Europe is falling behind in the AI race

November 30th will be the fourth anniversary of the public appearance of ChatGPT. It was when the majority of the population realized that artificial intelligence was for real and that it had such great power that it surprised even the specialists. AI has shown that it was not just another technological bubble, but that it has passed with flying colors from promise to reality. At the end of 2022, only OpenAI, the emerging company behind ChatGPT, was visible. But soon it became known that other US companies, some giants like Meta, Google and Amazon, and other emerging ones like Anthropic, had large language models ready similar to the one that powered ChatGPT. Without making much noise, China established AI as a national priority and, with its usual effectiveness, has managed to get its companies to obtain models that rival those of the US: although they are still slightly less powerful, they are more efficient in calculation cost and energy expenditure. Chinese models like DeepSeek and Qwen are increasingly popular in the West.

And Europe? Although a small handful of companies tried, it soon became clear that only Mistral AI, a startup created in France in 2023 with basically European capital, had the technology and resources to compete with the large models from the USA and China. Currently, this company remains the only European champion in large general-purpose AI models.

However, Mistral AI's investment in AI infrastructure is not comparable to that of the large US AI companies (Meta, Alphabet-Google, Microsoft, Amazon). In 2026 alone, each of these companies will invest between 130,000 and 250,000 million dollars in data centers, while Mistral AI's investment will be around 4,000 million over several years. If instead of money we compare the electrical power of the facilities, Mistral AI's goal for 2030 is to have an infrastructure of 1 gigawatt, while OpenAI already has more than 10 gigawatts at this time. Therefore, although Mistral AI allocates its resources very efficiently, the reality is that its investment is 2% or 3% of that of Google, Microsoft, or Amazon. The comparison with the investment of Chinese companies is more difficult, because costs in China are lower: in absolute terms, they are situated between the US companies and Mistral AI. Alibaba, ByteDance, Tencent, Huawei, and DeepSeek invest between 10,000 and 20,000 million dollars annually each.

On the other hand, does it make sense for European Union administrations to try to make up for the lack of companies and try to lead the AI revolution in Europe? If both in the home of global capitalism (the United States) and the champion of the planned economy (China) it is the companies that lead, it seems strange that in Europe we delegate this to the administrations. To give an example, let's imagine that the European Commission ends up choosing Móra d’Ebre as one of the 7 gigafactories it will finance, out of the 77 candidacies submitted (!). Well, the electrical power of this gigafactory will be 150 megawatts, a very small fraction of the power installed by large AI companies. Furthermore, due to the slowness of the entire selection process, in the best-case scenario it would not come into operation until the end of 2028, and a period of more than two years is an eternity in the current runaway race for AI infrastructure. This is not to mention that the GPUs (the chips that make up large data centers) that will be purchased will not be European technology, but rather from the American company Nvidia.

Speaking of our lack of sovereignty in chips, it is worth remembering that in the 1980s the European Commission and national governments (Spain included) tried to lead the microelectronics revolution by financing public R&D centers in this sector. The result of these policies is clear: 60% of the world's chip production is in Taiwan, the world's main chip designer is Nvidia (USA), and in Europe we are only left with the Dutch company ASML, which is the world's main supplier of photolithography equipment for chip manufacturers. We should avoid repeating in AI the mistakes that were made in microelectronics.

What could Europe do to avoid being definitively left behind in the AI race? To begin with, follow the advice of Mario Draghi and Christine Lagarde, who recommend a unified capital market in the European Union. As the president of the European Central Bank has said, European households have 440 billion euros invested in US technology companies, and they substantially finance the AI boom in that country. If the European capital market were competitive, many of these savings would go to European companies and perhaps the gigafactory in Móra would be promoted by a company and not by the administrations. Besides growing Mistral AI, European GPU manufacturers could be promoted, much like how China promotes its manufacturers. Meanwhile, European administrations could invest in these strategic companies by buying their shares instead of granting subsidies of dubious utility.