DeepSeek has dropped over $500 million on Nvidia chips, to keep AI low costs, according to SemiAnalysis.

Table of Contents

On Monday, shares of Nvidia experienced a dramatic decline, plummeting by 17% in value. This significant drop resulted in a staggering loss of nearly $600 billion from the company's overall market capitalization. The catalyst for this steep sell-off was the announcement from the Chinese artificial intelligence firm DeepSeek, which revealed a groundbreaking advancement in its AI capabilities. However, this development has prompted a wave of skepticism among analysts regarding the cost implications associated with DeepSeek's achievements.

DeepSeek claimed that it had successfully trained its December V3 AI model for a relatively modest sum of $5.6 million. This figure, while impressive at first glance, has come under scrutiny from experts in the industry. Dylan Patel, an analyst from the chip consultancy firm SemiAnalysis, pointed out that this reported cost does not accurately represent the total investments made by DeepSeek over the course of its operations. Patel noted, "DeepSeek has spent well over $500 million on GPUs throughout the history of the company." He further elaborated that although the training run of the V3 model was highly efficient in terms of cost, it necessitated an extensive amount of experimentation and testing to achieve the desired results.

The fallout from Nvidia’s stock plunge was not limited to the company itself; it also had broader implications for the semiconductor industry. The Philadelphia Semiconductor index endured its worst daily drop since March 2020, falling by 9.2%. This decline generated substantial profits for short sellers, amounting to approximately $6.75 billion, as reported by the financial data group S3 Partners. 

In addition to the cost discussions, DeepSeek's engineering team made headlines by demonstrating their capability to write code independently, without relying on Nvidia's Cuda software platform. This platform is widely recognized as a key factor in maintaining Nvidia's dominance in the AI development sector, and DeepSeek's ability to bypass it raises questions about the future competitive landscape of AI technology. The implications of these developments are significant, as they suggest potential shifts in the balance of power within the industry and could challenge Nvidia's previously unassailable position.