Those who follow the AI industry are well-versed in the former dominance of NVIDIA + OpenAI.

NVIDIA, with its AI chips, holds a commanding 94% share of the global data center market, leaving AMD and Intel trailing far behind. Its market dominance is undeniable.

On the software front, OpenAI provides AI models that are unparalleled globally. Other vendors strive to emulate OpenAI's success.

Together, NVIDIA + OpenAI formed an unbeatable duo in the AI world, integrating hardware and software seamlessly. Competitors could only aspire to follow in their footsteps, often with limited success.

However, the advent of China's Deepseek has altered the landscape, presenting a formidable challenge to the NVIDIA + OpenAI juggernaut.

Deepseek boasts training costs that are a mere fraction—approximately one-tenth—of those incurred by OpenAI, without the need for an extensive stack of AI chips. Furthermore, Deepseek operates independently of NVIDIA chips and the NVIDIA CUDA ecosystem.

Remarkably, Deepseek outperforms OpenAI.

This achievement strikes at the heart of NVIDIA's strength, demonstrating that advanced chips are not a necessity for such results. The question arises: why would anyone continue to invest in NVIDIA chips?

With lower training costs and emerging as a formidable player, Deepseek's superiority over OpenAI begs the question: what advantage does OpenAI still hold?

As a result, many believe that the global AI model industry is ushering in a new era, one where the dominance of NVIDIA + OpenAI has been significantly challenged.

Indeed, this transition is already underway. The Chinese market has witnessed the emergence of numerous Deepseek all-in-one machines, many of which utilize Huawei chips, while others incorporate domestic AI chips from various sources.

Experts predict that in the future, Huawei chips + Deepseek could potentially supplant NVIDIA + OpenAI. In terms of chips, Huawei's AI chips are more than capable of rivaling NVIDIA's in performance.

Regarding models, Deepseek's higher cost-effectiveness and reduced financial burden make it a viable replacement for OpenAI, aligning well with current market demands.

However, some caution that the situation is more nuanced. If OpenAI achieves 100 intelligence with 100 computing power, and Deepseek attains 1000 intelligence with the same 100 computing power, this exponential growth suggests that the potential for advancement is vast—perhaps even reaching 10,000, 1 million, or 10 million intelligence levels. Therefore, replacing NVIDIA may not be as straightforward as it seems.