For months, German media such as Handelsblatt, manager-magazin, and other media outlets in Europe have been painting a picture of impending disaster: the AI bubble will soon burst, leading to an AI crash. These warnings are primarily being voiced by European media, not so much in the US. That’s because the US sees potential in artificial intelligence that European experts and media outlets are unwilling—or unable—to see.

A few weeks ago, I wrote in my post AI Bubble? A Fairy Tale for Scaredy-Cats explained why these fears are unfounded. And also why this technology cannot be compared to the dot-com bubble.

Bubbles: AI versus the Internet

First, during the dot-com bubble around 2000, the companies that went bankrupt were often startups that had driven up valuations and expectations with venture capital financing. When no more fresh venture capital came in, everything collapsed. With artificial intelligence, it’s different. The companies involved are all highly profitable. The amounts that tech giants such as Microsoft, Google, Apple, and Nvidia are investing in the expansion of AI infrastructure and the training of new AI models are only a fraction of the profits these companies rake in every quarter. Even if the investments have to be written off, these companies are still highly profitable and liquid.

Secondly, skeptics and AI bubble warning voices consider AI to be too limited. When they talk about AI, they mainly have chatbots—i.e., large language models (LLMs)—in mind. However, these LLMs are the smallest part of the entire AI world, and not even the most computationally intensive. Even AI agents, where language models are combined with other software tools to create independent agents that can perform complex tasks on their own, require more processor time and thus also data centers.

And yet the age of AI has not even really taken off yet, because after standalone AI, such as chatbots, and autonomous AI or AI agents, physical or embodied AI is now following.

Artificial intelligence is embedded in cars, robots, machines, and drones, turning them into autonomous actors in our physical world. Autonomous cars such as those from Waymo and Zoox are just beginning to populate our roads. And they not only need local AI chips in the device itself, but also access to powerful data centers that offer the necessary computing power.

Humanoid robots, whose numbers are estimated to reach one billion or more by 2040, will perform so many complex tasks in households, factories, offices, and open spaces that they will not be able to function without data centers and AI tools. The computing power that chatbots require today, which seems enormous to us, will pale in comparison to the demands of physical AI.

This means that the estimated $600 to $700 billion that tech giants will have invested in the expansion and development of AI and AI infrastructure in 2025 alone is really just the beginning. The $7 trillion (i.e., $7,000 billion) figure cited by OpenAI CEO Sam Altman is already proving to be a fairly accurate and verifiable estimate of the sums that will be required.

But why are European media and experts so wrong?

Good question. Is it the pervasive and prevailing skepticism that sees risks and dangers in everything new, but not the opportunities? Is it a certain arrogance toward everything that comes from the US or China? Or the inability to see opportunities, possibilities, and potential? Is it a lack of vision that makes it impossible for them to view new technologies from the perspective of what is known, rather than future possibilities?

They also have projections about future consumers with a huge appetite for AI computing power. Robot taxis, humanoid robots, and the like will eclipse everything chatbots need. AI investments are therefore not only well spent, they must also be stepped up again. But future potential is more difficult and less lucrative to sell to readers and the public than issuing warnings, appealing to fear instincts, and positioning oneself as the only one who has the insight and has always said so.

Side Note

Germany itself would have had to invest between $125 billion and $150 billion in 2025 to keep pace with the US in terms of population. Bitkom estimates that data center operators in Germany will invest around €12 billion in IT hardware and a further €3.5 billion in buildings and technical equipment in 2025, bringing the total to around €15.5 billion.

This discrepancy is striking and also answers the question of the extent to which Germany and Europe will drive these developments forward or be able to lead the way in this technology.