Comment by ArtTimeInvestor
18 hours ago
Can Europe build AI datacenters though?
Europe has no wafer production and no companies that produce GPUs.
That means it is dependent on Taiwan for wafers and the USA for GPU design.
Then there is the question wether there is a will to invest. Gemini gives me this list of publicly traded companies in the US and what they invested in AI infrastructure in 2025:
Amazon: $100B
Alphabet: $90B
Microsoft: $80B
Meta: $70B
Tesla: $20B
For Europe, I get this list:
Deutsche Telekom: $1B
How much of the stuff that is under control of the US cloud companies has any need for being in an ‘AI’ datacentre?
Does a store of healthcare records need AI? The state portal for renewing passports? The tax administration?
I seemed to be able to use all of these things online before the latest boom in AI came along.
AI will be involved in all processing of data. Everything that is human work today will be done by AI tomorrow.
So if Europe will rely on US AI infrastructure, nothing is won by moving the old CPU bound processes off of US cloud infrastructure.
Two out of three of these things are effectively fully automated already in my experience.
And I'm afraid I do not share your belief that the work done by a healthcare professional today will be done by AI tomorrow.
Isn't every AI datacenter chip manufacturer critically dependent on EU (ASML)?
Sure but the US isn’t vowing to eliminate all dependencies on EU goods. (Just burning all their good will.)
Do Europe need AI datacenters to survive? AI is immature technology that is not yet critical to anything.
Define 'survive'.
Given Europe's productivity gap with the US, they appear they are becoming even further a vassal. They will survive, but they further lose their leverage with each year. We see this in international politics as the US pivots away from Europe and towards Asia. (Although Russia's decline has also made it less necessary too)
If you want Europe to rejoin 'great powers', as 'survival', yes they need AI.
Productivity in producing slop? Because that's the only thing LLMs are good for so far. And by the time this change the GPUs installed today will become obsolete.
Is there a use case for AI that open models can't solve ?
Are there really any customers who are demanding AI and threatening to leave if those AI features are missing in every tech adjacent product ?
I think the make or break situation of integrating cutting edge AI for any business is just the hype and fomo at leadership level.
> That means it is dependent on Taiwan for wafers and the USA for GPU design.
Both being dependant on ASML, that we're crippling to please our bully, the USA. We probably have more leverage that we want to admit, but that requires a lot of politic will and... planning the economy.
People tend to fixate about cutting edge technology, but my naïve intuition says the problem in Europe is not in lack of some secret sauce: it is hidden in plain sight lack of energy to run the DC - and worse - lack of long term desire to make the tough choices to get that energy
I read years ago that Hetzner placed data centers near inexpensive power, but I understand that the EU’s energy situation has deteriorated. So you are correct, they have the larger energy problem to contend with.
I would also add lack of technical competence at C level. In my previous job, I have dealt with quite a few European CEOs whos only background was an MBA. Unlike the US where a lot of CEOs have a deep technical background...
Of course. One only has to take a look at Microsoft, Apple, or Google, to notice that they're run by CEOs with a "deep" technical background. No MBA whatsoever...
1 reply →
Well it is like Thiel said in a recent interview - European companies and investors are very risk-averse and will never be a vanguard like the ones in the US.
You'll never get here that kind of cash for any risky project, it usually is low risk + low margin.
Feels like people write that like it somehow is failure on investors side.
If you are investor on US market having 300M people speaking roughly the same language and then high possibility to easily spill over the world upsides on the bet are really high, burning cash to have a chance hitting jackpot are much much higher than in EU.
In EU you are starting in a single country so like 60M people and your payoff is capped from start at most likely scenario you go big in a single country and then you basically have clean start in next country.
That is the reality of game theory, not some failure of imagination or being scared to take risks - payoff is just not there, in US you have a shot at insane payoff in relatively short term.
> If you are investor on US market having 300M people speaking roughly the same language and then high possibility to easily spill over the world upsides on the bet are really high
The topic is cloud providers. Do you think it would be critical for a EU-based cloud provider to translate their admin GUI to Elfdalian, Basque and Romansh in order to succeed? Or perhaps there are some deeper underlying causes for European failure in modern computer tech that you can think of?
Maybe we don't need that kind of money for things like office software, email, reasonable sized databases, VPS etc.
Thiel recently called Greta Thunberg the anti - christ. Thiel is maybe crazy as Musk. at least he is not an authorative source.
Besides the soure what does he mean with all of Europe: Berlin? London? Paris? Estonia? Sweden? The start up eco system is fragmented / decentralised. I doubt Thiel is a good overview and he argues probably not in good faith anyway.
Zero to 1 was good.
But I generally have a hard time thinking any person's opinion is some sort of fact.
However, it kind of tracks with my understanding.
Yes, this is a core issue. Most datacenter investments in Europe come from American big tech. If AI is going to be half as huge as predicted (kind of a big if), then Europe would depend even more on the US for compute. The highest energy prices in the world coupled with conservative investment mentality means Europe is practically out of the AI race.
If they wait a year or so, the new AI chips being used now in China will probably be available for LLM inference in Europe. It seems unfortunate for small and medium size countries, and also for the EU to be dependent on any IT infrastructure only from China or the USA, but perhaps being flexible enough to be able to switch venders or use both is safer?
exactly. in HPC we all understood that it was a tradeoff between money and time, and that the curve was exponential. if you wanted to race ahead of todays capabilities, you could, but you couldn't go very far without burning alot of cash.
because of the investment story about being first and building a moat, we have companies torching 100s of billions of dollars to see who can climb that exponential the furthest.
we have so much work to do, in infrastructure, and distributed computation models, and programmability, quantization, and information theory...just relax a little. you dont have to compete with OpenAI. OpenAI is just a giant waste of money. take your incremental gains and invest in research and I assure you we can get there without directing our entire economic output into buying the latest highest margin parts from Nvidia only to use them at 30%, if you're being generous.
This kind of reminds me of an old paper: https://arxiv.org/abs/astro-ph/9912202
Of course, we are no longer in the clock speed doubling era, but computer does get faster still.
1 reply →
Wow you're so right, you did such a good job asking computer mommy to confirm your priors!
But actually, that's not the goal here. AI, at least the kind of products that need dedicated datacenters ie. generative, isn't critical infrastructure. The focus is on documents, collaboration tools, file servers, single-sign on, databases etc. that are seemingly monopolized by US providers.
> documents, collaboration tools, file servers, single-sign on, databases
All being (or soon to be) fed through LLM agents running on fibers and datacenters controlled by NOT European entieties. And if you build DC you'll be powering them with energy imports.
Software being built on library repositories also under foreign jurisdictions. Network infrastructure built on imported tech running whatever backdoors "partners" see fit.
Its like you didn't notice the snowden revelations, the shift from dependence on Russian Gas to US gas, nordstream sabotage, stuxnet, etc
Also to be honest, suppose EU uses kimi model which is open source. They can literally swap out one word from the provider and move from say American datacenter companies to European.
Quite frankly, there is literally 0 moat and its great to see EU focus on the real moat/lock-in issues.
ASML & IMEC are European.
I mean, if after three years all we got is Mistral, it's obvious that EU is out of the current round of AI race.
It might even be a positive thing. If the AI 'bubble' bursts they might end up saving tons of money and can buy idle GPUs at a discount.
Yeah and we could run kimi. Th issue is energy, we should keep building nuclear reactor and renewable to power it.
Aren't Mali GPUs designed in Europe?
AI is as far away from useful and necessary as bitcoin and NFTs were. I'm sure society can survive without it
Strange to say it’s not useful, I personally get a lot of use out of it. It’s mostly replaced search engines for me, which is no small thing.
Agree that society can survive without it though, but seems a weird thing to just claim as useless.