The artificial intelligence (AI) semiconductor market, which was virtually monopolized by U.S. company Nvidia, has seen a disruption. With the release of Google's 'Gemini 3' shaking the AI market landscape, there are projections that Google's self-developed chip might replace Nvidia's AI chips.

On the 25th (local time) at the New York Stock Exchange, Nvidia closed down 2.59% compared to the previous trading day. Among the so-called 'Magnificent 7 (M7)' big tech companies, all except Nvidia saw an increase. Notably, following the release of Gemini 3, the stock price of Google's parent company Alphabet, which has been attracting market attention, continued to hit record highs, surpassing a market capitalization of USD 3.9 trillion (approximately KRW 5,720 trillion).

Since the release of Google's Gemini 3, the stock prices of Nvidia and Google have diverged. The decisive factor was Google's announcement that it trained Gemini 3 using its self-developed AI inference chip, the Tensor Processing Unit (TPU). Previously, Nvidia's Graphics Processing Unit (GPU) was considered essential for AI learning and training based on large language models (LLM), but Google released an advanced AI model without using GPUs. The fact that Nvidia's key clients, like Meta, are considering purchasing Google's TPU also supports the projection that Nvidia's monopoly might be weakening.

Nvidia holds a 90% market share in the AI data center GPU market. The widespread use of Nvidia's software platform CUDA by AI researchers has also acted as a network effect, broadening Nvidia's 'economic moat.' Thanks to this, Nvidia has been able to double its revenue annually over the past three years while maintaining an operating profit margin of around 60%. Due to Nvidia's growth and profitability, it has been valued at a higher level than other big tech companies.

Nvidia responded promptly. Through social media platform X, Nvidia stated, "We are pleased with Google's success," and added, "Nvidia continues to supply chips to Google. Nvidia's technology is a generation ahead of the industry."

In reality, it is widely believed that Google's TPU cannot completely replace Nvidia's GPU. While TPU can be used as an alternative for developing AI models in terms of energy efficiency, it is dependent on Google's cloud platform, and the versatility of Nvidia's GPU is much greater. Hwang Soo-wook, a researcher at Meritz Securities, stated, "It is reasonable that the innovation of Gemini 3 is a factor in Alphabet's stock price increase," but added, "Considering the era of physical AI that requires the ability to implement three-dimensional space, the interpretation that Nvidia's GPU will be completely replaced by TPU is somewhat excessive."

Popular News