NVIDIA is in advanced talks to acquire Inception, the two-year-old startup behind the Mercury family of diffusion language models, for at least $500 million plus performance-based earnouts tied to targets for the new employees. The report comes from The Information, citing three people familiar with the talks. This is a rumor: nothing has been signed or announced, and neither NVIDIA nor Inception has commented.
Instead of writing text one token at a time, like the models behind ChatGPT and Claude, Inception's diffusion models refine many tokens in parallel. The company says this makes them up to 10 times faster and much cheaper to run. Its latest model, Mercury 3, generates more than 1,000 tokens per second on NVIDIA GPUs, according to the startup.
According to the report, all three co-founders are expected to join NVIDIA if the deal closes: CEO Stefano Ermon, a Stanford professor, Aditya Grover of UCLA, and Volodymyr Kuleshov of Cornell. The technology and the team would help NVIDIA build out its Nemotron family of open models and its AI systems for robotics.
Inception has raised about $50 million from investors including Menlo Ventures, M12, Mayfield and NVIDIA itself, and has more than 30 employees. Microsoft had also shown interest in a similar deal, according to the report. Another concrete sign: Inception told developers that its Mercury API and platform dashboard will remain available only through December 28, 2026.
The talks are said to have started before Jensen Huang stepped down as NVIDIA's CEO — Colette Kress took over as interim CEO — a sign that the company's push into its own models continues under interim leadership.
Sources: Dealroom (summary of The Information's reporting), TipRanks, The Neuron.
Image: NVIDIA logo, via Wikimedia Commons.
Transparency: this article was produced with the help of artificial intelligence, based on cross-checking public posts and internet sources. Please consult the original sources listed above.
By Muse