Nvidia crashes

China AI DeepSeek sends Wall Street into a panic

DeepSeek is based on a significantly smaller database than the competition and is also open source.

(Bild: APA Österreich Bild/AP/Andy Wong)

A chatbot from China shocks Wall Street: panicked investors have caused the stock market value of chip company Nvidia to plummet by almost 600 billion dollars in one day. This was triggered by the realization that AI software can perhaps be trained with much less computing power than previously thought - at least according to an announcement to this effect by Chinese start-up DeepSeek.

DeepSeek, a supposedly uncomplicated and cost-effective AI model from a Chinese start-up, is causing a stir in the industry. If it turns out to be true that the cost of training artificial intelligence (AI) is only six million dollars instead of the usual more than 100 million dollars, this would be a breakthrough and would give an additional boost to the spread of this technology, according to portfolio manager Jon Withaar from asset manager Pictet.

"We don't know all the details yet and nothing has been 100% confirmed," added Withaar. But in the US, DeepSeek's AI has already overtaken ChatGPT as the most popular app in Apple's App Store. Venture capital investor Marc Andreessen therefore described the launch of the Chinese provider's AI application "R1" as a "Sputnik moment". He was alluding to the first satellite that the then Soviet Union launched into orbit in 1957. Following this, the USA and other Western countries intensified their space research.

"DeepSeek R1 is one of the most amazing and impressive achievements I've ever seen - and as open source, a great gift to the world," Andreessen wrote on the short messaging service X. With open source software, the program code is freely accessible and can be modified by anyone. The best-known examples include the "Linux" operating system and the "Firefox" Internet browser. Andreessen played a key role in the development of the "Firefox" predecessor "Netscape Navigator".

In contrast to ChatGPT, for example, which uses a large language model, DeepSeek uses many smaller models and is therefore more efficient.

(Bild: APA Österreich Bild/AP/Andy Wong)

Smaller database than the competition
According to DeepSeek, the new AI is based on a significantly smaller database than the competition. The company used Nvidia's AI chips from the "H800" model series to train the "R1" predecessor "V3", which was released a few weeks ago. This is a slimmed-down version of the processors available in the West, which is not subject to the US embargo on technology exports to China. Nevertheless, both "V3" and its successor "R1", which is only a few days old, could compete with rivals such as ChatGPT from OpenAI or "Llama" from Facebook parent company Meta.

Western technology stocks under pressure
According to official Chinese documents, DeepSeek is controlled by Liang Wenfeng, the co-founder of the hedge fund High-Flyer. The fund announced in March 2023 that it wanted to make a new attempt to develop artificial super-intelligence. Such an Artificial General Intelligence (AGI) can perform complex tasks without any human intervention.

Whether this is all exactly true is not known. There was speculation in the industry weeks ago that DeepSeek may have access to more Nvidia chips than it admits in view of US export restrictions. However, investors who have been driving Nvidia's shares higher and higher in recent months in anticipation of a future mega deal got cold feet on Monday. US President Donald Trump spoke of a wake-up call for American companies - and at the same time found it "positive" that AI could be cheaper.

Nvidia shares fell by almost 17%, erasing 589 billion dollars in market capitalization. Never before had a US company lost so much value in one day - although none had ever been as expensive as Nvidia with almost 3.5 trillion dollars.

What can the USA do with better chips?
But what does all this mean for the future? American companies have recently outdone themselves with announcements about how much money they want to invest in AI infrastructure. ChatGPT inventor OpenAI and several partners alone promised to invest 500 billion dollars in data centers in the coming years. At Facebook group Meta, founder Mark Zuckerberg promised 60 billion dollars this year alone. But is DeepSeek now showing that it will manage with significantly less computing power?

US experts would rather see it the other way around. The question is not whether DeepSeek can overtake current market leaders in the USA, emphasized X. Eyeé from the AI consulting firm Malo Santo. It is much more about how quickly the Chinese company's research approaches can be implemented. "If DeepSeek can develop this with old hardware, what can we do with newer hardware?" she asked on US broadcaster CNBC.

Doubts about the costs of DeepSeek
The tech industry will still "need a lot of chips", industry analyst Stacy Rasgon was also convinced. "It's clearly a panic today," he emphasized with regard to the stock market reaction. At the same time, Rasgon doubted the information from China: "They didn't train the model for five million dollars, that didn't happen."

Last week, the head of the AI company Scale AI, Alexander Wang, said that according to his information, DeepSeek had access to 50,000 H100 chip systems from Nvidia, but could not talk about them due to US export restrictions. The US government only allows Nvidia to sell chips to China that are less powerful.

Not only Nvidia affected
The DeepSeek shock also affected other shares. The shares of chip company Broadcom also lost around 17 percent. Energy companies, where investors had been speculating for months on good business in the supply of data centers, were hit even harder. Shares in Constellation Energy fell by a good fifth, while the price of Vistra plummeted by as much as 28 percent. Shares in Apple and the Facebook group Meta, on the other hand, rose on the prospect of cheaper AI.