Will Chinese AI startup DeepSeek change the outlook for the tech sector?

 

“The boom in artificial intelligence (AI) investment in 2024 was one of the most important drivers of the markets. A new player that appeared out of nowhere, Chinese startup DeepSeek, claiming to have developed a model that rivals US companies’ AI for a fraction of their cost, prompted investors to reconsider their assumptions about the value of tech companies. As a result, markets started this week with a particularly deep drop, and chipmaker Nvidia lost half a trillion US dollars in market value in just one day.

The rise in tech company prices seemed unstoppable. The so-called Magnificent Seven technology giants generated about half of the total return on US stock indexes last year.

Companies in the Seven, such as Meta, Google and Microsoft, developed increasingly powerful and intelligent AI models and sought to offer their end users ever smarter tools. Training AI models is a very computationally and energy-intensive activity, requiring thousands of powerful and expensive processors. Therefore, energy companies and data centers also got a chance to bask in the rays of the investment boom.

Recently, technology giants have announced investment programs worth hundreds of billions of dollars in order to get ahead in the AI ​​race. Huge investments in improving models have turned processor manufacturer Nvidia into one of the most valuable companies in the world. The expectation was that further development of AI would require more and more computing power, processors and energy, which is why shares of semiconductor manufacturers are extremely expensive.

Just as good, but much cheaper

Chinese startup DeepSeek has released a new large language model (LLM), which in its “intelligence” seems to be on par with today’s leading models. After this step, DeepSeek has received a lot of interest - on Monday their model became the most downloaded app in the Apple app store. In addition, the presented model is open source, so anyone can download it and adapt it for their own needs.

The appearance of a new competitor would be quite a bit of news in itself, but the developers announce that they managed to create AI for just a few million US dollars. Using clever training techniques, they did not need to use thousands of the most advanced processors and process a huge amount of data.

So far, many market participants are skeptical about the announced price of the model, but analysis of the model itself confirms that the innovative techniques used allow for much more efficient use of available data analysis resources and training target models cheaper. Such news is forcing the market to rethink all assumptions about the further development of AI.

Three new assumptions have emerged at the moment:

1. AI does not necessarily require gigantic investments, it is possible that it will be more widely available, at an affordable price.

2. Models can be improved without the most advanced processors. US semiconductor trade restrictions and Western technological dominance may not be as unshakeable as previously thought.

3. Applying more efficient training techniques to existing computing power may lead to even greater qualitative breakthroughs.

A new player provoked a stock sell-off

Stock markets, especially in the US, reacted painfully to the news. On Monday, January 27, Nvidia’s share price fell by almost 17 percent, wiping half a trillion US dollars off its value in a single day.

Shares of other major players in the semiconductor supply chain, ASML of the Netherlands and TSMC of Taiwan, reacted similarly to the news. The US technology index Nasdaq 100 fell more than 3 percent in a single day.

It has been argued for some time that the US stock market, and the technology sector in particular, is too expensive for its financial performance. It is not surprising that technology stocks are even more sensitive to news because of this. The DeepSeek news also acted as a catalyst for a price correction that had been brewing for a long time.

A bubble burst or a new beginning?

The initial reaction to the breakthrough that allows for cheaper development of AI models is absolutely negative for the market leaders. The prevailing opinion is that there will be no need for so many processors, so semiconductor manufacturers will perform worse. In addition, the democratization of AI development means that the current market leaders will soon feel increasing competitive pressure from both China and other new market participants.

On the other hand, the cheapening of technology may mean that the AI ​​market will not be dominated by a few large players. Personalized AI solutions will be cheap enough for many businesses and individuals to implement. This may mean that the circle of customers for the same Nvidia will expand, since until now it has been dependent on orders from just a few large players.

It is possible that even in the new conditions, the same technology companies will dominate. By applying DeepSeek's innovations and their own computing capabilities, they will create much higher-quality and more widely applicable AI models. In addition, high spending on data centers for training AI has been dragging down the financial indicators of the technology sector for some time. More efficient use of available resources can be extremely beneficial for the results of technology giants.

So far, the only thing that is clear is that the end user of AI models will definitely win the technological race - cheaper and smarter AI models will allow the technology to spread faster and in the long term can serve the most outside the technology sector.

Investors will be cautious about the technology sector for some time, so the local sector correction may be prolonged. However, a technological breakthrough is undoubtedly great news for the markets, but the question remains which market participant will become the ultimate winner of the AI ​​race.”