The thing is that AI is not winner takes all market. AI models and server space is all fungible. It is very valuable, but no one will hold exclusive AI capabilities in the long run.
In the HPC space, it has always been a "winner takes all" market. 90% or more of the business goes to the biggest provider with few exceptions (in terms of dollars, obviously all compute gets used).
And, yeah, quite a bit of that is government having ... suboptimal ... developers who want to run bad code that nobody's allowed to look at. And yes, it is about weapons design, which is why nobody is allowed to look at (and certainly not fix) the code.
I think the idea is that AI will dominate a bunch of industries, from government administration, to intelligence (arguably they have that one already), to software, to insurance, to entertainment (produce 10 new lord of the rings series per day). And the first one to get there will "own" Government administrations, Netflix, Microsoft, Google, OpenAI, ...
If the past 20 years have taught us anything, it's that the first one to have a great popular success wins. Not in one market, on ALL of them.
Very few governments outside of the US will let a US company "own" them. It is a national security issue. They just take a good enough locally made model.
> They just take a good enough locally made model.
You're right that the question if that is an expensive mistake or not ... is for now an open question. However, even if Europe starts using it's own models, they'll be running on US chips.
And, yeah, quite a bit of that is government having ... suboptimal ... developers who want to run bad code that nobody's allowed to look at. And yes, it is about weapons design, which is why nobody is allowed to look at (and certainly not fix) the code.
I think the idea is that AI will dominate a bunch of industries, from government administration, to intelligence (arguably they have that one already), to software, to insurance, to entertainment (produce 10 new lord of the rings series per day). And the first one to get there will "own" Government administrations, Netflix, Microsoft, Google, OpenAI, ...
If the past 20 years have taught us anything, it's that the first one to have a great popular success wins. Not in one market, on ALL of them.