The small world of AI seemed well established, with clear prospects and a well-defined roadmap. But then, patsy! A small grain of sand has changed all that. As soon as it was launched, China’s new generative artificial intelligence (AI) solution DeepSeek shook up the American technology industry. Developed by a small team with limited resources and without recourse to Nvidia’s high-end chips, this open-source AI boasts performance comparable to that of OpenAI’s models. Its transparency and accessibility contrast with the proprietary approach of its Western competitors, sending shockwaves through the financial markets.
The impact was immediate: Nvidia suffered a historic loss of $589 billion in market capitalization in a single session, an all-time record. The entire semiconductor sector and the digital giants, who had invested hundreds of billions in closed solutions, were also affected. In the space of a single day, almost $1,000 billion in value was wiped out, revealing the excessive overvaluation of certain players. This correction comes just days after US President Donald Trump announced the launch of Project Stargate, a program to invest up to $500 billion in AI infrastructure in the United States.
DECOUVREZ DANS CET ARTICLE
An unsettling success
The response to this breakthrough was swift. DeepSeek was immediately accused of plagiarism, intellectual property infringement, ideological bias and problems of data security and confidentiality. These criticisms, though predictable, could well be turned against American companies, who also rely on massive databases and similar learning techniques. But polemics aside, the arrival of high-performance, open-source, low-resource AI opens up unprecedented prospects, particularly in the field of business intelligence.
A revolution in business intelligence?
DeepSeek offers a number of advantages that could transform the way companies and institutions conduct their information and decision-making intelligence.
🔹 Flexibility and independence
Unlike proprietary solutions such as GPT-4 or Claude, DeepSeek relies on open source, enabling companies to integrate its models directly into their infrastructures, without relying on third-party suppliers. This approach guarantees total control over the data analyzed, and offers greater flexibility in adapting algorithms to the specific needs of each organization.
🔹 Security and digital sovereignty
One of the major challenges of business intelligence lies in protecting sensitive data. As DeepSeek can be hosted in-house, it guarantees total confidentiality, a strategic asset in a context where digital sovereignty is becoming a priority for many economic and political players. This independence reduces the risk of industrial espionage and limits exposure to foreign regulations.
🔹 Low cost and accessibility
Current generative AI models require colossal investments: developing GPT-4 cost almost $100 million. DeepSeek, on the other hand, reduces these requirements by a factor of 10, making these tools accessible to SMEs and governments wishing to develop their own sovereign solutions. This democratization of AI could rebalance global competition.
A historic opportunity for Europeans?
As usual, Europe is absent from this technological revolution, content to be a spectator and dependent on American players. Yet DeepSeek represents a unique opportunity to reshuffle the deck and relaunch a European AI industry on new foundations. Its openness means that it no longer belongs to China alone, but becomes a global technological heritage that everyone can appropriate and develop. European companies and institutions therefore have every interest in seizing this opportunity to build a credible alternative to the dominant solutions, without being at the mercy of American or Chinese regulations.
The upheaval caused by DeepSeek is a strong signal that the era of closed AIs, ultra-dependent on Big Tech, may be over. It remains to be seen who will best capitalize on this.
Alexandre Sonnet
President of Humind
