The artificial intelligence sector saw a notable shift in January 2025 with the introduction of DeepSeek.
The arrival of this Chinese startup sparked substantial market reactions, including a historic $600 billion drop in Nvidia’s market value—the largest single-day decline recorded in US market history.
But beyond the market ripples, a key question stands: Does DeepSeek AI represent a real challenge to established players like OpenAI, Google’s Gemini, and Anthropic’s Claude?
A New Approach to AI Development
DeepSeek AI’s rise stems from its distinctive strategy. The company built its large language models using approximately $6 million in computational resources—a fraction of what major competitors spend.
Their pricing structure reflects this efficiency, charging around $0.55 per million input tokens compared to OpenAI’s $15.
This cost reduction comes from their “Mixture of Experts” method, which turns on only the necessary computational parts for specific tasks. The approach helps reduce energy use while maintaining speed and adaptability.
DeepSeek’s path also shows steady progress.
Starting with DeepSeek Coder for programming tasks, they moved to a 67-billion parameter model in December 2023. By December 2024, their V3 version reached 671 billion parameters. Each step brought improvements in both size and capability.
How DeepSeek Measures Up
When comparing DeepSeek-R1 with other AI models, the results paint an interesting picture:
Matched against OpenAI’s models, DeepSeek-R1 shows strong performance in mathematical reasoning, scoring 79.8% on advanced multi-step reasoning tests versus OpenAI’s 79.2%. However, OpenAI maintains an edge in general knowledge tasks.
Google’s Gemini brings different strengths, particularly in processing various data types like images and videos. Its integration with search functions sets it apart from DeepSeek’s text-focused approach.
Anthropic’s Claude stands out in coding tasks and creative writing, showing that raw parameter count isn’t everything in AI performance.
Market Effects and Industry Response
The ripples of DeepSeek’s market entry spread far beyond initial stock reactions.
The technology sector experienced a substantial shake-up, with Nvidia’s $600 billion market value drop being the most significant single-day decline in US market history.
The response from established tech companies proved equally significant. Meta established dedicated teams to analyse DeepSeek’s methods, particularly their achievement of such high performance at remarkably low costs.
This move signals a broader industry shift toward examining cost-effective AI development strategies. The effects have extended to investment strategies as well.
DeepSeek’s ability to create its R1 model at a relatively low cost prompted investors to question current AI business models. This cost difference sparked discussions about sustainable development practices in AI technology.
President Donald Trump characterised DeepSeek’s rise as a “wake-up call” for US technology companies, highlighting potential shifts in global tech leadership. Analysts now watch for changes in semiconductor market demand, as DeepSeek’s cost-effective approach might alter requirements for high-performance chips.
Looking at Reality
DeepSeek’s advantages in cost and mathematical reasoning are clear. Their open-source approach allows broader access to AI technology. However, this needs to balance against the established track record of companies like OpenAI in general knowledge tasks and Claude in coding.
The actual test will come from real-world applications. While benchmark scores matter, practical usefulness determines long-term success. DeepSeek’s ability to maintain quality while keeping costs low will be critical.
Looking Forward
DeepSeek AI represents a shift in how companies might approach AI development. Their success in creating capable models at lower costs could lead to more accessible AI tools. This accessibility might speed up AI adoption across various industries.
The competition between DeepSeek and established players will likely drive innovation in AI development methods. As companies work to balance cost, performance, and accessibility, users stand to benefit from improved AI tools at better prices.
