Nvidia, which grew up at the top of the stock market by selling computer chips that promote the world’s artificial intelligence boom, has been treated with a harsh control of reality by a small Chinese company that showed it could do more less of what Nvidia does.
On Monday, Nvidia shares were plunged more than 16 percent after the company, called Deepseek, indicated that it could train a advanced system of it with a portion of Nvidia chips that were used in the past by Openai, manufacturer Chatgpt.
The release of Deepseek challenged a technology industry consensus that, to build larger and better systems of it, companies will have to spend billions and billions of dollars in new databases. At the center of those databases would be the only thing that, perhaps so far, no project he could do without: a large cache of Nvidia chips.
Silicon Valley, with some ratings, controls 90 percent of the market for specialized chips used to build it. She has had a wonderful run since Openai released the chatgt in the end, peaking with $ 3.62 trillion in November.
But Deepseek’s apparent progress has shown that appetite for nvidia chips may not be as boundless as some had imagined it just a week ago. While Nvidia is still in an enviable position – there is little competition for its chips – companies that have purchased its technology can slow down their expenses.
“Previously, he was bigger, better, faster. Larger chips equal to him, “said Patrick Moorhead, Moor Insights & Strategy chief executive, a technological and semiconductor research firm. True for Nvidia and if people will need as many chips as possible in the future. “
The release of Deepseek also withdrew the shares of other semiconductor companies, including Broadcom, Micron Technology and Taiwan semiconductor production Inc.
The arrival of Deepseek has crystallized a concern that was already casting a shadow over the Nvidia business. At the end of last year, he began to warn that improvements in chatbots were slowing down. They had previously relied on a simple formula to provide advances: by removing as much data as possible from the internet and pushing into large linguistic models-technology that empowers chatbots in increasingly strong computers.
But that concept, which is known in the industry as laws for scaling, has begun to fall out of favor because technology companies are finishing data. This has led companies to start experimenting with new techniques to continue improving their systems. She has also prompted questions from Nvidia investors on the consequences for her business.
At the January -Jann Huang CES trade show, Nvidia’s chief executive, said the new techniques were “running the big demand for Nvidia chips”. He said companies are using Nvidia chips in power models that make more justification in databases, which means there will be more demand for its technology, no less.
In a statement on Monday, Mylene Mangalindan, Nvidia spokesman, said Deepseek shows that those new techniques are working. She added that “Deepseek is an excellent progress of him.”
But investors have been seeing to see if changes in the way he is being made would kill the Nvidia business, and Deepseek’s success speaks of uncertainty if it will be cheaper to build and provide systems.
“This was something we worked on,” said Daniel Newman, chief executive of Futurum Group, a technology research firm. “Everyone has pursued this goal because the training costs of it were very high.”
Not everyone is convinced that raising Nvidia is stagnating. Its stock has been unstable and collapsed in August with more than 10 percent on reports that the company will delay the transport of its newest artificial intelligence chip. Later withdrew.
In a note for investors on Monday, Stacy Rasgon, a semiconductor analyst at Bernstein Research, said Deepseek spent more money to build his system than he claimed. He added that being able to build more systems he more efficiently needs to mean more demand because more companies can allow to invest in them.
Panic over the weekend, Mr. Rasgon said, “seems overloaded.”
Changes in the value of Nvidia are the last will of how it continues to increase the stock market. Last year, Microsoft replaced Apple as the world’s most valuable enterprise for its early push in June, Nvidia grew for both companies to search for the crown.
Now, Apple has returned to the lead after leaving his system called Apple Intelligence for the iPhone. But there are signs that may not hold that position for a long time. This month, it disabled one of the signature skills – the collection and summary of news announcements – as clients and companies complained that its software was abusing news reports.
“There was no winner here,” said Mr. Newman. “The tools have been just okay. But if everyone can make better models at lower costs then people can start to adopt he “