In January 2025, DeepSeek AI introduced two powerful large language models (LLMs) that shook the AI world. Developed at a fraction of the cost of its existing peers, DeepSeek holds the potential to transform the AI development landscape globally.

DeepSeek’s efficiency enables better cost-effectiveness by reducing computational needs

DeepSeek’s V3 and R1 models focus on efficiency and require less computing power than rival models while delivering equivalent performance. Its efficiency stems from using the “Mixture-of-Experts” (MoE) architecture, which activates only parts of the model for a given task, minimizing computational needs. This targeted use of computational memory reduces operational costs, giving it a significant edge over competitors who rely on more resource-intensive approaches.

The arrival of DeepSeek has sent shockwaves throughout the US tech industry, marked by a significant decline in stock values. The key headline event was Nvidia’s US$600 billion drop in market capitalization. The fact that a Chinese company was able to achieve groundbreaking results at a fraction of the cost by using low-power Nvidia H800 chips challenges the investment poured into the market by established players.

The open-source model enables widespread applications at a budget

DeepSeek has embraced a fully open-source model, allowing anyone to utilize their technology for commercial purposes. DeepSeek’s open-source approach democratizes access to AI, enabling a wider range of applications.

The availability of DeepSeek’s advanced APIs at a very low cost also appeals to customers who have previously been priced out of advanced AI applications due to the higher costs of proprietary LLM models such as OpenAI’s GPT.

The AI ecosystem already feels the impact of DeepSeek’s triumph. Its free AI assistant has also made a significant splash in the consumer market, with DeepSeek’s app surpassing ChatGPT in Apple Store charts. Its cost-effectiveness has even attracted the attention of major players such as AWS and Snowflake, which are now offering DeepSeek’s technology on their platforms.

Following DeepSeek’s success, several other Chinese companies may follow suit by developing more efficient yet high-performing AI models, further driving the costs down. Alibaba already released a new version of its Qwen 2.5 model at the end of January 2025.

Initial success of DeepSeek does not guarantee dominance

DeepSeek’s success does not guarantee its dominance in the AI landscape. We have had precedence of a similar company making headway and then falling off in the AI space. Mistral’s open-source AI model, Mixtral 8x7b, initially seemed poised to disrupt the field. However, it quickly fell off the radar when other closed-source models incorporated Mixtral’s innovations.

DeepSeek’s continued success will depend on whether it is able to maintain its edge through continuous innovation, particularly with limited access to high-performance chips.

EOS Perspective

DeepSeek’s emergence as a serious contender has intensified the global AI race, challenging the dominance of established players such as OpenAI, Meta, and Google.

DeepSeek-R1, with its open-source foundation, has already demonstrated impressive abilities in handling complex text-based tasks, such as summarizing documents, answering technical questions, and generating codes. Moreover, it offers these capabilities through APIs at a fraction of the cost of its competitors, potentially disrupting the market and driving down prices for AI services.

With DeepSeek’s AI models requiring less computational power and hardware, they will offer significant cost savings for users. Combined with its open-source model, which fosters customization, collaboration, and broader access, DeepSeek is expected to gain traction rapidly. While it is currently limited to text-based queries, its potential is undeniable.

While questions about Chinese government influence and censorship persist, DeepSeek presents a compelling vision of AI disruption. In the short term, we can anticipate lower AI adoption costs and shrinking profit margins for established AI providers. Furthermore, DeepSeek’s emphasis on efficiency could spark a shift in the industry, prioritizing resource optimization over simply increasing computing power. The full scope of DeepSeek’s impact, however, will only unfold over time.