SAP CEO Christian Klein considers it unnecessary to implement a massive AI-driven expansion of data centers in Europe, as is being done in the US. According to him, a European equivalent of Stargate would yield less than a focus on the practical application of AI. Why?
Klein responded during a media interview to European plans for a large-scale AI infrastructure. These plans include a proposal to build five “AI gigafactories,” which many cities are competing to host. The SAP CEO doubts whether this is necessary.
Commoditization of LLMs
According to Klein, LLMs are rapidly becoming commonplace, despite the high system requirements for groundbreaking AI training rounds. He points to the Chinese company DeepSeek as an example, which surprised friends and foes alike earlier this year with a model that is quite competitive with OpenAI’s and Google’s best options.
Instead of investing billions in infrastructure, Klein suggests that European sectors such as automotive and chemicals should focus on practical AI applications for their business operations. He believes this approach would be much more effective than trying to keep up with US investment levels.
Incidentally, this is already happening, albeit in a fragmented form. Think of startups such as Juvoly for AI speech recognition in Dutch or Silo AI from Finland, which was acquired by AMD and has focused strongly on the development of language models in all Scandinavian languages. But other applications of AI can also be found at a medical innovator such as Cradle, which wants to make biology programmable.
In the automotive sector, which Klein explicitly mentions, there is less positive news to share on the European side. Perhaps he would prefer to see the current allocation of €20 billion for AI infrastructure within the EU go to initiatives such as improving batteries and breakthroughs in self-driving cars. However, this also seems to be a battle that Europe is already losing.
Large investment gap
In any case, the figures speak volumes about the gap between Europe and America when it comes to AI infrastructure. American tech companies have announced the “Stargate” plan with much fanfare, with potential investments of up to $500 billion. Europe, on the other hand, has only committed 20 billion euros, or twenty times less, for the five AI “gigafactories” at locations yet to be determined.
These differences became even clearer when Nvidia CEO Jensen Huang claimed during his European tour in June that the region’s AI development was being held back by insufficient computing power. He announced partnerships to strengthen Europe’s AI infrastructure with thousands of Nvidia chips. Needless to say, he has a vested interest in seeing all continents compete for the limited AI hardware. It drives up the value of every Nvidia GPU; the pickaxes in the AI gold rush.
Should we therefore completely reject Huang’s argument? It can’t hurt to at least ensure that the expected demand for AI applications can be met entirely within Europe, even if that means using cloud services from American hyperscalers and American chips.
Klein’s perspective fits in with the broader discussion on digital sovereignty that Europe is having. Earlier this year, he called plans for European data centers to compete with US cloud providers “nonsensical.” He emphasized at the time that Europe’s high energy costs make such investments unattractive. That too remains a limitation, even with more than enough ambition to realize the AI rollout.