【AI+NVDA】NVIDIA Develops AI Large Models - Direct Challenge Throughout the Entire Industry Chain?

NVIDIA’s Q3 2025 financial report shows that the company will invest a total of $26 billion over the next five years in open-source AI large models across the entire industry chain. The first self-developed open-source AI models are expected to be released between late 2026 and early 2027. This investment far exceeds the $3 billion spent on training GPT-4 by OpenAI. Market analysts believe this move makes the GPU giant a direct competitor to its clients OpenAI, Anthropic, and DeepSeek.

Foreign media point out that NVIDIA’s carefully chosen open-weight models could become a key differentiator in its strategy. These models can compete with OpenAI and Anthropic’s closed-source models while also distinguishing itself from Meta’s LLaMA open-source models.

Analysts note that companies increasingly need model transparency and customization but lack resources to train from scratch. If NVIDIA can offer hardware-optimized, competitive open-weight models, it could build a strong moat.

Some analysts believe that if NVIDIA maintains its hardware dominance and captures 10% of the foundational AI model market within three years, its annual revenue could increase by $50 billion. However, if this leads to customer loss to AMD or other chip manufacturers, it could backfire and cause significant losses.

Documents show that the $26 billion will be used for model development, computing infrastructure, R&D talent, and “ecosystem building,” with expenditures gradually increasing over the next 18 to 24 months.

Foreign media indicate that NVIDIA already has all the necessary conditions. Over the past two years, NVIDIA has quietly assembled an AI research team and controls key computing infrastructure. CEO Jensen Huang has repeatedly stated that NVIDIA is a “full-stack AI company.”