translation
Report: Nvidia to Spend More Than $500 Million to Acquire Inception, the Company Behind “AI Faster Than ChatGPT”
Download the IT Home app for an exclusive first look
November 20, 2026, 14:27:24, from Shandong
IT Home, November 20 — Technology outlet The Information published a report today (November 20) saying that Nvidia plans to acquire AI startup Inception. The two parties are currently in advanced talks over a deal valued at more than $500 million (IT Home note: approximately RMB 3.346 billion at current exchange rates), potentially with additional performance-based incentives.
Amir Efrati is the executive editor of The Information.
According to publicly available information reviewed by IT Home, Inception was founded in 2024 and is headquartered in Palo Alto, California. It was co-founded by Stanford University professor Stefano Ermon, UCLA professor Aditya Grover, and Cornell University professor Volodymyr Kuleshov.
The company primarily develops “diffusion large language models” (dLLMs). In February 2025, it released Mercury, which it describes as the world’s first commercial-scale diffusion large language model. Its products are available to developers through its own API, Azure AI Foundry, OpenRouter, Baseten, and other channels.
The Mercury model’s main highlight is parallel generation. Unlike the “autoregressive” approach used by OpenAI’s GPT series and Anthropic’s Claude series, the model uses a diffusion architecture that begins with rough, “noise-like” content and then performs multiple rounds of denoising and refinement, generating multiple tokens in parallel.
In terms of cost-effectiveness, Inception says Mercury can achieve first-token latency of less than 300 milliseconds, throughput five to seven times that of conventional models, and cost reductions of up to 70% per task. The outlet described it as “AI faster than ChatGPT.”
According to the report, if the deal is completed, Inception’s three co-founders are expected to join Nvidia. Its team may also help develop Nvidia’s Nemotron open-source model ecosystem and extend diffusion technology into robotics and other areas of “physical AI.”
Neither Nvidia nor Inception has commented publicly. As negotiations have not yet been finalized, the purchase price, closing date, and whether products such as Mercury will continue to operate independently remain uncertain.
“Special notice: The above content (including any videos, images, or audio) was uploaded and published by users of Dafeng Hao, a self-media platform operated by Phoenix New Media. This platform merely provides information-storage services.
Notice: The content above (including the videos, pictures and audios if any) is uploaded and posted by the user of Dafeng Hao, which is a social media platform and merely provides information storage space services.”
AI Trends