The publication season of swing From the largest technology companies in the United States begins next Wednesday (29) under pressure from a new competitor from China. On Monday (27), even before the opening of the market, the influence of the chat of generative AI yes startup DeepSeek It led to a drop in shares of the most representative companies of the S&P 500 index.
The market reaction at the beginning of the week is mainly due to the signaling that artificial intelligence models can be created cheaper and efficiently, precisely at a time when large companies have drew rivers of money to expand their infrastructure to date CENTERS.
On Wednesday (29), Microsoft (), Goal () and Tesla () inaugurate the season of balance sheets of the so-called “Magnificent Seven“Seven largest companies in the sector in the USA. The next day, Thursday (30), Apple () presents its results. The following week is Alphabet’s turn () and Amazon () make the ads.
Participate in free dividend manual training and discover the simple and powerful strategy to live in income.
“We had already begun to realize a greater concern of investors regarding the return on the high investments made in AI in the last quarters,” says the global strategist of the XP Research, Paulo Gitz. “Now, with this Deepseek event in mind, Big Techs will have to detail how they intend to be profitable with their projects in the area, at the risk of, if the explanations are not convincing, to be punished by the market.”
Who closes the group’s list of balance sheets is the company of the moment when it comes to artificial intelligence: Nvidia (). It is she who is at the center of the United States and China dispute for the semiconductor industry and cannot sell many of her newest generation products to the Asian country.
Apparently, however, access to the avant-garde GPUs of the American chip developer did not prevent the Deepseek-R1 model from being able to achieve similar performance to market leaders such as OpenAi and meta. “Nvidia, due to the high expectations of sales of its new product, Blackwell, may have estimates reviewed down,” says Gitz.
Continues after advertising
As in 16% around 4 pm on Monday.
For the magazine Natureexperts estimated that the hardware rental cost to train Deepseek-R1 was approximately $ 6 million, compared to over $ 60 million for Llama 3.1 405b of the goal, whose demand for computing resources was 11 times bigger.
Gitz points out, however, that DeepSeek used existing models as the basis for its development: “Their model does not take the need to develop and improve the current LLMs [sigla em inglês para modelos de linguagem ampla]. It is an industry in still initial stadiums of evolution and, like this event, many others will come. ”
Continues after advertising
The arrival of DeepSeek happens just the week when the less enthusiastic advertisements for Big Techs results in years begin. Second, the profits of the seven largest technology companies in the world are designed to grow 22% in the fourth quarter of fiscal year, the lowest evolution since the first quarter of 2023.
“There should be no visible direct impact on companies’ results. The focus will be on the answer to management Regarding the possibility of being more efficient in investments, ”says Gitz.
