DeepSeek’s $6 Million AI Model Makes Waves in the Tech Industry
The topic of DeepSeek, a Chinese artificial intelligence (AI) model, has taken center stage in the tech industry this week, with a focus on a single figure: $6 million. According to a paper published by DeepSeek, the company’s total training costs for its newest AI model amount to $5.576 million, based on the rental price of Nvidia’s graphics processing units. However, the number only includes “official training” and excludes costs related to “prior research and ablation experiments on architectures, algorithms, or data.”
DeepSeek’s AI Assistant became the most-downloaded free app in the US on Apple’s App Store, dethroning OpenAI’s ChatGPT. The news sparked a sell-off in global tech stocks, with chipmakers Nvidia and Broadcom losing a combined $800 billion in market capitalization on Monday.
A new report from SemiAnalysis, a semiconductor research and consulting firm, provides additional context to DeepSeek’s expenses. The firm estimates that DeepSeek’s hardware spend is “well higher than $500 million over the company’s history,” pointing out that research and development (R&D) costs and total cost of ownership are significant. The report also notes that generating synthetic data for the model to train on requires a “considerable amount of compute.”
DeepSeek’s model is considered to be a remarkable achievement, especially considering the US curbed chip exports to China several times in recent years, leading to concerns that the US is falling behind China in the AI market. The US chip-making industry is predicted to reach $1 trillion in revenue within a decade.
Experts have praised the quality of DeepSeek’s model, noting that “catching up to the reasoning edge this quickly is objectively impressive.” DeepSeek’s R1 model is also open-source, allowing any AI developer to use it.
The buzz around DeepSeek has led to concerns about the potential implications for the AI infrastructure complex. Some experts believe that DeepSeek’s achievement may mark a turning point in the AI market, while others think it is just a sign of increased competition.