AI investors have valid reasons to feel uneasy today.
On Jan. 21, OpenAI launched the $500 billion Stargate Project with partners like Arm, Microsoft, NVIDIA, and Oracle. This initiative aimed to secure American AI leadership, create jobs, and boost the economy. Initially, this boosted the stock prices of these companies.
However, six days later, China’s DeepSeek introduced a cost-effective and efficient AI model, causing these stocks to plummet. DeepSeek’s model raised questions about the future demand for AI chips and infrastructure like data centers and power plants.
DeepSeek’s open-source approach and impressive performance, achieved with fewer resources than its Western counterparts, challenge the massive spending strategies of companies like Nvidia and Meta. As AI competition intensifies, the returns on AI investments are uncertain. Investors should be cautious to avoid losses during the market’s ups and downs.
When OpenAI launched ChatGPT in late 2022, it revolutionized AI with its ability to engage in natural language conversations, generate coherent text, and perform tasks like coding and content creation. Within just five days of its release, ChatGPT gained over one million users, making it the fastest-growing consumer application in history at that time. By January 2023, it had reached 100 million users, captivating a global audience and prompting widespread integration of AI into various business operations, from customer service to content generation.
This rapid adoption led to a surge in demand for AI infrastructure, particularly high-performance Graphics Processing Unit (GPUs) from companies like Nvidia and AMD. These chips became essential for training and running large language models (LLMs), driving the AI revolution. Data centers and power plants also saw increased demand to support these AI models.
However, OpenAI faced significant challenges in developing ChatGPT 5.0, including technical difficulties with scaling the model, managing costs, and ensuring safety and reliability. These setbacks highlighted the uncertainties in AI development, which requires vast computational resources and innovative approaches to ensure powerful and safe models.
Amid these challenges, a Chinese AI startup called DeepSeek was founded in 2023 by Liang Wenfeng in Hangzhou, Zhejiang. The startup gained significant attention with its open-source AI model, DeepSeek-R1.
Unlike proprietary models from OpenAI and Google, DeepSeek-R1 was available to the public under an open-source license, allowing developers to access, modify, and build upon it. This approach marked a significant shift from the closed systems dominating the AI landscape. DeepSeek-R1’s efficiency, achieved with fewer resources, made it a serious competitor to established AI models.
DeepSeek-R1’s training process required just 2,000 second-tier Nvidia chips, compared to the 16,000 chips used to train Meta’s Llama 3.1. This efficiency reduced development costs and made the model more accessible.
DeepSeek-R1’s pricing was also significantly lower than OpenAI’s models, making it a cost-effective alternative. Specifically, DeepSeek-R1 costs $0.55 per million input tokens and $2.19 per million output tokens, where tokens are the basic units of text that the model processes or generates. In comparison, OpenAI’s models, such as GPT-4, are priced at $2.50 per million input tokens and $10.00 per million output tokens.
Furthermore, in benchmark tests, DeepSeek-R1 performed reasoning tasks at the same level as OpenAI’s o1. Also, DeepSeek-R1 claimed it outperforms OpenAI’s models in mathematical reasoning. This performance positioned DeepSeek-R1 as a serious competitor in the AI field.
DeepSeek-R1’s success sent shockwaves through the global AI industry, particularly in the United States. The announcement of DeepSeek’s model coincided with a downturn in the stock market, as investors questioned the sustainability of growth for companies like Nvidia and AMD. Tech giants like Google and Meta, which invested heavily in proprietary AI models, also faced uncertainty about maintaining their competitive edge.
DeepSeek’s rise has been compared to the “Sputnik moment” for the United States, prompting calls for a reevaluation of America’s approach to AI development. The U.S. government’s efforts to ban TikTok raised questions about whether similar measures should be taken against DeepSeek.
However, banning an open-source model like DeepSeek-R1 would be challenging due to its freely available code.
DeepSeek-R1’s open-source architecture offers significant advantages. Unlike proprietary models, its code is available for anyone to use, modify, and share, fostering collaboration and innovation. Developers worldwide can build upon DeepSeek-R1, creating new applications and improving its capabilities. This openness democratizes access to AI technology, making it accessible to smaller companies, researchers, and individual developers.
DeepSeek’s rise challenges the West’s dominance in AI development. Rather than responding with fear and protectionism, the West should learn from DeepSeek’s open-source model. Embracing open-source AI development can foster a more collaborative and innovative ecosystem, accelerating AI advancement.
The challenges faced by OpenAI in developing ChatGPT 5.0 highlight the uncertainties in AI development.
Investors should be cautious, recognizing that while AI represents the next industrial revolution, not all companies will emerge as winners. Picking the right investments in this rapidly evolving landscape is risky. Managing a well-balanced portfolio of stocks would be a more prudent approach.
