DeepSeek, a Chinese AI startup, has gained global attention with its AI model R1, reportedly trained on minimal hardware resources. The company claimed it spent only $6 million and used 2,048 GPUs for training. However, findings from analyst firm SemiAnalysis contradict this, revealing DeepSeek’s vast infrastructure and significant investments. According to SemiAnalysis, DeepSeek operates a large-scale computing infrastructure with 50,000 Nvidia Hopper GPUs. DeepSeek might not be as disruptive as claimed because its AI model R1 reportedly relies on a $1.6 billion hardware investment. The company has reportedly spent $1.6 billion on hardware alone, including 10,000 H800s and 10,000 H100s. Additional investments in H20 units further bolster its computing power. Operating costs are estimated at $944 million. These findings challenge the narrative that DeepSeek achieved high-level AI training at dramatically lower costs.
Origins and Business Model
DeepSeek was established as an AI-focused venture by High-Flyer, a Chinese hedge fund that invested heavily in GPUs. Unlike many AI startups, DeepSeek remains self-funded, which allows flexibility in decision-making. Reports indicate that the company has invested over $500 million in AI development since its inception, far exceeding the initial $6 million training cost claim.
A major differentiator for DeepSeek is its ownership of data centers, unlike many AI firms that rely on cloud services. This self-sufficiency enables the company to conduct AI research, training, and optimization without external limitations. DeepSeek’s strategy emphasizes efficiency and algorithmic advancements over scaling up hardware resources.
Talent Strategy and Competitive Salaries
Unlike many Chinese AI companies that seek talent abroad, DeepSeek exclusively hires from within China. The company recruits graduates from top institutions such as Peking University and Zhejiang University. Reports indicate that some AI researchers at DeepSeek earn over $1.3 million, surpassing compensation at rival firms like Moonshot.
DeepSeek might not be as disruptive as claimed as its $6 million training cost excludes infrastructure, research, and operational expenses. DeepSeek has developed advanced AI techniques such as Multi-Head Latent Attention (MLA), which required extensive GPU usage. The company prioritizes algorithmic efficiency, which some experts believe could reduce demand for high-end GPUs in the long term. This has sparked concerns about AI infrastructure overcapacity, potentially affecting companies like Nvidia.
Market Reaction and Fitch Ratings’ Analysis
Industry analysts suggest that DeepSeek might not be as disruptive as claimed, given its reliance on large-scale computational resources similar to its competitors. DeepSeek’s claims about its low-cost AI model led to major fluctuations in the tech market. Following the company’s announcement, Nvidia’s market capitalization dropped by $589 billion. Fitch Ratings noted that DeepSeek’s emergence raises concerns over excessive AI infrastructure investments. The agency predicts a slowdown in hyperscaler spending on AI infrastructure, which could impact semiconductor companies.
DeepSeek’s expansion faces hurdles beyond financial scrutiny. Italy’s data protection authority recently requested details about the company’s data usage. Following an inadequate response, the regulator ordered service providers to block DeepSeek’s chatbot in the country. Additionally, security concerns and geopolitical factors may limit DeepSeek’s adoption in Western markets.
Future of AI Investments
The rise of efficient, open-source AI models could accelerate broader AI adoption. Fitch Ratings suggests that companies like AMD and Intel may benefit from the growing demand for AI chips suited for inference tasks. However, Nvidia, a dominant player in AI training, could face challenges if the industry shifts towards cost-effective AI models.
DeepSeek’s advancements highlight the shifting landscape of AI development. While its claims of low-cost AI training stirred excitement, deeper analysis reveals substantial investments driving its success. The company’s future depends on balancing innovation with regulatory compliance and evolving market dynamics.
