Nvidia Will Spend $26 Billion to Build Open-Weight AI Models, Filings Show
The chip giant's five-year investment aims to create models tuned specifically for its hardware.
Nvidia is making a massive $26 billion strategic pivot into frontier AI model development, according to its 2025 financial filings. This five-year investment signals the chipmaker's evolution from a hardware and software provider into a direct competitor with labs like OpenAI and DeepSeek. The core of the strategy is building open-weight models—where the model parameters are publicly released—specifically tuned to run optimally on Nvidia's own hardware. This move is designed to further entrench its dominance in the AI chip market by creating a symbiotic ecosystem where its models showcase the full capabilities of its systems.
Alongside the funding announcement, Nvidia released Nemotron 3 Super, its most capable open model to date. With 128 billion parameters, it's comparable in size to OpenAI's GPT-OSS but claims superior performance, scoring 37 vs. GPT-OSS's 33 on the Artificial Intelligence Index. The company also revealed it has already pre-trained a massive 550-billion-parameter model. Nvidia executives state that building these models internally helps "stretch" and test their entire hardware stack, from compute to networking, informing their future architecture roadmap. This initiative also serves as a strategic counter to the growing influence of top-tier open models from Chinese companies like DeepSeek and Alibaba, which are currently favored by many global researchers and startups.
- $26 billion committed over five years to develop open-weight AI models, shifting Nvidia from chipmaker to frontier lab.
- Launched Nemotron 3 Super, a 128B-parameter model claiming to beat OpenAI's GPT-OSS on key benchmarks.
- Strategy aims to create hardware-tuned models to strengthen Nvidia's ecosystem and counter open Chinese model dominance.
Why It Matters
This massive investment could reshape the AI landscape, creating a powerful, hardware-optimized open model ecosystem directly controlled by the market's leading chip supplier.