The turning point of generative artificial intelligence is a new opportunity for Europe 

Alberto Iglesias Fraga, Journalist specialized in innovation and digital economy

Artificial intelligence has reached a turning point. After two years of being constantly on the public agenda, many analysts are beginning to see the beginning of the end of the bubble of expectations surrounding this technology. But more importantly, experts from all walks of life are beginning to warn of the immediate limitations to the growth and scalability of the current large language models (LLMs) that underpin some of the best-known generative AI tools, such as ChatGPT, LlaMA, or Gemini.  

There is no doubt that large language models have captured the global imagination and laid the foundations for a new era in the race for artificial intelligence. However, technical limitations and the enormous costs of training and operating them make their medium-term future unviable. Hence, we are seeing a certain shift towards smaller, more specialised, and efficient models. A paradigm shift that not only opens new opportunities in the technology sector but also places Europe and Spain in a privileged position to develop alternatives to the US giants. 

The AI models we are all familiar with, despite their impressive capabilities, require massive infrastructures to function optimally. The high power consumption, the need for specialised GPUs, and the constant maintenance of these systems pose major challenges. Accenture estimates that the cost of processing these models will double in the next five years, making them difficult to adopt by smaller companies or in budget-constrained industries. 

In addition to their high cost, LLMs are limited in many other respects. Firstly, their “universal training” prevents them from being adapted to the specific needs of a company or organisation, which must retrain the model to meet its expectations. This, of course, comes at an extra cost and has an associated environmental impact. Otherwise, these models can answer a wide range of questions but lack the precision and customisation that many organisations require to address complex or highly technical problems.  

What is the alternative then? What path can we follow to mitigate these brakes on the AI adventure? As we mentioned earlier, there is a clear demand in the market for lighter AI solutions tailored to specific use cases. Smaller models, also known as edge models, have the potential to meet these demands, allowing companies to deploy AI on less powerful devices, reduce operational costs and, at the same time, gain in adaptability and accuracy. 

In other words, instead of having to adapt a generalist LLM to a specific context, companies can use models designed for specific sectors, such as finance, healthcare, logistics, or industry. This customisation capability results in a significant improvement in accuracy and efficiency, which is essential in critical applications such as medical diagnosis or real-time fraud detection. And results similar to general questions can always be obtained by linking different models together as a network.  

On a technical level, these specialised models can also run on less powerful devices, such as smartphones or tablets, and do not require expensive server infrastructure to run basic operations. According to a Gartner report, the AI market for low-power applications will grow by 40% by 2026. That is a lot. 

A geopolitical opportunity  

The promise of the century, one might think. And even more so if we land it in a geopolitical key, where these new models may find fertile ground in Europe, where regulatory standards, public funds, and growing market demand are driving the development of solutions of this type as opposed to large LLMs. Not surprisingly, the European AI ecosystem has been distinguished by a focus on privacy, sustainability, and ethics, in clear opposition to the more unregulated environments of classic powers such as the United States or China.  

Thus, the EU has allocated significant resources to promote safe, efficient, and privacy-friendly AI. Programmes such as AI4EU or the Digital Europe Programme support research and development of decentralised and accessible AI for small and medium-sized enterprises. In this context, the EU funds of up to 7. 500 million euros for AI and cybersecurity projects seek to promote precisely this type of new model. Spain is also moving in this direction, especially through the National Strategy on Artificial Intelligence. 

And this is true on the supply side, but also on the demand side: it is estimated that European companies will spend more than 10.000 million euros on artificial intelligence by 2030, with smaller models becoming a viable and attractive solution for our productive fabric.   

All in all, the transition towards smaller and more specialised models may represent a before and after in the structure of the AI industry. Currently, this market is dominated by US and Chinese giants, such as OpenAI, Google, and Baidu, which operate with massive language models and their own infrastructures.  

But in the face of the walls they will face, European companies such as Aleph Alpha in Germany or Mistral in France can seize the opportunity and develop lighter models, applicable to specific industrial and commercial uses. These startups can challenge the hegemony of the big foreign models. And that is probably the closest we have ever come to putting the Old Continent at the forefront of this revolution.   

Sustainability as a claim  

Sustainability is a priority in Europe, especially in the field of artificial intelligence. European regulations, led by the Green Deal and the AI Act, require companies to develop technologies that have a low energy impact and meet rigorous ethical and privacy standards. According to a Forrester study, the demand for sustainable and regulated AI in the EU will generate a market worth 35.000 million euros by 2027.  

Again, the answer to this demand involves smaller models, as they require less energy to operate. A green alternative, which, if it emerges in Europe, would allow our companies to focus on a niche market that values adaptability, resource savings, and efficiency rather than competing directly with the technological giants.   

The turning point is here. It is now time to seize the opportunity that lies before us.