How to invest in AI after the DeepSeek bomb

As Big Tech throws billions at advanced chips and giant data centres, along comes DeepSeek with a surprising alternative. An inexpensive and efficient AI model that challenges the established order and is making investors nervous. Has the AI industry come to a turning point?

For most big US tech companies, it was a Monday morning to remember. On 27 January, they woke up to the news that an unknown Chinese startup may have beaten them in their own game.

In just a few months, DeepSeek has developed an advanced AI model (R1) that easily competes with the models of established American tech giants. Without cutting edge chips and at a fraction of the cost.

Nvidia, the flagship of everything AI, saw USD 600 billion in market value evaporate in a single day: a 'world record'. Although the stock recovered afterwards, investors in Nvidia and other AI stocks were left reeling.

Now that the first dust has settled, investors are shifting their focus to the key question: how can you (re)position your portfolio in a landscape that can easily be turned upside down?

What consequences will DeepSeek have in the long term?

1. More money = better AI?

The DeepSeek model questions the idea that top performance can only be achieved by investing billions of dollars in huge data centres. This means generative AI might develop in a much faster and cheaper way than the expensive US big tech models have done so far.

DeepSeek has said that the model it has built cost less than 6 million dollars. The model is more or less similar to Open AI’s ChatGPT 4.0, which cost no less than USD 100 million to train.

A deliberate move or not: DeepSeek launched its model on the day Donald Trump announced Stargate with a great deal of fanfare. This is also an AI project with investments worth 500 billion dollars. (It actually took a week for the news to spread properly). Mark Zuckerberg later announced that hundreds of billions more were to be invested in AI. However, by developing advanced AI models with less advanced chips, DeepSeek has raised the question whether all those billions are actually going to pay off. And that is worrying investors.

2. Will US tech dominance continue?

Even if DeepSeek is soon overtaken again by US (or European?) competitors, this development does make it clear that the global AI race is heating up even more. US companies such as Nvidia, Microsoft, Meta and Alphabet are at the top today, but that doesn't mean they’ll stay in the lead forever. Today’s winners are not necessarily tomorrow’s winners. Investors should not be afraid to challenge what they believe. DeepSeek proves that innovation doesn't have to come from Silicon Valley.

3. High valuations: time to let off some steam?

Big Tech is expensive. Investing in US tech companies doesn't come cheap. The so-called price-to-earnings (PE) ratio of many big tech companies is higher than the historical average, which means investors are paying a significant premium for the expected growth. This is all the more relevant because the US tech industry has been the driving force behind the broader US stock market rise for years.

Because these stocks are priced so high, they can afford few slip-ups. Chinese tech companies, however, are trading at significant discounts due to geopolitical concerns. The emergence of DeepSeek could pique investors’ interest in undervalued Chinese tech companies and offer an alternative growth story.

4. Should everyone have their own model?

While some AI stocks were hit hard on 27 January, others – such as Apple – surprisingly had a lucky escape. The reason for this offers clues on where the industry could go. Everyone blamed Apple for missing the AI boat, but according to The Motley Fool, Apple’s decision to rely on third parties for most of its AI strategy suddenly sounds like the smartest approach.

5. Will open source be the way forward?

DeepSeek takes a different path from many US companies that are keeping their AI models private. DeepSeek uses an open source strategy. This means the source code and training methodologies are publicly accessible to anyone who wants to use or improve them. Meta did not emerge from the storm unscathed either, but the damage is limited for now. This was probably because Meta’s open source vision shows similarities to DeepSeek’s.

6. What are the consequences on the geopolitical stage?

DeepSeek has said it uses Nvidia’s H800 chips, a less advanced variant designed specifically to meet (older) US export restrictions. If this is the case, the startup has successfully built a model that can compete with GPT-4. According to the Financial Times, this raises questions about the effectiveness of US export restrictions. If China can develop advanced AI in this way, it may well shift the balance of power in tech. If US export restrictions fail to curb Chinese innovation, we can expect policymakers to explore new tactics, which may result in volatility in the semiconductor industry and other segments.

7. What impact will this have on other sectors?

The consequences for other sectors are not yet clear, but the sell-off of energy companies and dip in the copper price did give us some indication. If it turns out that data centres will not need as much power, this may affect the demand for energy sources. In recent years, the exponential growth of AI models was expected to lead to an explosive increase in demand for electricity, which has benefited energy companies and infrastructure players. However, if DeepSeek’s approach takes hold and AI systems become more efficient, some energy companies may see their growth prospects differently.

AI is still hot, but be (more) critical

The rapid rise of DeepSeek has illustrated once again that AI innovations are evolving at lightning speed. Every new player can really disrupt the market, which may lead to significant price fluctuations. Also bear in mind that we have only discussed generative AI here, which is only one aspect of AI as a whole.

At the same time, the underlying trend of digitalisation, automation and data analysis has remained unchanged. In other words, the big picture hasn't changed. AI continues to drive innovation and has a long path of growth ahead of it. If the cost of AI continues to fall, there is also a good chance that its use will increase drastically. In the past, we have often seen leaps in efficiency lead to more use rather than less. Steam engines, electricity and even fossil fuels followed a similar path: when they became cheaper and more efficient, overall demand skyrocketed.

For AI, this means fewer obstacles may result in much wider adoption, not only by large multinationals, but also by SMEs, public institutions, the services industry, education, everyday applications and so on. In practice, this may lead to growth for the entire AI industry (chips, servers, networks, software), even if individual costs fall. The lower the price, the more industries will embrace AI, which in turn will lead to new innovations and ideas.

What is the clever way to include AI in your portfolio?

• Stay calm and think long-term The AI industry remains promising, but volatility is inevitable. Historically, diversified investments in the tech industry are rewarding in the long term, but the road to get there can get bumpy.
• Diversify within AI Rather than focus on infrastructure alone, also include AI applications, in healthcare or education, for example. The real winners of the AI revolution may be companies that implement AI to solve problems.
• Diversify in tech rather than focus on AI alone By investing only in pure AI players, you run the risk of exposure to strong price fluctuations and hype-driven valuations. So think of broadening your exposure. Consider cloud computing, cyber security, robotics and automation and semiconductors. These segments benefit from AI indirectly without any of the risks associated with AI-specific players. Be careful not to concentrate all your investable assets in one sector. Instead diversify them across different sectors and asset classes (such as bonds).
• Use ETFs or funds for diversification ETFs offer broad exposure to AI without relying on a single player’s performance. They reduce the risk of stock market corrections. An actively managed fund can also be an interesting option as this gives you a team of behind-the-scenes experts thoroughly investigating which companies are future-proof.
• Take advantage of market dips A market correction of AI shares can also be a buying opportunity. Buy stocks gradually through regular investing, and always keep some cash tucked away to take advantage of buying opportunities. If you want to venture into investing in the shorter term, be aware of the greater risks.
• Look beyond the US DeepSeek’s success shows that not all AI innovation comes from Silicon Valley. Consider exposure to Asian and European tech companies as well.

Interested in diversified investments in AI?

• Log in to Keytradebank.be on your laptop or desktop
• Click on Advanced at the top of the search window
• Tick Tracker and/or Fund
• Search for the term “AI”, “Artificial Intelligence” or related terms