NVIDIA is spending $7 billion to acquire over 100 top talents, aiming to build the world's most powerful open-source model and capture the Chinese AI market.

NVIDIA is not only the world's most profitable AI hardware company, but is now also focusing on large model development, spending a whopping $7 billion on talent acquisition to build the world's most powerful open-source large model.

US media reports that NVIDIA is set to reach a cooperation agreement with AI startup Poolside, investing $1 billion in the company at a $12 billion valuation, while spending $6 billion to purchase the company's technology intellectual property and some of its R&D personnel.

This cooperation model is similar to NVIDIA's deal with LPU chip company Groq late last year—not a direct acquisition, but a heavy investment in technology licensing and talent, with the company continuing to operate independently.

Around 100+ employees from Poolside will join NVIDIA, working on NVIDIA's self-developed Nemotron large model series, aiming to make it the world's most powerful.

As is well known, NVIDIA strongly supports open-source large models and has previously praised Chinese open-source models like DeepSeek and Kimi. Its own Nemotron model is also open-source, though its performance hasn't been outstanding, falling short of Chinese open-source models like DS V4 and Kimi K3.

As the world's largest AI GPU supplier with substantial profits, NVIDIA holds unique advantages in funding and computing resources. Now with a pool of new talent, Nemotron has strong potential to break into the top tier in the future.

NVIDIA developing the most powerful AI model itself isn't good news for VIP customers like OpenAI and Anthropic, but Jensen Huang is simply giving them a taste of their own medicine.

That's because OpenAI and Anthropic are both heavily purchasing AI computing power beyond NVIDIA GPUs, including from Google and Amazon, and are developing their own AI chips. They don't want to rely on NVIDIA chips, so NVIDIA naturally can't pin its large model hopes on them either.

For NVIDIA, developing the strongest open-source AI has other benefits as well. The US will eventually crack down on Chinese open-source AI, and NVIDIA's large model could seize that market share in the future. They don't want DeepSeek, Qwen, Kimi, GLM, and other Chinese open-source models to become the primary large models used by American companies.