Great piece here Carey! One thought I had when it comes to the first part of your two-track approach regarding building - what do you think Europe should be building specifically? Should it be trying to rebuild a European version of the US-dominated tech stack it relies on, or should it be building something else? In the context of AI specifically, I am not sure if it is worth it to try and compete on frontier models as I think Europe might be too far behind to do it well. What might be more worth it is building the infrastructure that allows AI to be implemented across industries more effectively and responsibly. This is not necessarily just physical, but also building the know-how for integrating models into wider systems and helping their diffusion across the economy. So it is more about education and training, expanding both the number and quality of people who can build and implement the technology across sectors, and standardising best practices so knowledge flows smoothly between the builders and implementers. Is this something worth focusing on you think?
This was actually a condensed version of a larger draft I wrote, which included more build-stage material. Here’s what I wrote: Treat the EU Cloud and AI Development Act as the spine of an industrial strategy, not a procurement footnote. It already directs public money toward European products, cuts data centre red tape, and lets the Commission vet providers for sensitive sectors.[24]
Use the Chips Act's crisis toolbox and capital, and lower procedural barriers. The first Chips Act already allows the Commission to compel priority production in a crisis—an EU analogue to the US Defense Production Act. Member-state co-investment (e.g. Infineon's €5bn Dresden fab) shows demand exists. The binding constraint is not ambition but friction: permitting, fragmented grids, procurement complexity. Reduce it deliberately and strategically.
Preference compliant EU firms through demand, not prohibition. Ursula von der Leyen's 2024–2029 political guidelines promised to revise public-procurement legislation to "enable preference to be given to European products in public procurement for certain strategic sectors." France's SecNumCloud certification, which grants structural immunity from extraterritorial US access and legal mechanisms, shows how to make market access conditional on staying within the bounds of EU law. The logic is light-touch: make it easier for lawful EU entities to compete, rather than harder for everyone across the board.
Pursue a "brain gain." US immigration hostility and constrained H-1B access, plus China's human rights deficit, create an opening for talent. Europe cannot out-spend either rival, but it can compete on autonomy, rule of law, quality of life factors, and research freedom—if it funds the institutions to host returning talent.
Fund it like the strategic priority it is. Both tracks need capital at a scale Europe has so far avoided—the Draghi competitiveness report put the gap at roughly €750–800bn/year in additional investment. But not all of it is new money: redirecting even a fraction of the ~€264bn Europe spends each year on US tech toward compliant European providers is itself an industrial-policy instrument.
I am probably missing a lot. First thing though, is I think for a quick win, Europe should really take advantage of how hard the Trump administration is fucking over researchers in the US and scoop up all the smartest minds. Treat it like America did in the 1940s-1960s — openly encourage the best and brightest to move for better opportunities and unfettered research access. Incentivize them to do so.
Great piece here Carey! One thought I had when it comes to the first part of your two-track approach regarding building - what do you think Europe should be building specifically? Should it be trying to rebuild a European version of the US-dominated tech stack it relies on, or should it be building something else? In the context of AI specifically, I am not sure if it is worth it to try and compete on frontier models as I think Europe might be too far behind to do it well. What might be more worth it is building the infrastructure that allows AI to be implemented across industries more effectively and responsibly. This is not necessarily just physical, but also building the know-how for integrating models into wider systems and helping their diffusion across the economy. So it is more about education and training, expanding both the number and quality of people who can build and implement the technology across sectors, and standardising best practices so knowledge flows smoothly between the builders and implementers. Is this something worth focusing on you think?
This was actually a condensed version of a larger draft I wrote, which included more build-stage material. Here’s what I wrote: Treat the EU Cloud and AI Development Act as the spine of an industrial strategy, not a procurement footnote. It already directs public money toward European products, cuts data centre red tape, and lets the Commission vet providers for sensitive sectors.[24]
Use the Chips Act's crisis toolbox and capital, and lower procedural barriers. The first Chips Act already allows the Commission to compel priority production in a crisis—an EU analogue to the US Defense Production Act. Member-state co-investment (e.g. Infineon's €5bn Dresden fab) shows demand exists. The binding constraint is not ambition but friction: permitting, fragmented grids, procurement complexity. Reduce it deliberately and strategically.
Preference compliant EU firms through demand, not prohibition. Ursula von der Leyen's 2024–2029 political guidelines promised to revise public-procurement legislation to "enable preference to be given to European products in public procurement for certain strategic sectors." France's SecNumCloud certification, which grants structural immunity from extraterritorial US access and legal mechanisms, shows how to make market access conditional on staying within the bounds of EU law. The logic is light-touch: make it easier for lawful EU entities to compete, rather than harder for everyone across the board.
Pursue a "brain gain." US immigration hostility and constrained H-1B access, plus China's human rights deficit, create an opening for talent. Europe cannot out-spend either rival, but it can compete on autonomy, rule of law, quality of life factors, and research freedom—if it funds the institutions to host returning talent.
Fund it like the strategic priority it is. Both tracks need capital at a scale Europe has so far avoided—the Draghi competitiveness report put the gap at roughly €750–800bn/year in additional investment. But not all of it is new money: redirecting even a fraction of the ~€264bn Europe spends each year on US tech toward compliant European providers is itself an industrial-policy instrument.
I am probably missing a lot. First thing though, is I think for a quick win, Europe should really take advantage of how hard the Trump administration is fucking over researchers in the US and scoop up all the smartest minds. Treat it like America did in the 1940s-1960s — openly encourage the best and brightest to move for better opportunities and unfettered research access. Incentivize them to do so.