A major tremor shakes the tech world as Chinese start-up DeepSeek unveils a revolutionary AI model developed at low cost, sending shockwaves through the financial markets.
Disruptive innovation at a low price
DeepSeek is hitting hard with its new DeepSeek-V3 model, an intelligent assistant that rivals giants such as ChatGPT. The innovation lies less in its performance than in its development cost: between $5 and $6 million, a derisory sum compared to the massive investments made by the American giants. The announcement sent shockwaves through Wall Street, with spectacular falls in technology stocks: Nvidia (-17%), Broadcom (-17%), Oracle (-14%) and Cisco (-5%).
Open-source as a weapon of massive disruption
DeepSeek’s strategy is clear: democratize AI by making its model accessible and modifiable under an open-source license. This approach is reminiscent of the historic successes of Firefox and Android, which revolutionized their respective sectors by lowering barriers to entry.
A paradigm shift in the industry
This breakthrough challenges the “bigger is better” dogma that has dominated the AI industry until now. If DeepSeek’s model proves successful, the huge computing infrastructures currently under construction could find themselves oversized, threatening the entire semiconductor and cloud computing ecosystem.
Major geopolitical implications
The emergence of DeepSeek is part of the Sino-American technological rivalry. Based in Hangzhou and headed by Liang Wenfeng, the company offers an alternative to American technological restrictions. This frugal approach could benefit not only China, but also other regions such as Europe, traditionally limited in AI infrastructure.
Still some grey areas
Despite the enthusiasm, experts temper the euphoria by pointing to a likely underestimation of actual costs. Expenses linked to data cleansing, network equipment, energy and human resources have reportedly not been taken into account in the announced calculations.