From Shanghai, Yereth Jansen informs us about the latest developments in the field of artificial intelligence. Naturally, this includes the recent introduction of DeepSeek and the reactions from the high-tech industry. As the AI market has been significantly shaken, a reassessment of the European position is urgently necessary.
The world has been captivated by the AI revolution for several years now. ChatGPT was launched on November 22, 2022, over two years ago. Since then, the "magnificent seven" – America’s largest technology companies – have jumped on the AI train en masse, and tens of thousands of AI startups have emerged worldwide. For a long time, the United States was the frontrunner in this race, benefiting from its enormous market, high GDP, and the determination of venture capitalists.
In January 2025, something remarkable happened in the world of artificial intelligence: the relatively unknown Chinese AI company DeepSeek released its latest model, called Deep Seek R1. Within a few days, it reached the number one position in the Apple App Store – not only in China but globally. What made this achievement even more remarkable was that DeepSeek’s model was not only powerful and significantly cheaper than other models, but it was also trained with hardware that was considered inferior compared to the advanced chips used by American companies like OpenAI and Google.
In comparison, OpenAI’s CEO, Sam Altman, has indicated that the costs for training GPT-4 exceeded $100 million. DeepSeek, on the other hand, claimed to have trained its R1 model for only $6 million. This significant cost saving was made possible by more efficient model architectures and the use of less advanced hardware.
Price Tag Under Discussion
This revelation had immediate consequences for the market. Nvidia, a leading supplier of AI chips, saw its share price drop by 17%, resulting in a loss of $589 billion in market value. This loss reflected investors’ concerns that demand for expensive Nvidia chips might decrease if more efficient AI models, such as those from DeepSeek, could be developed with less advanced hardware.
However, the mentioned $6 million is a subject of debate. Critics argue that this estimate only includes the direct costs of the latest training round and does not account for the full development costs, such as previous experiments and infrastructure investments. Some analysts estimate the total costs for DeepSeek R1 – including hardware, data acquisition, and failures during previous training runs – to be many times higher.
Moreover, DeepSeek is backed by High-Flyer, a Chinese hedge fund that uses AI for investment strategies. This financial model implies that DeepSeek may be less reliant on direct revenue from its AI models, unlike companies like OpenAI, which need to raise capital from the market.
Corporate Espionage?
The success of DeepSeek has also brought other controversies with it. OpenAI accused DeepSeek of using its models to train its own system, a practice known as “distillation.” In this process, a smaller model is trained to mimic the output of a larger, more advanced model. OpenAI claimed to have evidence that DeepSeek used its models in this way, which would be a violation of the terms of service and potentially infringes on intellectual property rights.
While the United States continues to view AI as a strategic power tool and Europe struggles with how to maintain its sovereignty in technology, DeepSeek demonstrates that progress may not solely be determined by brute computing power or the thickness of the wallet. This raises a pressing question: how does this change the balance of power in the global AI sector?
Transforming AI Landscape
DeepSeek operates within a context of unprecedented advancements in the Chinese AI industry. Baidu’s Ernie and ByteDance’s Doubao primarily target consumers, while Alibaba is developing a versatile ecosystem with its Qwen series. These Qwen models are also accessible via platforms like Hugging Face. Unexpectedly, on March 14, 2025, DeepSeek and Alibaba announced a collaboration, indicating a possible new phase in the market. Instead of fierce competition, there seems to be a consolidation taking place where various players specialize and form strategic partnerships.
This has direct consequences for the West. The notion that Chinese AI companies are merely “copycats” has long been outdated. DeepSeek, along with Moonshot AI’s Kimi and other emerging players, is innovating in a way that forces Western companies to make their own models more efficient and cheaper. The question arises: how long can OpenAI maintain an advantage if a Chinese startup delivers the same quality at a fraction of the cost?
Supply Chain Under Pressure
The cost savings don’t just come from smarter software. It is also a sign that China’s strategy of becoming less dependent on Western semiconductors is beginning to pay off. Nvidia’s most advanced AI chips are no longer being exported to China, but instead of waiting for a political breakthrough, DeepSeek simply switched to less powerful, locally available alternatives. And successfully so.
This also has consequences for companies like the Dutch ASML, which is still struggling with export restrictions for its advanced chip machines. China is heavily investing in the development of its own semiconductor technology, and if this trend continues, the West could find itself in a difficult position. The idea that technological dependence would slow down China seems increasingly untenable.
A European Wake-Up Call
For Europe, which is struggling to determine its position in the world of artificial intelligence, the rise of DeepSeek is both a warning and an opportunity. The recently announced €200 billion InvestAI program is intended to counterbalance the dominance of the US and China. However, as long as it’s unclear how this amount will be made available and how it will be implemented, the impact remains uncertain. DeepSeek shows that it’s also possible to break through without billions in investments from venture capitalists or governments.
Moreover, this development forces European companies to look beyond familiar defense mechanisms like regulation and ethical guidelines. Of course, responsible AI usage remains crucial, but as technological leadership diminishes, Europe must also focus on implementation power and market relevance. If it doesn’t, it risks not only falling behind but also becoming structurally dependent on ecosystems that are determined elsewhere. Now is the time to show that even a more risk-averse investment climate can lead to groundbreaking innovation – but it needs to happen for real.
Outlook: The Future of AI Beyond Language Models
Perhaps the biggest question is not how DeepSeek is changing the market, but what the AI race itself means. Large Language Models (LLMs) are impressive, but far from a complete imitation of human intelligence. They predict words, not worlds. The idea that we can build a comprehensive system with artificial intelligence that understands and simulates reality remains science fiction for now. And that raises a fundamental question: is the ultimate goal of AI to achieve artificial general intelligence and digitally replicate humans? Or is it just a tool that, as DeepSeek shows now, rewrites markets and power dynamics before we even truly understand where it’s leading us?
Although recent developments in LLMs are impressive, it’s important to recognize that this is just one step in the evolution of AI. The next phases could involve better generation and editing of videos, sound, and even dynamic 3D worlds. Furthermore, AI systems could contribute to simulations in biology, chemistry, and physics, where they model complex processes at atomic or even quantum levels. This ongoing growth within the AI industry suggests that we are only at the beginning of what’s possible, with applications extending far beyond generating text and images.
Conclusion
DeepSeek is not just a Chinese startup competing in the AI race. It is a symbol of a broader shift: a world where technological supremacy is no longer exclusively in the hands of Silicon Valley. This is changing not only how AI models are developed but also how we think about software, hardware, and dependence on specific infrastructures.
For Europe and the US, this means a new reality. The AI revolution seemed for a long time to be dominated by a few American tech giants, but China is proving that it can produce companies that can compete with them on multiple levels – not just in raw computing power, but also in strategic efficiency. While the US becomes increasingly defensive in its AI policies and Europe struggles with regulation and investment, China has set up a model in which AI development is closely intertwined with industrial and geopolitical goals.
What this means for companies like ASML, Nvidia, and OpenAI is still uncertain. But one thing is clear: DeepSeek has sent shockwaves through the industry, and the aftershocks will be felt globally – especially for those who think AI innovation remains an exclusively Western game.
About the Author
Yereth Jansen is an entrepreneur and strategic advisor based in Shanghai. As a co-founder of China Insights Weekly, he analyzes the latest trends in China’s tech and business world weekly. He is also the CEO of Darling Advertising + Design in Shanghai and advises international companies on branding, digital transformation, and market strategy in China. With over 15 years of experience in cross-cultural business development and a strong focus on AI and platform economies, he helps companies worldwide learn from China’s technological innovations.
