I think there's some truth to the matter for the reasons.

As Google's Argon and Meta's muse wonder has shown: Even reaching today's SoTA is possible provided you have enough compute, data and the moat for building it is comparatively low and it would still take years to play catch up while burning money for comparatively low point as China open sources its SoTA models that are just few months behind.

Also suppose such a model is built, then what? what matters would be running something valuable on top of that model which can provide top value in it of itself.

Now this is something which is more possible and plausible and where the demand is.

What might matter more is the deployment of open weights models on EU datacenters with proper ZDR or optimzing more on the speed aspect of things.

There is also a factor that EU VC funding can simply not have so much money to infuse similar to how AI labs got injected by money for so long, and also even if they have, the question of "should they?" still arises.

I think that this is sort of a playbook similar to how Apple is operating. I think that it's a good enough and a realistic playbook at that.