DeepSeek, a game changer or a buying opportunity?

The news from DeepSeek hit the stock markets hard on Monday. Shares of companies related to any form of AI developments fell sharply. This included not only semiconductor stocks from the IT sector, such as Nvidia, ASML, ASM International, TSMC and Broadcom. Also stocks from the industrial sector, such as Schneider Electric, Trane Technologies and Legrand, and the utilities sector, such as Siemens Energy and Constellation Energy, were hit. Data-center stocks such as Equinix and Digital Realty also lost value, as their share price had risen rapidly on expectations that more data centers would be needed.

DeepSeek’s announcement stirred markets as the Chinese startup claims that its open-source large language model (LLM) was built in just two months, costed less than USD 6 million to build and uses chips from an older generation. However, these claims could not yet be verified independently. DeepSeek’s AI app is currently the most popular and most downloaded free app in app stores,  and does not contain commercials.

It is quite impressive how DeepSeek was able to build this new model, despite strict semiconductor restrictions that the US imposed on China. Chinese companies do not have access to the most powerful and advanced chips of Nvidia, such as Hopper 100 and Blackwell chips. DeepSeek claims that it has trained its AI-model with ‘older’ Nvidia H800 chips. DeepSeek might have ‘trained’ a smaller model to get smarter on just a certain part. This is cost efficient and energy efficient.  

The release of Deepseek is raising three main topics for investors to keep an eye on:

• First, if it is possible to create successful models for substantially less money, will the large investments in chips and other AI requirements continue? One example is the Stargate joint venture recently announced by US President Donald Trump to invest USD 500 billion in AI. Or will companies reconsider and scale down these significant investments, or even postpone them?
• Second, an important assumption underpinning the strong performance of companies such as Nvidia was that the strongest and newest chips would be essential to roll out AI tools. Is this still the case, or can it be done with cheaper and less efficient chips, while challenging the monopoly position of companies like Nvidia?
• Third, it demonstrates that the restriction imposed on China has not stopped them from developing competitive AI models. What does it mean for US exceptionalism if companies in other countries are able to seriously compete with the big US tech companies?

It is too soon to answer these topics. They appear, however, when the US tech sector is “priced near perfection” with a price-earnings ratio trading at around 30 - significantly higher than other US stocks. In the past decade, investors rerated US tech considerably, as they were willing to pay for strong revenues and earnings growth with strong visibility. The news, if confirmed, will certainly cause investors to look with greater scrutiny at AI investments.

Is DeepSeek a gamechanger or could it present a buying opportunity for some companies that saw their shareprice drop significantly on Monday? This is something all investors are struggling with, but not enough is known currently to answer. The news has surprised markets and statements issued by DeepSeek need to be verified and might not be completely reliable. At least, it warrants further investigation.

In case of high volatility and many uncertainties, we believe the best strategy is to not make large changes in the portfolio for now. We therefore stick to our strategy and remain overweight towards the IT sector. We will continue to look for more clarity, starting with the earnings announcements of companies such as ASML, Microsoft and Meta Platforms later this week.

Joost Olde Riekerink
Equity Research & Advisory Expert

Olivier Raingeard
Global Head of Equity Strategy

Richard de Groot
Global Head Investment Center