Last week, markets were rocked by the news that DeepSeek, a recently emergent Chinese company, has achieved a significant breakthrough in the artificial intelligence domain. Now the most downloaded app in Apple’s App Store, surpassing ChatGPT, DeepSeek claims to have developed an advanced neural network capable of outperforming existing models, including OpenAI’s ChatGPT, in natural language processing tasks. According to the company’s announcement, their AI system has achieved unparalleled accuracy in understanding and generating human-like text, making it a formidable competitor in the AI race between the US and China.
DeepSeek’s innovative approach involves a novel architecture that enhances the AI’s ability to learn from vast datasets while maintaining efficiency and reducing computational costs to a fraction of those experienced by other companies like OpenAI. This led to a collapse in tech-related stocks, and in particular in NVIDIA share prices, as Nvidia provides the expensive last-generation CPUs that are considered to be essential to performing the most sophisticated large language models (LLMs). On Monday, Jan. 27, 2025, the Nasdaq Composite dropped by 3.4% at market opening, with Nvidia declining by 17% and losing approximately $600 billion in market capitalization.
OpenAI claimed that DeepSeek has stolen its dataset and technology, and that the company has probably exaggerated the savings that achieved with its new architecture. Additionally, customers found that DeepSeek may be subject to censorship, as the system prefers not to answer questions regarding the actions of the Chinese government, for example about the Tiananmen Square protests in 1989. DeepSeek responded to these criticisms by highlighting the ethical considerations embedded in their AI development process, ensuring that their technology adheres to strict guidelines to prevent misuse and promote transparency. As in most cases, the truth likely lies in the middle.
What really matters at this stage is that the AI battle between US and China, one of the key battlegrounds of Cold War 2, is in full swing, and, if anything, is intensifying. From the Chinese side, Alibaba has also entered the game, with the release of its own AI model, Qwen 2.5, which promises to push the boundaries of natural language processing, even those of DeepSeek, even further. Alibaba’s new release boasts state-of-the-art performance metrics and aims to integrate seamlessly into various applications, potentially reshaping the competitive landscape of AI development and deployment worldwide.
On the other side of the pacific, Jen-Hsun Huang, Nvidia’s CEO, was received by US President Trump, most likely for two reasons. First, to receive reassurances that the US government will continue backing the nascent AI industry in the US, protecting it from international competition and interference.
Second, Trump probably reiterated to Huang that the US has imposed a ban on export of critical US technology, including chips, towards its geopolitical rivals, including China.
We can be sure that further developments will accompany these in the coming months and years, until eventually further breakthroughs in quantum computing will probably revolutionise the AI landscape once again.
