Reconstructing a trillion-dollar crash: how a small Chinese start-up shook finance, business and politics
A company from a Chinese province is calling US AI leadership – and billions of dollars in investment – into question. Germany and Europe could benefit.
What it’s about
In late January 2025 a previously unknown start-up from the Chinese city of Hangzhou rocked the tech market like no company before: Nvidia lost 600 billion dollars in market value in a single day – more than any company in history. Meta called emergency meetings. Donald Trump put US industry on alert. The trigger: a language model called DeepSeek that, supposedly built on a tiny budget and without US high-end chips, caught up with the best models from Silicon Valley. We reconstructed how that could happen.
What DeepSeek does differently
DeepSeek is efficient. The model needs less compute, less memory, less training. Exactly that puts the business model of US tech companies in question: if AI of this quality can be built on a small budget, does Microsoft really need 80 billion dollars for AI data centres? Is OpenAI rightly a unicorn at a $157 billion valuation? Or is the AI boom mainly an infrastructure boom built on the wrong assumptions?
The bigger contest
The question is not only commercial but geopolitical. The US has done everything it can with export restrictions on high-end chips to slow China’s AI development. DeepSeek shows that the strategy isn’t working. Beijing has in parallel built its own AI strategy, leaning on open source, open weights and efficient training. A second AI ecosystem is emerging that competes with the American one.
Why Europe should pay attention
One of the most interesting aspects: DeepSeek is small. The team is manageable, the budget modest. If a Hangzhou start-up can do this, there is little reason why it could not also be done in Munich, Paris or Helsinki. France’s Mistral and Finland’s Silo AI prove that European AI champions are, in principle, possible. DeepSeek could shift the European debate from complaining back to building.
I wrote the full story together with Sabine Gusbeth, Felix Holtermann, Larissa Holzki, Christof Kerkmann, Lina Knees and Philipp Alvares de Souza Soares for Handelsblatt.