Analysis examines how DeepSeek's AI breakthrough challenges industry assumptions and opens competition in AI development.

Chinese start-up DeepSeek's chatbot overtook ChatGPT downloads on the Apple and Google App Stores this week, demonstrating an open-source AI model that matches leading US models' performance at a fraction of the cost by using compute and memory resources more efficiently. According to Forrester Research, DeepSeek is more than China's "ChatGPT"; it represents "a major step forward for global AI by making model building cheaper, faster, and more accessible," and should be "celebrated as a milestone for AI progress."

The cost advantages of DeepSeek's methodology extend across the entire infrastructure stack. DeepSeek's "DeepThink" feature, which leverages the R1 model to provide enhanced reasoning capabilities using advanced techniques to break down complex queries into smaller, manageable tasks, cost only about $5.5 million to train—tens of millions of dollars less than similar models. This efficiency addresses what Rita Sallam, a distinguished vice president analyst at Gartner Research, identified as a critical barrier: "Current leading vendor costs and resulting pricing models that pass on the high cost of innovation and developing, training and running LLMs have resulted in only the largest most well-funded companies with the deepest pockets being able to develop LLMs." High costs have hindered GenAI adoption, leading to negative returns on investments for many organizations' use cases at scale, which has stifled innovation even with vendor subsidies.

DeepSeek's approach cuts hardware and computational costs, allowing smaller companies to compete in LLM development and drive new GenAI innovation. Brendan Englot, a professor and AI expert at Stevens Institute of Technology in New Jersey, emphasized that "DeepSeek's technology provides an excellent example of how disruptive and innovative new tools can be built faster with the aid of open source software." The arrival of DeepSeek caused Nvidia's stock to tank as investors realized the impact of more efficient processes on AI processor and accelerator sales. However, Forrester Research noted that while the lower cost will reduce short-term demand for Nvidia, it will likely drive more startups and enterprises to create models, boosting demand long-term. Meta's planned $65 billion investment in a data center with a footprint almost as large as Manhattan, expected to come online at the end of this year and housing 1.3 million GPUs, illustrates the scale of current infrastructure spending.

While DeepSeek's base technology isn't pioneering—the company's published research paper shows that Meta's Llama and Alibaba's Qwen models were key to developing DeepSeek-R1 and DeepSeek-R1-Zero—the breakthrough demonstrates significant efficiency gains. Englot stated that "less computing hardware is needed to train and deploy a powerful language model than we originally assumed. This can permit AI innovators to forge ahead and devote more attention to the resources needed for multi-modal AI and advanced applications beyond chat-bots." Though the cost to train AI models has declined significantly with DeepThink, Forrester noted that "the cost to support inferencing will still require significant compute and storage," and that "this shift shows that core AI model providers won't be enough, further opening the AI market." Mel Morris, CEO of startup Corpora.ai, observed that DeepSeek's affordability and open-source model allows developers to customize and innovate cheaply and freely, and "the idea that competition drives innovation is particularly relevant here, as DeepSeek's presence is likely to spur faster advancements in AI technology, leading to more efficient and accessible solutions to meet the growing demand."