The DeepSeek panic was short-lived. New AI models open up new opportunities for investors and for Norwegian entrepreneurs, writes Salvador Baille, partner at Aggrator, in Shifter
On Monday morning, the world woke up to what appeared to be an AI Ragnarok. Panic spread among investors, who dumped everything that smelled of artificial intelligence as if it were toxic. DeepSeek had taken the stage and promised achievements on the same level as OpenAI, Metas Llama, Google's Gemini and all the other American generative AI models.
At the same time, it was quite remarkable how calmly the top executives of these companies reacted to the chaos. Even The Donald behaved surprisingly civilly, without theatrical threats or announcements of new dramatic measures.
They were right.
At the moment, the stock values of Microsoft and Google are roughly where they were before the crisis. Meta is almost at an all-time high. Nvidia (which was hit with the hardest penalty) is only six percent cheaper than the company was a month ago.
All of this for good reason.
Firstly, many of the techniques used by the language and reasoning model DeepSeek are already well-known. The major breakthrough, «Pure Reinforcement Learning,» which should allow the model to train itself and make it much cheaper to build, can also be replicated relatively quickly. OpenAI has already begun experimenting with such models. Meta launched Llama 3.3 in December, a language model that also drastically reduces the need for computing power compared to previous models.
Just as it was logical to expect, AI models are becoming increasingly simpler, cheaper, and more accessible.
Secondly, language is only one of the many aspects of generative AI. The next milestone is realistic image generation, video, and emotion interpretation. The demand for computing power remains unstoppable. The AI chips from Nvidia, which are supposed to be the cheapest in terms of computing power versus price, are therefore still needed. New advanced applications based on artificial intelligence will emerge, and they will also require increasingly efficient use of computing power. Nvidia CEO Jensen Huang can sleep well at night.
Thirdly, DeepSeek has weaknesses as well. The security aspect regarding potential surveillance by Chinese authorities is naturally problematic. In terms of hardware, DeepSeek and other Chinese competitors also face major challenges at the moment; the stock of Nvidia chips is used up and cannot be replaced due to the export ban on chips from the US to China. Moreover, DeepSeek is still a research project. The extent to which their Open Source model or organization will be able to scale is an open question.
This development has consequences nonetheless, and it will force the AI industry to split into three main segments.
On the one hand, we will find companies that Anthropic, Companies such as OpenAI/Microsoft, Google, or Amazon will develop AI models with multiple applications for increasingly large markets. These will be forced to develop business models and user-friendly interfaces in fierce competition with each other, often delivered via their cloud services, which customers are already dependent on.
Only a handful of these companies will survive, but they will deliver solid results to their shareholders.
Some other players will develop AI as a tool to support their own core business. These will subsidize the development with the company’s cash flow and will make the models Open Source and cheap to use to accelerate the development. Here we will find players like Meta or Tesla.
The disadvantage of this strategy is that the models will be more difficult to deploy for ordinary mortals. Their applications will become more limited and related to the company's needs.
Nevertheless, better and cheaper AI means better products and services for their customers, and hopefully higher profits and more satisfied shareholders.
Finally, we will see more Open Source AI models like DeepSeek emerge; many of them have roots in universities and research communities. They will die quickly, become irrelevant in the long run, be acquired, or attract investors if they specialize in very specific problems.
Based on these scenarios, some good opportunities may open up for Norwegian entrepreneurs.
Start-ups with expertise in generative AI can develop user-friendly interfaces and support services for Open Source models. The company Red Hat did just that with the operating system Linux and sold it to IBM for 284 billion kroner in 2018.
Those who want to include artificial intelligence in their products can also rejoice. For the vast majority, the most common AI applications will become increasingly cheaper, better, and more accessible.
A third possibility is to develop AI models with very concrete applications for high-profitable niches. In these cases, hands-on experience and excellent contacts within the selected niches will be critical. In such cases, investors will have a laser focus on the team.
The major challenge will be to find the right talent and expertise in a country that is already suffering from an acute shortage of IT specialists. If the situation does not change soon, and politicians do not take these opportunities seriously, our society risks seeing them come to light in another country – or even on another continent.
The post was first published in Shifter.