After the DeepSeek shock: What opportu­nities does AI offer investors?

Considering that their business model was practically declared dead just a few months ago, the large American technology companies are proving to be surprisingly resilient. For instance, the sector's earnings in the first half of 2025 climbed by an average of 23% compared to the previous year. As a result, the U.S. tech giants far exceeded market expectations and outperformed most other industries.

Why did the DeepSeek announcement cause the largest single-day loss in history?

Last January, the situation looked very different. At that time, the industry suffered overnight book losses of nearly USD 1 trillion; the stock of the leading manufacturer of processors for artificial intelligence (AI), Nvidia, experienced the largest single-day drop in market capitalization in history, losing nearly USD 600 billion. The reason for the stock market panic: the Chinese AI startup DeepSeek announced that it had trained its AI model R1 at a fraction of the cost of comparable Western AI. DeepSeek's AI models proved to be many times more efficient than previous models, requiring significantly less computing capacity.

During the subsequent "DeepSeek shock," investors assumed that demand for AI infrastructure, such as expensive high-performance graphics processors (GPUs), as well as network technologies and energy equipment for data centers, would plummet. A sell-off of stocks from AI infrastructure providers was the direct result.

By now, it has become clear that the market reactions at the beginning of the year were highly exaggerated. Moreover, the opposite of the fears from back then has occurred. Not only has demand for AI infrastructure continued to rise, but it has also exceeded previous growth expectations. The investments of leading tech companies in AI data centers (so-called hyperscalers) have been revised upward. For example, the consensus estimate of industry analysts is that Google will increase its capital expenditures by 41% between the end of December 2024 and the end of July 2025, Amazon by 30%, and Oracle by 32%.

Hyperscaler Capex of Amazon, Microsoft, Google, Meta and Oracle since the beginning of 2025 (consensus estimates in USD billion)

In our opinion, there are several explanations for this: Even though the costs of AI models are decreasing, the trained models are becoming increasingly powerful and complex. In the race for AI leadership, investments in training even more potent models continue to be made.

Additionally, inference – the processing of queries by existing AI models – requires increasingly more computing power. This is necessary to deliver the countless responses that millions of users worldwide request from the models: AI "as a service" is already a global product, AI applications are spreading further, and they are becoming part of everyday life.

Furthermore, in our view, the so-called Jevons paradox likely explains the increasing demand for computing power. This paradox states that more efficient technology – and thus lower costs – leads to exponentially higher usage. This, in turn, results in greater demand for resources overall.

What role do government interests play in the demand for AI?

Last but not least, countries are also participating in the AI race. The US and the United Arab Emirates (UAE) are supporting the private-sector project “Stargate,” which aims to invest around USD 500 billion in this future technology. Last July, US President Donald Trump's administration also granted the industry numerous freedoms with its “AI Action Plan”. Countries such as Canada and the United Kingdom have directly promised billions in investment, while the EU plans to invest up to EUR 20 billion in the establishment of 20 AI centers in France, Germany, Italy, and Spain.

All of this suggests that the long-term trends supporting growth in future technology can be considered intact. The same applies, in our view, to the three phases of AI adoption: building the necessary infrastructure, developing software, and eventually deploying it in numerous applications. Accordingly, we find the timing interesting to invest in selected companies in the fields of AI technology and cloud infrastructure.

In addition to the risks of market fluctuations and technological changes, we identify the following investment opportunities:

• Demand for GPUs (graphics processors) continues to grow; with the transition to the phase of scaled AI applications, additional application-specific chips (application-specific integrated circuits, so-called ASICs) will be needed to handle tailored AI tasks.
• Cloud-based AI platforms and services are expanding rapidly. In addition to private use, more and more applications for businesses are being demanded.
• To handle the new requests generated by AI applications, network and connectivity intensity is also increasing.
• AI data centers require a lot of energy. Therefore, electrification infrastructure must also be expanded. Since AI also offers exciting potential for greater energy efficiency, this expansion does not necessarily have to come at the expense of the climate. This is particularly significant with regard to sustainable investments.

Will AI one day become as ubiquitous as the internet?

Numerous applications for AI are still in their infancy. In the coming years, increasingly domain-specific applications are likely to emerge, such as in healthcare, industry, or robotics. Such applications could contribute to the physical and virtual worlds merging more and more – and AI as a technology becoming as ubiquitous as the internet is today.

Why do we consider the investment potential of AI and cloud technology intact even after the DeepSeek shock?

1. The AI model presented at the beginning of the year by the Chinese startup DeepSeek proved to be many times cheaper in training and inference. This unsettled investors. However, demand for AI solutions and infrastructure has since increased.
2. Investments by major technology companies in AI have also continued to grow.
3. From an investment perspective, a wide range of products and services can offer return potential. This includes cloud technology and data centers, networks and energy efficiency, as well as various forward-looking applications.
   

Investment theme «Digital Economy»: Insights