Europe is projected to hold about 5% of global compute, even though weighing 22% of the world economy. It will be the final straw if we don't act, and we cannot yet again rely on the government. We must give our heroes the weapons to play the AI race, while de-risking our already exposed industrial giants.

There is most likely no AI bubble. $500-600bn annual CAPEX are being spent in the world to build and operate AI, which means that $500-600bn is the kind of revenues that we are looking to find in the next few years. Per seat basis, it means that about $40 a month per knowledge worker would be enough to offset those CAPEX. Firstly, it seems highly reasonable considering the already obvious reliance on those tools in most of our offices, secondly, it doesn't consider most of where the real value will be created. AI is heading towards replacing a large part of the knowledge workers and will rapidly also replace most of physical labor work. Evolution has done an amazing job at creating beautiful species, but the simple fact we exist and can do what we do should convince us that there is no wall below human's level intelligence and physical ability. We will get there, it's only a question of time, and it doesn't seem like we are that far nor that we are going slowly right now.

What does the world look like in 10, 20, 1000 years?

Nobody knows, nobody can predict accurately what will the world look like when labor as we know it is mostly occupied by AI. There is a small thing that we can however predict: Compute won't be distributed equally in the world. Some regions, countries, some continents will have an almost infinite access to labor, having invested enough early enough, and others will be stuck in the past. Compute will convert in intellectual and then physical labor, and therefore ability to accumulate more compute and goods. A new currency that builds itself, a bit like accelerated capitalism. In such a world, in hindsight, would you have rather invested more, or less when money still had value?

Where is Europe headed?

Europe currently represents ~22% of the global economy. We hold 5% of worldwide AI compute, and the numbers paint a grim picture. While global data center CAPEX in 2025 is expected to hit $500-600 billion with the big four hyperscalers accounting for the majority, Europe's deployed facility CAPEX caps at roughly $12-18 billion. CBRE expects 937 megawatts to complete in 2025 across Europe, with build costs running $9-13 million per megawatt, yielding only $8-12 billion delivered in-year. Add ongoing draws on the ~2.6 gigawatt under-construction pipeline, and you're still in the low to mid teens. We are in big trouble. The committed capital is not really committed, and it's not like our governments' main focus is currently AI.

Can we fix it?

We can try. Contrary to the United States, which has a hyper powerful and rich technological sector, and to China, who has a hyper powerful, centralized and rich state, who believes more in AI than even I do, we are going to convince people for whom it's not the main business yet. What is reassuring is that we have our destiny between our hands: we have the cashflow output, and the EU governments, even though incapable of funding more than a percent of what we need, don't seem to be trying to get in the way, and are even passing bills to help us. We therefore have one thing to do: pour enough money in a money generating, productive asset, even if it's not related to our core business. Money is the main problem, but obviously at scale so many issues will get in the way of this investment, some predictable, others yet to emerge. The path forward requires coordinated private sector investment at unprecedented scale, strategic allocation of resources toward compute infrastructure, and decisive action before the window of opportunity closes.

What exactly should we do?

I don't know. All I know is that we have a problem, that the order of magnitude of the effort is doable, but the only solution seems to be forcing an absolute existential FOMO onto those who hold Europe's capital: pension funds, dividend-generating companies, even those unrelated to technology, and large institutional investors. We need to introduce them to special investment vehicles, maybe similar to the recent xAI deal with Nvidia, that make compute infrastructure accessible and compelling as an asset class. Contact me if you're willing to work on that.