Rise of AI Conference 2025 · Berlin · Applied AI Stage

The German Genius: Invention & Innovation

Tina Klüwer · Director of AI Strategy, German Federal Ministry (described in the talk as recently renamed to cover research, technology and space)

At Rise of AI 2025 in Berlin, Dr. Tina Klüwer, Director of AI Strategy at what she describes as Germany's federal ministry for research, technology and space (renamed from the education and research ministry the week before the talk), argues Germany suffers from a persistent gap between invention and innovation. She cites rankings placing Germany third worldwide in AI research publications, sixth in patents, ninth in AI startups and twelfth in enterprise AI adoption, calling this the German or European paradox: strong basic research, weaker commercialization. She describes a policy shift from funding research alone toward supporting businesses that adopt and build AI, including four regional AI service centers for SMEs and two EuroHPC AI factories built in Germany. She also previews a planned EU AI gigafactory initiative with up to five facilities across Europe, each with about 100,000 GPUs, for which Germany is seeking to be a host. A lively Q&A covers energy consumption, ministerial coordination, public trust and whether Germany is too late to the infrastructure race.

Key takeaways

  • Klüwer places Germany third in AI research publications but only ninth in AI startups and twelfth in enterprise AI adoption.
  • She attributes the gap to what she calls the German or European paradox: strong invention, weaker innovation and commercialization.
  • Germany has built two EuroHPC AI factories, in Stuttgart and Jülich, with half their capacity reserved for SMEs, enterprises and startups rather than research alone.
  • The planned EU AI gigafactory initiative envisions up to five facilities with roughly 100,000 GPUs each, and Germany is seeking to host one.
  • Klüwer cites the Stanford AI Index to say German public acceptance of AI rose by 10 percentage points year over year.

Detailed notes

Germany's track record: strong invention, weaker innovation

Klüwer opens with well-known German inventions, from aspirin and the printing press to MP3 and the coffee filter, and the origin of Barbie in a doll based on a Bild newspaper comic that Ruth Handler saw in Europe. She argues Germany still holds a reputation abroad as a country of 'poets and thinkers' as well as engineers, but that this creative reputation does not automatically translate into commercial success in AI today.

From publications to commercialization: where Germany ranks

Klüwer cites Germany as the world's third-largest economy and third in AI research publications, behind the US and China. It ranks sixth in patents, ninth in the number of AI startups, and twelfth in enterprise AI adoption rate. Among the most-used AI websites by monthly visits, she notes ChatGPT ranks first and, to her surprise, Canva second, while the first German tool, DeepL, ranks thirteenth. She calls the pattern of ranking well in research but progressively worse further down the innovation pipeline the German or European paradox.

Shifting focus toward business adoption

Using the technology readiness level framework, she places Germany strong in basic research (levels 1-3) and reasonably solid in development within large enterprises (levels 4-6), but weaker in deployment and commercialization (levels 7-9). She describes a policy shift toward supporting businesses that use, adopt and create AI, including joint industry-research transfer projects and four regional AI service centers that give SMEs access to compute, GPUs, inferencing and AI expertise, while also functioning as regional talent hubs near universities.

Building compute: AI factories and the planned gigafactory

Germany has built two EuroHPC AI factories: Hammerhai in Stuttgart, focused on engineering and industrial HPC, and one in Jülich offering training and inference of foundation models for science, business and society. Klüwer previews a planned EU AI gigafactory initiative, part of the European Commission's AI Continent Action Plan, envisioning up to five gigafactories across Europe with around 100,000 GPUs each, roughly four times the size of the Jupiter exascale computer, alongside EUR200 billion earmarked more broadly to boost AI development. Germany is in discussions to host one; unlike the research-first AI factories, gigafactories are dedicated primarily to industry.

Q&A highlights

Asked about the gigafactory's location, Klüwer says the decision rests with the lead industry consortium, not the ministry, with several German states interested. Dell's AI CTO notes that half of the AI factories' capacity is reserved for SMEs, enterprises and startups rather than research, and that EU sites must federate compute and data across borders. On cross-ministry coordination, Klüwer agrees more is needed and distinguishes digitizing public administration from enabling AI innovation. On energy, she says efficiency is a funding condition and cites Jülich's exascale chip as among the ten most energy-efficient worldwide. Citing the Stanford AI Index, she notes German public acceptance of AI rose by 10 percentage points year over year. Asked if Germany is too late on compute infrastructure, she concedes it was not first but says the best time to act was yesterday, and now is also a good time.

Quotes

We're very good at inventions, but we are not that strong at innovation.

Tina Klüwer

Best time was yesterday. Now is also a good time.

Tina Klüwer

There's no option for not getting into the race now.

Tina Klüwer

Questions answered in this talk

Why is Germany strong in AI research but weaker in AI commercialization?

According to Dr. Tina Klüwer at Rise of AI 2025, Germany ranks third worldwide in AI research publications but only ninth in AI startups and twelfth in enterprise AI adoption, a pattern she calls the German or European paradox of strong invention but weaker innovation.

What do Germany's regional AI service centers do?

Klüwer describes four regional AI service centers that give small and medium businesses access to AI compute, GPUs, inferencing and expertise, while also acting as regional talent hubs located close to universities.

What is the planned EU AI gigafactory initiative?

Per Klüwer, it is part of the European Commission's AI Continent Action Plan, planning up to five gigafactories across Europe with around 100,000 GPUs each, about four times the size of the Jupiter exascale computer, with Germany seeking to host one.

Mentioned

AI gigafactory, technology readiness level, AI service centers, European paradox, AI compute infrastructure, Jupiter exascale computer

Keep exploring

More from Tina Klüwer

More from Rise of AI Conference 2025

More on Enterprise AI & Adoption

Transcript

Auto-generated transcript, not edited. · 4,477 words

Show the full transcript

All right. So, I was asked to do that uh quickly because we all know it's actually lunchtime and uh I'm hungry too. So, I'm going to be as fast as I can so that you can get to your lunch soonish or we do have enough time for questions. So, exactly. My name is Tina Clever. I'm from the federal ministry of actually research technology and space. But that is very fresh. So that change in the name and also in the responsibility was done last week. So I'm still also uh kind of getting used to this new name. Um I want to talk a little bit today about um invention and innovation in Germany. Uh there was a similar talk last time but now we want to dig a little bit deeper in what comes next.

Um so as you may know many of you may know there are several worldwide known innovations that were actually done in Germany. So for example aspirin cars, printing press, MP3. There are many examples that are very well known and there are some examples that are not so well known but still are German um inventions like the coffee filter. The first telephone that was the built by a physicist uh German physicist didn't work so well but there was the first idea there and Barbie for example as well maybe some of you have heard about that in the during the Barbie movie that came out lately because actually that originally was a doll that was um invented by the build newspaper um based on a comic that they had and that was a build lily that you can see in that picture down there.

So, and Ruth Hendel actually saw that on a business on trip to Europe and was inspired and then she built the Barbie out of that. So, uh interesting facts and many people know about it. So, um this is also why there is some certain interest in these German creative genius which is often caused. So, there are books about the German genius even from outside of Germany. So it's not only uh self-reflecting. Yeah. So we are the genius creatives but also from outside of Germany and it's I think germs are often perceived um as poetics and thinkers but also engineers. Um but what I would like to look a bit better deeper now is how is the situation today because many of the examples that I just brought are not from just yesterday.

Yeah. So they have been here for decades. um and how's the situation now and how is it in artificial intelligence? If we check on research which we can see is the basically the starting point of new innovation. So u what can we see is that Germany as the third largest um economy in the world is also the third um country in the world leading in um research publications. So we do have the US on the on rank first and then China and there's a big difference uh also from year to year because China invests more and more in research and publishes a lot of articles in uh artificial intelligence um but still Germany of course smaller numbers but um ranked three if we go further and say okay what comes out of those publications how's the situation about patterns Then we can see that Germany is on rank six worldwide which is still in the upper top 10 is which is clear but you can definitely see that there's a difference between rank three and

rank six and if you go further down the I would say the the the innovation process uh pipeline and look at the gym AI startups then you can see that in the number of startups that We do have in artificial intelligence, we are ranked nine worldwide. And one last number, if you check the AI adoption rate in enterprises, we are ranked 12th. So a lot of countries that we haven't had seen so in the top 10 or in the earlier positions in the publication rank more higher than us. And one last statistics um that I found really interesting is this is the uh analysis of the most used AI websites by monthly visits um with numbers from January this year.

And as you can see GPT is the um web tool AI web tool that was um used as the most but um there are other interesting things I didn't know about Canva. I was really surprised that Canva is actually the second AI um tool mostly used. Um but then if you check so the first German one that we have that comes here is ranked 13 with DeepL. So thanks a lot guys for doing that. Um so as you can see there's a certain degree of ranks um regarding Germany. So we're very high in the rankings of publication and research and the more it gets into application the more it gets into commercialization we rank lower and lower and that is well known as the European paradox which is also the German paradox.

So I would say we are one of the um countries that really incorporates that paradox. We're very good at inventions but we are not that strong at innovation. Um so if you check for the full innovation then we can see that Germany somewhere in the middle starts to derank and doesn't be isn't as uh active as before and the problem is that innovation is not invention. So if we check the technology readiness level um which basically give an idea about how long an innovation pipeline goes what the process includes for a technology. So that we can see we have the technological uh readiness level one to three which is basic research where we can define a principle have a concept and have a first proof of concept and then we have the development part from readiness level four to six with a lab validation and other validations in the real environment.

And then finally we have the stage of deployment from level seven to level 9 where it comes more and more market oriented including the final commercialization of that. And I mean obviously we do have a very good stake in the research part. We have a okayish good uh stake in the development part um especially in our u great enterprises. But there seems to be a certain issue in this phase and um therefore the government is also very interested in finding solutions to support those later readiness levels and to come up with more support for the businesses and to improve the adoption and the AI building part in Germany and Europe. So what comes next? We want to support businesses to use, adopt and create AI.

And that's a different viewpoint from many years that we had before where we solely looked at the research part of it. Um which turned out to work fine. Um and now it's not of course we don't stop invest in research still very important. We need to have certain ch we we have challenges in AI that we definitely need to tackle. For example, energy consumption is something where we need to have research on but also in more and higher quality artificial intelligence and trustworthy artificial intelligence. But additionally to that the support for businesses to use, adopt and create AI is one of the main u pillars and goals. So we want to strengthen the transfer by joint research basically between industry and research departments and the start of support and one additional very important aspect of that often overlooked but very uh crucial is the regional AI services and talent offering.

So if you have a business in a certain region and you need support then you need people. Yes. So it's not only about someone coming to your um company and just tell you like five minutes what what comes next and what would be a good off idea but it's also to deliver and offer those people that can work with you on the topic. What we have done is already we built the AI service centers. Uh the goal of the service centers is to support uh small and medium companies to get access to compute AI compute GPUs and uh additional things inferencing um offerings and AI expertise. There are four different services centers in the geographical distribution among Germany.

So we can see they provide services for every business that is interested and has questions in artificial intelligence. But they also act as a regional ecosystem um for talent. They are often very close to universities and research but they do this uh they offer this outline that you can as a umme contact them and they they can support you. Another thing that is currently undergoing is the AI factories by Euro HPC. Um many may have heard about that. It's a big initiative of the um EU Commission to build new uh supercomputing centers in Europe. The goal is that they enforce dynamic ecosystems around those new machines. Um they are all AI optimized supercomputers. They offer data capabilities and pray programming and training facilities and also human capital.

So if you need people then most probably you may find some there. So we do have in Germany already we have built up two of those AI factories. Um they are financed by uh Euro Vision by the uh um Euro HBC uh and by the public uh government and we have the Hammer High in Stoodkat. That's the first one. They focus on engineering science and global challenges and industrial HBC. And we do have a second one, the one in Uulish. Um it's an additional one to the supercomputer that is still there, the uh and they offer training and inference of basic models for science, business and society. And they have this experimental platform for inference which is like the USP of their machine.

And this is not the end of it because many of you may have also heard about that. Um there's more to come and this is the AI giga factory initiative. So um this is part of the new commission's AI content and action plan to become a global leader in artificial intelligence and we are of course as a federal government also part of that initiative and our this the objective here is to develop and uh provides AI models and AI applications of a completely new scale which means there are in the planning we have up to five gigaf factories in Europe with approximately 100,000 GP GPUs which is huge. Um it's four times the grow uh the size of the Jupiter exoscale um computer and we want u of course to have one of those um gigaf factories here in Germany.

So we are currently in conversations with everybody who's interested in that goal as well. And if you feel like you would like to get in conversation with us about that, please contact me, talk to me. Now is the time to actually think about the next steps here. Additionally to that big machine, there are 200 billion euros um to boost AI development. So this is more about uh projects and programs that are um done on the site to um that the machine is used and also to come up with more uh usage and adoption here. Yeah. And what is actually the main difference between the gigafactories and the AI factories that I've shown in the earlier slide is that um the gigafactory is dedicated to businesses.

So it's a it's an industrial machine industrial machine and there is research also available computing power available for research but it's second priority. So first priority is really to have a a big machine and this offering of services dedicated to enterprises and um businesses. So if you feel like this is something that you need uh just uh let me know. All right that's it. Thanks a lot. Thank you. First of all, I have to apologize because I actually knew the new name because I worked last week with Rabia Rogo, our first female space, um pilot, so to say, our fa Yeah. astronaut. Um uh as someone who comes from research and technologies, I'm not really happy that education is not part of your ministry.

It's about the talents of the Yeah. Okay. Well, well, what to do? But space that's amazing moonshots. Um, do we have questions from the audience? Yes, please. Can I have the microphone here in the third row? Um, hi. I wanted to ask if you already know where this Gigafactory is going to be located at and al also if you're considering decentralized uh alternatives like to spread the compute across the infrastructure cause uh latency might be an issue for robotics. Yeah. So that's the the most uh hard the hardest question that you can ask. No, I don't have the final location. Um because it's not yet decided on. Um and I also want to emphasize that the final location is not something that the ministry decides but um decides about but it's something that mostly is um in a decision of the lead industries that will later then build up the the the Gigafactory.

So um there's a lot of support as you can imagine many different um lender are actually really interested in getting the gigaf factory and that's great to see um but how this race in the end will turn out we can't say now yeah first question here Johan I got it's it's again for those who haven't been here yan stra chief technology officer of Dell um regarding the 12 or 13 Ju sites before the gigaf factories. There's there's one good news for you which I didn't see on your slide. 50% of that capacity has to be brought to the public. Mhm. So yes, research first. So their priority is research, but 50% of the capacity of Hammerhai and ULI chief has to come to Germany to SMBs, enterprise and startups.

Just a site comment. Thanks. Thanks. And this is also a second site comment on the federation of these technologies. All these sites across the European Union have to be able to federate the compute and the data. So that means pabytes of data that are sitting in Finland you could tap into and use over here. if you have a data contract, you know, we can discuss that later, but that's that's that's the thought process we're going through. And we we've been sitting about four months ago with the JU sittings like what is the platform you need to build rather than just Yeah. acceleration technology. Which by the way, um there is over a dozen accelerating technologies, you know, it's it's not only Nvidia GPUs, but another time for that one.

There was somebody else. Yes. Here. Can you hand over here to Frederick I think? Yes. Thanks. Um so yeah uh having great what you what you are doing till here and and and what you are trying to achieve in the next months. I mean for all of us we are very happy to have now digital at least maybe be a bit more coordinated having worked up until 7 months ago in the federal transery uh as a co um uh I I've seen also what great work you did in the BMBBF back then and for me that's one of the biggest task I think right now is AI will still be split between the ministries that's something that I'm really thinking a lot around and uh also like the switch in the last moment from Agitra to the other AI with the gigafactory stuff when the sentence was moved and stuff.

H how do you think we can better coordinate that? I mean the AI Act was a classic example of how we totally failed to be honest um to coordinate ourselves good that was not on on your side from my point of view but uh maybe that's for later but I I'm I'm really I I would really emphasize this should be something that we all have to work up on now in the ecosystem that AI really gets a strong voice from German politics. Definitely. So I totally agree to your point. Yeah. Thanks for taking the time to to highlight that again. Yeah. So because it's crucial. Yeah, I agree. And yeah, as well. Yeah. And I think what we need to do is definitely to double check and what are the concrete tasks that we have regarding artificial intelligence in in in what ministry?

Right? Because I think basically artificial intelligence is of course it's a very broad technology. So it's somehow obvious that it occurs at many different places. Yeah. Um but what we can do is of course that we double check what is the exact duty and the exact task at hand. And I think there is a diff difference between getting the public administration being more digital and also using artificial intelligence for that between this task and the task of enabling Germany to become even more innovative in AI to bring up the new ideas and to also bring up new startups and new businesses here. So I think those are different tasks and although they can say well we have to do some departments somewhere.

Yeah it can't be allin-one unfortunately. So but if we have the chance to really make that very clear then I think we do we can we can have very good support for artificial intelligence throughout these different duties. Yeah thank you. You see that we already are at 1 p.m. but I have three more questions in the audience. Is that okay for you? are too hungry. Too hungry. I think it's too interesting. As long as we're got not getting hangry, please go ahead. Um, thank you. Each of these data centers will consume at least 300 megawws of energy and you are competing with the electrification of cars and other things. We don't have the energy. We don't have the transmission grid and we don't have the cooling water.

Especially if you see the draw and this entire thing will produce within five years when the Nvidia graphic cards are completely outdated and AI waste of whatever factor of thousand compared to what data centers produce at the moment. How do you deal with the envir environmental constraint on that? Is this part of a concept or is this just ignored for No, it's definitely part of the concept as I already said. So energy efficiency is one of the biggest uh things that we still have to look at and also find more and better solutions but already now there is a big and huge part of the of the different ideas and also uh it's a condition of um being successful in these applications that you have a very concrete ideas about your energy uh consumption and if you compare it for example with the Jupiter exoscale uh computer that is now built in in Ulis.

It is has already now a very very good um consumption rate compared with other computers that are even a lot smaller. Yeah. Because um also the the the energy efficiency technology evolves all the time and everything that can be done is definitely be done at the moment it comes into serialization and commercialization. So which means that the new exoscale uh chip that is now enabled there is actually one of the 10 most power um energy uh saving chips that you can have worldwide. So there's a lot done there of course always in the boundaries of what is available now. Yeah. Thank you. You have question please. Hi thank you. Um I'm Diana Hines. I'm from Bavaria Gamin.

Um my question goes um is not in the data center direction but more in the research and education direction. Um I'm in the management board for the geriatronics project in gam patin. It's a research project where robotics should be um should support older people. So for elderly care together with the technical university of Munich and Karas as well and uh yeah we as um yeah in this management board and generally we are really con getting confronted every day uh to yeah convince actually normal people like my grandma um what does the rob the robotics are doing with these people and how so we trying to find programs and initiatives to help these people understand um that um they can actually also trust uh the technology and um that this um yeah technology is also controlled by humans and my question is actually um I mean I talked also with um Marcus Bloomer um about that and yeah my question goes more into the what what initiatives do we have um not only in Bavaria but

maybe also in in Germany that helps to un helps the people to understand what is AI general AI programs things like that to really democratize it. Yeah, thank you. Yeah, thanks. That's a very good sounds like a very good initiative and also very good question. Of course, uh I'm not sure if you've seen that, but in the Stanford AI report that came out this year, compared to last year, our acceptance of AI in Germany has increased by 10%. That's a lot. It was really low. Yeah. Yeah. According to your presentation last year, it was very low. But so it increased. So there is a more and more acceptance of that technology and I think also people now more and more look at the chances that we have with it and exactly what you just described I think is one of the biggest chances that we have and it's one big chance as well from my perspective for for Germany especially because we do have this big and industrial strengths in robotics and if you combine

that together I think this is really like some it's a big Yeah, could be a big benefit for us. And um there are certain um yeah like more educational programs that we have. We have some from the federal ministry as well, but there are many on the lender uh as well. So we have the the the KA campus website for example. that has been online for quite a while already where you can have different coachings and um things and there are more and more courses available and I think that's also one of the reasons why this acceptance increase has happened in the last year. So yes, so there are things available, but I totally agree there's more to do because if you compare it with China or the attitude that other Asian countries for example have towards artificial intelligence, there's still a big gap and I think um with all the risk management that we do and I think that we do it of course for good reason.

Yeah. So definitely we need to do the regulation uh conversations and we need to double check every risk that may evolve of that but still we need to also focus on the chances and that's something that from my perspective and from the numbers as well has significantly changed from the last year as well. Thank you. Well there is there's improvement but there's also room for improvement. Definitely. Last question. Thank you. Uh my name is Wfgila. I'm from the German edition of Technology Review magazine. And uh I don't want to be the party crusher here. Uh and I really like to see that we will get these AI factories. Uh but um my question is isn't that a little bit late?

I mean it it was no secret that there was a revolution in deep learning. It was no secret that uh companies like Google built the transformer architecture and we had a year-long discussion and discussion and discussion about more compute here in Germany uh and more compute in in Europe and it was delayed and delayed and delayed and my question is if you agree that uh this buildup of compute infrastructure is a little bit late what can we learn from it and how can we avoid that latency in the future? future especially looking at elsewhere countries like China which is building up a massive open-source infrastructure besides that buildup of compute. Yeah. So I would love to have the final answer to that.

Um I agree. Yeah. So we we haven't been the fastest in the game. Um maybe what's interesting about that sometimes it's possible not to be the fastest but still succeed. Um if we check with the situation with cars for example it wasn't the Germans that came up with the cars first. It was more something from UK and for example, right? But we still managed and I think um yes, you're right. We weren't the first, but still there's no option for not getting getting into the race now. Yeah. So, it's just not an option. We need need to do that because if you double check on what comes next, new technologies, right? like quantum computing for example or neuromorphic chips or more technologies that we do have in the pipeline worldwide now and who's active in that technologies research is of course but it's not only research it's also the big digital businesses from the US and they know how to build those things and they may be fast with that too so and we are

lagging behind because we didn't do do the first step and therefore from my point of view there's no uh alternative to not just to not come up with that business uh and artificial intelligence ecosystem here. Yeah, best time was yesterday. Now is also a good time. Yeah, nicely put smarter smarter. Well, we learn from we learn from our mistakes from the la from the past and now have ministries for digitalization on um and the ministry for research technology and space. And with that um thank you so much Dr. Applause. Have a nice day. Thank you. We now we now eat lunch. I see you in at uh 13:45 1:45. So bit more than half an hour and please come back because then we talk about entrepreneurs, we talk about diversity and we talk about my favorite bird if someone is interested.

We talk about penguins.

Rise of AI 2027 · Berlin · May 2027

Be in the room for the next one.

Ten years of talks started as conversations in one curated room in Berlin. The next edition brings 300 selected leaders together in May 2027.