Lookout OpenAI, DeepSeek is Behind You

While the United States has dominated AI development, led by firms such as OpenAI, China is rapidly closing the gap. DeepSeek, a Chinese AI model, presents a formidable challenge to American AI dominance. 

DeepSeek beats OpenAI on many fronts. 

One of the most striking advantages of DeepSeek is its dramatically lower training cost. While leading American firms spend tens of millions of dollars to develop large language models (LLMs), DeepSeek achieved similar performance for under $6 million. 

This cost-efficiency stems from the use of 2,000 lower-quality chips instead of the 16,000 high-performance chips required by US models such as Meta. Despite this hardware disparity, DeepSeek outperforms Meta on some benchmarks. 

By reducing the barrier to entry for AI model training, DeepSeek’s approach challenges the traditional cost-intensive paradigm of AI development. 

Unlike OpenAI’s proprietary models, DeepSeek operates on an open-source framework. This makes its model accessible to researchers and developers worldwide, accelerating improvements and modifications. 

American companies can technically adopt DeepSeek’s techniques, but its open-source nature gives China a strategic advantage in fostering AI innovation both domestically and internationally. 

The Chinese AI industry has found ways to optimise efficiency despite U.S. restrictions on high-end AI chips. Through software and algorithmic enhancements, DeepSeek has managed to compensate for a lack of premium hardware. This suggests that China can innovate around American sanctions and continue developing competitive AI models without relying on the latest NVIDIA or AMD chips. 

DeepSeek introduced a new “reasoning” model that competes with OpenAI’s state-of-the-art offerings such as O1 and O3. Unlike traditional LLMs, reasoning models engage in self-reflection before generating a response, leading to more accurate and coherent answers. 

However, this improvement comes at a higher energy cost. As AI models become more reasoning-intensive, operational costs will increase, potentially favouring firms that can develop cost-effective alternatives like DeepSeek. 

Historically, AI competition has been defined by massive investments, where a handful of dominant firms like OpenAI, Google DeepMind, and Anthropic have controlled the market. However, DeepSeek’s cost-effective training strategy could democratise AI development, enabling more countries and companies to build their own models. 

This shift could weaken the monopolistic hold of U.S. firms, as more AI models emerge worldwide, each tailored for specific needs. 

DeepSeek’s success signals a shift in AI competition, proving that cutting-edge AI can be developed at a fraction of the cost typically associated with American models. While OpenAI and other U.S. firms still hold a quality advantage, the decreasing cost of AI development in China could disrupt the industry’s economic model. If AI becomes more widely accessible and proliferates outside the U.S. ecosystem, America’s technological and geopolitical dominance could be at risk. 

The U.S. must take proactive steps to maintain its AI leadership while acknowledging that China’s AI capabilities are advancing rapidly, with DeepSeek at the forefront of this transformation. 

Bill Owens, Managing Director of Veracity, has decades of experience in global business consulting and technology and is committed to raising digital literacy levels of business leaders across Australia. Bill often presents to Boards and senior leaders on data governance, privacy, AI, and cybersecurity to help leaders feel at ease discussing and making decisions about tech and IT.

Continue the conversation on LinkedIn at Digital Literacy for Leaders.