Monday morning January 27, 2025, the NASDAQ financial world was in a mini -fusion on the headlines concerning a Chinese company called Deepseek, which announced major IA breakthroughs – even with a limited budget. For example, An ETF Nvidia lost 51% of its value in one day.
So why did Deepseek cause such a reaction?
According to Reuters“The launch by the Chinese startups Deepseek of its latest AI models, which, according to her, are on peer or better than advanced models in the United States to a fraction of the cost, threatens to upset the world of Technology.
“The company drew attention to the world circles of AI after writing in an article last month that the training of Deepseek-V3 required less than $ 6 million in computing power of the Nvidia H800 flea.
“The AI assistant of Deepseek, powered by Deepseek-V3,” exceeds Rival Chatgpt to become the best rated free application available on the Apple App Store in the United States. … ”
You can find out more about Deepseek in this video:
The first lessons learned from the ads of the deep AI
1) Investing in artificial intelligence companies will continue to be a roller coaster. This Reality, as Nvidia investors have learned it even before Deepseek announcements, applies to financial investments. There are many companies in this space looking for attention and time investments and end and business user drivers. In the years to come, there will be many claims and counterators by many companies that want you to jump to their solutions because they offer revolutionary results or different prices. The challenge will be to know when to change products / companies, or to add more products as a tool that you are currently using.
2) The American-American companies will present many challenges and claims to unpack for individuals and governments. Beyond the data and security problems (which we will cover in n ° 3), there will be allegations and arguments on different sides, in the same way as Tiktok beats youtube, Facebook, Instagram and others for the Be careful with global users. We are in a global generative AI race (Genai) for supremacy.
As CNBC pointed out“Chinaseek ai dethrones chatgpt on the App Store.”
“Monday, Chinese artificial intelligence Startup Deepseek took over the coveted spot of Rival Openai as the most downloaded free application in the United States AppleApp Store of the app, destroy the Chatppt for the AI assistant of Deepseek. Global technological actions were sold and were at the rate billions in market capitalization. …
“The buzz around Deepseek began to steam earlier this month, when the startup published R1, its reasoning model that rivals the O1 of Openai. It is open-source, which means that any AI developer can use it, and has gone bankrupt at the top of application stores and Industry rankingsUsers rent its performance and reasoning capacities. … ”
3) There are many data, cybersecurity and even the national security problems to be resolved with Deepseek. Here are some of the articles on this security subject in the past few days:
CBS News: “”Deepseek AI raises national security problems, say the US officials“-” “in the direction of the president, the NSC and other members of the US government work in many ways to respond to concerns concerning AI, China and data security,” said the spokesperson for The agency, Brian Hughes, in CBS News in an email. “As the president has pointed out, American policy is to ensure that the United States runs the world in AI.” “”
New York Post: “”Deepseek Ai Collection, stores American user data in China – arousing strangely similar national security problems that have forced Tiktok repression“-” “What distinguishes this context is that Deepseek is a Chinese company based in China,” said Angela Zhang, law professor at the University of South California, focused on Chinese technological regulations.
“” It raises the question of whether data collection such as IP addresses and typing models could represent a national security threat, “added Zhang.”
Reuters: “”Deep sensitive data exposed to the web, says the Cyber Cabinet“-” The cybersecurity company based in New York, Wiz, says that it has found a mine of sensitive data from the startup of Chinese artificial intelligence In depth exposed inadvertently to the open Internet.
“In A blog article, WIZ Wednesday, WIZ said that Deepseek infrastructure analyzes showed that the company has accidentally left more than a million lines of available data guaranteed. These included digital software keys and discussion newspapers that seemed to capture prompts sent by users to the company’s free AI assistant. “”
Time: “”Why Deepseek triggers debates on national security, just like Tiktok” -” However, other confidentiality experts argue that Deepseek data collection policies are not worse than those of its American competitors – and fear that the rise in the company is used as an excuse by these companies to call for deregulation. In this way, the rhetorical battle on the dangers of Deepseek takes place on lines similar to those of the insimbo Tiktok Banwho deeply divided the American public.
“” There are confidentiality and data security problems completely valid with Deepseek, “said Calli Schroeder, AI and human rights leaders at Electronic Privacy Information Center (EPIC). “But all are present in American AI products too.” »»
4) Companies will have an even greater struggle with “Bring Your Own IA” (Byoai) or “Shadow AI” because of these frequent AI and free online announcements. I have raised this subject several times in the past two years, and the trend should speed up with new “hot” AI tools out of China and elsewhere. See these detailed articles on this subject: “Generative ai Guar-editor: how to approach the shadow ai” and, “Has generative discreetly inaugurated a new era of shadow on steroids?“”
5) AI disruption will be commonplace, as well as the claims that will be closer to Deepfakes than the IA breakthroughs – so do your homework and try before buying.
See this Venturebeat article To find out more: “Deepseek and its R1 model do not waste time rewriting the rules of cybersecurity in real time, with everyone, from startups to business suppliers piloting integrations to their new model this month .
“R1 has been developed in China and is based on learning pure strengthening (RL) without supervised fine adjustment. It is also open source, which makes it immediately attractive for almost all cybersecurity startups which are all on the ‘open source architecture, development and deployment.
See also Yahoo: “Analysts offer a blunt vision of panic in depth“-” But the veteran of Wall Street and portfolio director Chris Versace recently highlighted That his team tried to first avoid a “shooting, to ask questions later” when evaluating the impact of Deepseek on leaders in the technological sector.
“I must say that it is a one-year startup, and it goes with the head with some of the best and the brightest minds,” he noted, expressing a certain skepticism that the new company will continue to push NVDA Stock Down. “”
Final reflections
While financial week was coming to an end on Friday afternoon, January 31, 2025 (when I write this), the Nvidia stock and the Nasdaq on the whole had rebounded, but not always near their heights All time. The global technological sector had calmed down and Panic of AI had calmed down for most market observers.
Nevertheless, this experience with Deepseek predicts more more bumpy bullets on the roller coasters of AI which have implications for Wall Street, rue Main, private sector companies and government companies.