Nvidia is in advanced talks to acquire diffusion AI startup Inception for over $500 million, according to The Information. The deal would bring Inception's co-founders and its unique diffusion-based language models into Nvidia full-time.
TL;DR
- Nvidia is reportedly acquiring diffusion AI startup Inception for over $500 million.
- Inception's Mercury models are 5 to 10 times faster than competitors while matching their accuracy.
- The acquisition aligns with Nvidia's strategy to expand beyond AI hardware into models and software.
What happened
Nvidia is in advanced discussions to acquire Inception, a two-year-old AI startup developing diffusion-based large language models, for over $500 million, according to The Information. The deal would bring Inception's co-founders—Stefano Ermon, Aditya Grover, and Volodymyr Kuleshov—into Nvidia full-time, along with additional earnout awards for employees, subject to performance targets. Inception, based in Palo Alto, California, has over 30 employees and raised a $50 million seed round in November 2025, with Nvidia's venture arm NVentures among its backers. The startup's Mercury family of models uses an iterative refinement process designed to accelerate text generation, claiming to be 5 to 10 times faster than speed-optimized models from OpenAI, Anthropic, and Google while matching their accuracy. Mercury has been integrated into developer tools including ProxyAI, Buildglare, and Kilo Code.
Why it matters
The potential acquisition fits into Nvidia's broader strategy to expand beyond its role as a leading supplier of AI computing hardware. By bringing Inception in-house, Nvidia would gain greater control over the development of its diffusion-based model technology, strengthening its wider ecosystem. This move follows Nvidia's agreement to acquire Hugging Face for about $12.9 billion in September 2026 and its $20 billion non-exclusive licensing agreement with Groq in December 2025. For AI developers and startups, this acquisition could mean faster, more efficient diffusion models integrated into Nvidia's offerings, potentially accelerating AI development and deployment. However, the discussions remain at an advanced negotiation stage, and neither company has publicly confirmed an agreement, leaving room for changes in price and terms.
Key facts
- Inception was founded by Stefano Ermon, Aditya Grover, and Volodymyr Kuleshov, all of whom would join Nvidia full-time under the acquisition deal.
- Inception raised a $50 million seed round in November 2025, led by Menlo Ventures, with participation from Nvidia's NVentures, Microsoft's M12, Snowflake Ventures, Databricks Investment, and others.
- Inception's Mercury models are claimed to be 5 to 10 times faster than speed-optimized models from OpenAI, Anthropic, and Google while matching their accuracy.
- The acquisition price is notably lower than the figure Inception was reportedly valued at earlier in 2026, when Microsoft and SpaceX were also courting the startup.
- Nvidia has been investing in AI models, software platforms, and research teams to strengthen its ecosystem, including planned minority investments in European startups like Doctolib, Genesis AI, Pigment, and Entalpic.
- Nvidia agreed to acquire Hugging Face for about $12.9 billion in September 2026 and signed a $20 billion non-exclusive licensing agreement with Groq in December 2025.
Context
Nvidia's push to acquire Inception aligns with its strategy to expand beyond AI hardware into models and software, following its agreements with Hugging Face and Groq. Diffusion models, like those developed by Inception, offer a different approach to text generation, potentially improving inference speed and computing efficiency. This acquisition, if finalized, would give Nvidia greater control over the development of diffusion-based model technology, strengthening its position in the AI ecosystem. For developers and startups, this could mean faster, more efficient tools integrated into Nvidia's offerings, accelerating AI development and deployment.
