European AI – A Drop in the Ocean

Europe remains a rounding error.

• The fact that Mistral is rolling out 13,800 GB300 chips in Europe is newsworthy, is a sign that AI in Europe remains far behind both the USA, China and the Middle East, mostly as a result of an uncertain and generally hostile regulatory environment.
• Mistral, the French full-stack AI services provider, has raised $830m in debt financing that will go towards its plan to build 200MW of data centre capacity by the end of 2027.
• This amounts to roughly $10bn of capital expenditure, of which roughly half has been allocated to building facilities in France and Sweden.
• Mistral is pretty much Europe’s only home-grown AI platform contender and its best hope to have a relevant position as the AI industry emerges and matures.
• Almost all of the other capacity being rolled out in Europe belongs to the US hyperscalers or non-European neo-clouds (except Nebius).
• By contrast, the USA will roll out 13GW of capacity in 2026 (65x Mistral’s 2027 target), and the Middle East will bring 400MW in line in 2026, which is double Mistral’s 2027 target.
• Mistral also expects to have $1.1bn in annual, recurring revenues by the end of 2026, which is a rounding error when compared to either OpenAI or Anthropic.
• This is a pretty poor state of affairs as the USA’s GDP is only 50% larger than Europe’s while its AI position remains a rounding error compared to the USA.
• Hence, given the relative size of Europe’s economy, it should be doing far better than it is.
• The problem, as I see it, is both Europe’s conservative culture and its regulatory environment, which is still pretty hostile to AI.
• Despite a big push from Nvidia, Mistral and most of the political class at the VivaTech Trade show in June 2025, most of the talk has not been converted into any visible action.
• A great example of the malaise is that in June 2025, I spoke to a number of EU policymakers and government officials from EU member countries who all admitted that the AI Act had gone too far and risked harming the rollout of AI in Europe.
• Many participants are also concerned that if they build the AI infrastructure, no one will show up despite the clear evidence from the US and elsewhere that demand for AI compute continues to outstrip supply by a very wide margin.
• In the last 10 months, Europe’s response has been to delay the implementation of the rules that have been causing all the trouble, as opposed to ripping the whole thing up and starting again.
• The problem with the EU AI Act is that it seeks to regulate an industry that does not yet exist, and because no one knows how it will work, it is impossible to set appropriate rules for it.
• The result has been a series of well-intentioned rules that place obligations upon the creators of the algorithms that they have no hope of ever being able to meet.
• The result is that anyone with a great idea for AI goes to America to turn that idea into reality.
• The best example of this is OpenClaw, which was invented by Austrian developer Peter Steinberger in Europe, but the minute it became popular, he relocated to the USA and joined OpenAI.
• This should have been an innovation that was fostered and grown in Europe, and given its popularity with open source, the hyperscalers and the technology suppliers, OpenClaw could have really put Europe on the AI map.
• Instead, the fruits of OpenClaw will now be earned by the American and Chinese economies, resulting in Europe falling further and further behind.
• This is why I think Europe needs to move very quickly to clarify the regulatory environment that will be in force so that businesses and entrepreneurs can decide whether they want to build an AI business in Europe or not.
• The more it fiddles and prevaricates, the more start-ups, businesses and entrepreneurs will decide to set up elsewhere, and many other places are more than ready to welcome them.
• The most immediate beneficiary will be the Middle East (once the war is over), which sits on Europe’s doorstep and has plenty of space, money and energy to build the infrastructure for small private companies to use.
• I suspect that we will also continue to see an influx of talent into the USA, which has taken the opposite approach to AI and is already by far the global leader for AI outside of China.
• The time for the EU to rethink the rules is running out, and the signs are that it won’t do anything until it is much too late.