Rise of AI Conference 2025 · Berlin

Alternate Paths to AGI

Rafael Laguna de la Vera · Director, SPRIND (Federal Agency for Disruptive Innovation)

At Rise of AI 2025 in Berlin, Rafael Laguna de la Vera, director of SPRIND, Germany's Federal Agency for Disruptive Innovation, argues that Europe needs a more optimistic, self-built vision of the future and describes how SPRIND tries to fund the breakthroughs that create it. He traces SPRIND's founding to Angela Merkel's 2017 decision and the four years it took to pass the SPRIND Liberty Law freeing the agency from standard ministry procurement rules, modeled on how the US founded NASA and DARPA after Sputnik. He argues Germany produces twice as many patents per capita as the US from strong basic research, yet only a twentieth as many unicorns, because later-stage funding is scarce and 80 percent of trade sales above a 25-million-euro valuation go outside Europe. He details SPRIND's two funding tools, fast 'challenges' with concrete deadlines and always-open project proposals, illustrated by a 35-million-euro broadband antiviral challenge launched during COVID, and describes SPRIND's push into new AI architectures, distributed compute and alternative computing hardware such as neuromorphic and optical computing to reduce dependence on massive GPU data centers.

Key takeaways

  • Laguna says it took four years to pass the SPRIND Liberty Law freeing SPRIND from standard federal procurement rules, after ministries had stripped its promised freedoms.
  • He argues Germany produces about twice as many patents per capita as the US but only a twentieth as many unicorns, due to a later-stage funding gap.
  • He says 80 percent of German trade sales above a 25-million-euro valuation go to buyers outside Europe.
  • SPRIND's 35-million-euro broadband antiviral challenge, launched during COVID, funded nine teams at 700,000 euros each within about three weeks of application.
  • SPRIND's budget grew from zero in 2020 to 250 million euros this year, having launched 11 challenges and funded roughly 200 projects.

Detailed notes

A positive narrative for the future

Laguna opens by arguing that Europe has lost optimism about the future, partly because social media algorithms feed negative content, and that the best way to predict the future is to build it. He tells the audience that as AI builders they carry responsibility for shaping a positive future narrative, comparing powerful technology to a knife whose use depends on the people building it, and argues creating and communicating a positive vision is the necessary first step.

Why a government agency funds breakthrough innovation

He describes SPRIND as the government agency for breakthrough innovation, founded because very early-stage, very expensive innovation, such as AI gigafactories with 100,000 GPUs, rarely attracts private capital. He draws the parallel to the US founding NASA and the DARPA after the 1958 Sputnik shock, freeing those agencies from normal bureaucratic constraints. Angela Merkel decided to create a German equivalent in 2017; Laguna was brought in in 2019 to help build it, only to find the promised freedoms had been stripped away by ministries during the founding delay. It took four more years of work, credited partly to colleague Mario, to pass the SPRIND Liberty Law that finally gave SPRIND freedom from standard federal budget and procurement rules, signed on December 30 after Laguna made his contract renewal conditional on it.

Germany's patents-to-unicorns gap

Laguna argues Germany has strong basic research, writing roughly twice as many patents per capita as the US, but only about a twentieth as many unicorns, because startups can raise seed and Series A funding but struggle to secure later-stage capital or complete an IPO domestically, often exemplified by companies like BioNTech listing on Nasdaq instead. He says 80 percent of trade sales above a 25-million-euro valuation end up outside Europe, calling this both an economic loss and, in his view, a threat to democracy if Europe fails to build the industries of the coming decades.

How SPRIND funds innovators

He explains SPRIND runs two funding mechanisms. 'Challenges' pose a concrete problem with an eight-week application window and a fast selection process; he illustrates this with the 2021 broadband antiviral challenge, launched with a 35-million-euro budget during COVID to find fast, mutation-adaptable antiviral treatments, which drew 47 applications by the deadline and selected nine teams for 700,000 euros each in year-one funding, paid out as revenue-generating orders within roughly two to three weeks of application. The second mechanism, always-open project proposals, offers 'validation contracts' of up to 200,000 euros to evaluate promising ideas, followed by grants up to one million euros for new companies, co-investment with private venture up to 70 percent of a funding round with no fixed cap, or grants up to 35 million euros for early technology requiring a roughly 25 percent private contribution. He says SPRIND's budget has grown from zero in 2020 to 250 million euros this year, having launched 11 challenges and funded around 200 projects, including a 45-million-euro laser project and an 80-million-euro Alzheimer's clinical trial.

Applying the model to AI

Turning to AI specifically, Laguna says SPRIND's strategy, published on its website, argues Germany should not simply build ever-larger GPU data centers but invent new model architectures, new data modalities and 'compound compute' that can train models while keeping data distributed. He cites the agency's compound learning challenge, already running with seven teams, aimed at reducing dependence on centralized 100,000-GPU clusters, and describes plans for a further challenge combining new AI algorithms with new computing hardware, including neuromorphic, in-memory, optical, reservoir and quantum computing, aiming eventually to shrink AI compute the way DARPA shrank GPS receivers down to fingernail size inside smartphones.

Q&A highlights

Asked how he manages expectations for technologies like fusion energy or AGI, Laguna says honesty is the only approach, comparing his own 15-years-plus-or-minus-five estimate for fusion to the wide range of answers AI experts would give for reaching broadly useful AI tools. He rejects the idea that outcomes can be reliably predicted, describing SPRIND's approach as a 'serendipity culture' that allows for happy accidents, and says that if outcomes could simply be calculated in a spreadsheet, an agency like SPRIND would not be needed.

Quotes

The best way to predict the future is to build it.

Rafael Laguna de la Vera

Who owns AI will own the world.

Rafael Laguna de la Vera

I'm a general manager not an astrologer.

Rafael Laguna de la Vera

Questions answered in this talk

Why did SPRIND take four years to become fully operational?

According to Rafael Laguna de la Vera at Rise of AI 2025, although Angela Merkel decided to create SPRIND in 2017, ministries had stripped away its promised administrative freedoms by the time he joined in 2019, and it took another four years to pass the SPRIND Liberty Law restoring them, signed on December 30 of that year.

Why does Germany have fewer unicorns despite strong research output?

According to Rafael Laguna de la Vera, Germany writes about twice as many patents per capita as the US thanks to strong basic research, but only about a twentieth as many unicorns, because later-stage funding is scarce and roughly 80 percent of trade sales above a 25-million-euro valuation go to buyers outside Europe.

How does SPRIND fund breakthrough AI and technology projects?

According to Rafael Laguna de la Vera, SPRIND runs time-limited 'challenges' with fast selection and payout, such as its COVID-era antiviral challenge, alongside always-open project proposals offering validation contracts, grants up to 35 million euros, and co-investment with private venture up to 70 percent of a funding round.

Mentioned

SPRIND Liberty Law, compound learning challenge, compound compute, neuromorphic computing, in-memory computing, optical computing, reservoir computing, quantum computing, broadband antiviral challenge

Keep exploring

More from Rise of AI Conference 2025

More on AGI & Future of Intelligence

Transcript

Auto-generated transcript, not edited. · 5,717 words

Show the full transcript

starting now our afternoon session. Um really glad that you're all here again at the main stage. Our next keynote speaker has its own freedom act. is his main point of operations. Rafael Laguna de la Vera is the director of the Bundesour. He's a businessman, a venture capitalist focusing on technology and information. And I think he founded his first startup when he was 16. Yeah, Elephant Software. Really cool. Please visit with a warm welcome. Say hello to Rafael Laguna. Thank you very much. Thanks everyone. That was a long time ago as you can see right 16 but it was a real company. Um in a way I was always driven by one thing which is sort of being curious about the future right the future is the sort of exciting uncharted land and being optimistic for the future is just a lot of fun.

Unfortunately, we've lost some of that because we're not being all that optimistic anymore, which is kind of a problem. And I will tell you why. And then I will tell you what we're trying to do so that we change that. Because somebody much smarter than I said when he was probably younger than I was when I was founding my first company said, "The best way to predict the future is to build it, to make it." Right? And this room is full of people building future. You have a higher responsibility for everyone else that may be worried about the future and is not so deeply involved in building futures. Most people just want to live a normal life and be left alone with problems and do their thing.

Whatever it is, you guys are here because you want to build a future. Otherwise, you wouldn't be here at an AI conference. whatever that is, right? I know we all don't know, but we don't tell them, right? It's something that's kind of cool and that will obviously do something to change the future in a big way. Now, if we believe that to predict the future is the best way is to build it and we're building it, what's missing? It's a vision of a positive future that we want, right? And if you're opening your social media accounts and you're seeing orange men doing wild things, bald men doing wild things, mostly men doing wild things, it seems, really bad things.

All our stable world is changing and we're reading all this crap from algorithms that we're not controlling that are feeding us, you know, with things that take away our optimism for the future. Now that needs to change with good journalism and we have to reinvent how we do that and how we finance that. That has to change with us by first of all creating a positive future narrative. Yeah, I have a t-shirt that says pessimism is a waste of time. Now, if you want to sound smarter, you can say Carl Pa once said that optimism is the only viable option. Yeah, it is. I mean, what do you get out of pessimism? A bad day, right?

And then the next day is bad again. So, creating narratives and telling them to the world is the first thing that we need to do, especially when you're working in technologies like this. Every technology is like a knife. And the more powerful it becomes, the more powerful a knife it is. Now, you can do different things with knives. And what we're going to do with knives is decided by people like you and by what you have in your head as your perceived future that you want. So I was telling you I will be unloading a lot of responsibility on you right don't worry but it's true because you're the future builders. Now how do we go about this?

Let me give you the example of SPRIND because I think it's an interesting story and some of you sort of have have ch shared the experience. Mario, Mario, show yourself, please, because you have to come to his presentation at five at three, I'm sorry. Mario was was, you know, helping us get this sorry, bloody law. It took four bloody years to get this SPRIND liberty law so that we can work the way we always wanted to. It took four years, but that, hey, it's democracy, right? It's hard. A lot of people have a say in it more than should maybe. And then you need, you know, state secretaries that just, you know, ram through, push it through.

Thank Thank you very much for that, Mario. Thank you very much. Now, now he needs a vacation of five years or something, I believe, because it's been so hard. Now, what does it all mean? You may ask. I mean, what is SPRIND to begin with? We're the government agency for breakthrough innovation. And somebody said it's like the ministry of funny walks, right, from from Montipython. And in a way, it is, as you can see, but it is a ministry. And you may ask why is a ministry even involved in in in breakthrough innovation? Why a ministry? I mean we have well the answer is most breakthrough innovation governments were involved because at the very early stages you need a lot of money.

The private sector very more often than not is not willing to give you that kind of money. All of you being in fundraising know that or if all of you being VCs know that because you are fundraising all the time as well. It's just hard to get money when you have a vision. Now really big B breakthroughs are also very expensive. So I mean we're talking about AI gigafactories with 100,000 you know GH100s or so. I know why Nvidia is a sponsor, right? That's great. Fantastic. But it's billions. It's billions and billions and billions. And so why don't we have them? Well, because nobody is putting down that money from the private sector because we don't have it.

So if you want it, you have to put down the money to get it. in the US, of course, always being a little faster in with this stuff. In 1958, after the Sputnik shock, after a little metal ball was flying around the world, I think two or three times, sending down beeps, that was all, you know, President Eisenhower realized, oops, you know, the Soviet Union is ahead of us in the space curve and who owns space, just like AI, who owns space will own the world. Who owns AI will own the world. Same thing. So, he said, obviously, our ministries you know have fallen behind. Our policies have not led to us being the leaders in these technologies.

And so he founded NASA which we all know moon program following successfully spawning off tons of technologies that we wear in our pockets and elsewhere and an agency called ARPA the advanced research projects agency later renamed to DARPA defense. It's pretty much what we do now right now know everything is defense right they they did that also somewhere somewhere down there and they freed them from the normal shekels on how ministries have to work and that was the idea of SPRIND now Angela Merkel decided I want one in 2017 and then two years nothing happened and then an arbit was started right and u I was called in to help designing SPRIND this agency innov Deutseland.

That's what SPRIND means. And I thought, what a funny name. Never heard, but it sounds like fun. And of course, building something, you know, it's truly supporting innovators, you know, in a fast way that love them, that are friends of these innovators and not people sending forms that you have to fill in 400 pages and stuff. sounded like a really good idea because other than Eisenhower in 58, we have very good basic research and some of you come from basic research in AI. We've been doing that for decades, right? Unfortunately, the companies are not here. So, there is a gap between basic research and unicorns. Germany writes half the patents that the US writes, but we only have a fourth of the people.

So per capita we're writing twice the patents. That kind of speaks to our basic research. And researchers hate writing patents. You know that specifically basic researchers don't like doing that very much because it's just work. They want to publish papers in nature, right? That's what they do. Well, anyway, we still write twice as much patents as as the US. But at the end, we only have a 20th of the unicorns. So somewhere on this way we're losing it and all of you entrepreneurs know that after well you probably get seed funding here you get a series A here 20 million some get 100 million or if your name is Yonas you get 300 million right that but that's about it right there's it's very hard to sort of get the later stage funding and it's super hard to do an IPO here even if you're successful here like Biontech you're doing an IPO at NASDAQ.

So you can see, okay, there's a problem. And it always ends in a trade sale when you should be doing a series B, when you start scaling. And 80% of the trade sales above a valuation of 25 million go where? Outside Europe. That's where our unicorns go to die. And this is what we need to fix. It's a threat to democracy because if we are not creating the industries for the next decades where we'll be having more people voting the wrong stuff and democracy will go to die and it's not far out. You can see that. So SPRIND's mission, very simply put, is to make sure we have more unicorns to. So we went out and said, "Okay, what's the problem?"

And we found out the first big problem is administration because we've been burdened with an administrational load like a ministry even though Angela Merkel and our cabinet decided to make SPRIND totally free of everything. You know, you get we get a budget over the next decade of a billion. We're not you don't have to adhere to the bundes households ornong to the baselong for board and all this super compound nouns that they invent. Unfortunately within those two years where nothing happened the ministries had enough time to kill all of that and none of the freedoms were there when I came in 2019 and was asked to help build this agency and then they talked me into taking the job.

I was running my own company at the time with 300 people and all that, but I said, "H, I'm 55. You know, I've been to Harvard. I learned time to give back is a good thing to do. If you're that old and you have enough money, you don't need more money. I mean, how many schnitle can you eat, right?" And so, I said, "Okay, let's let's try that." I thought I had the freedoms that were in this in this uh uh original document. What I did not know is that they had all changed to something that couldn't operate the way we wanted to. And then the next four years, as part of my normal job, which was funding innovators, I also had to invent myself and eventually come up with a SPRIND liberty law that that Mario helped bringing into the world.

It took four years and now we have it. Actually, when I when I was asked to sign my contract for another four years, I said I sign for one year and then maybe extend for another three if I get this law. The law was signed on on December 30th of that year. Right? So, a little bit of pressure helps, right? To to get it to where we want to. But now, what we are now, and that's the good message, we are a very unique government agency that does not exist in the world in the same way because we can do funding tools that governments usually have. So we give you, you know, for example, I'll explain how that works in a second, but also funding tools that private venture usually does and we can co-invest with private venture.

So at the early stages of an innovation when there is no private capital, we just give you money. I'm sorry, do it, but please deliver results. We we have a lot of trust in our innovators because innovators want to make that thing work, right? And the only thing that we check is are they making progress? And if that sort of changing path, which happens a lot too, we're saying okay, is it still a leaprog innovation? Is it still breakthrough? Is it still disruptive? The answer is yes, we say go. If the answer is no, we say stop. No more money. Not only money is important, innovators. I mean there are some types that are business people and technical people at the same time that have a lot of social competency that can speak well on stage and blah blah blah but let me tell you this is very rare.

So you need to find a team and they don't show up with complete teams. So we help them complete their teams. We train them how to pitch. We train them how to read a cap table and all these things because mostly technical innovators don't know right because eventually they need to pitch in front of private investors. So we make them bankable. We give them all the money they need to continue developing. We give it to them very quickly because I mean if you who who of you has ever applied for government funding for your venture did you like it? Was it fast? How was the documentation task? Wonderful. Right. So, that needs to change because we don't I mean it's kind of crazy when you think, right?

Tina Ka I think she she already talked. Um she uh she's a director general at at the Ministry of Science and and these other things that they've now put there. Um she she said if I want to if I would need to set up a program for something, right? Let's say AI gigafactories. I have to make a call for a project. That's that's an or government organization near government organization that puts these things to work. But this call takes 6 months, 12 months or so because there are several of them and you have to be fair and what have you. So then you select one of those and then you say please develop this program.

Takes six months. program comes back the politicians minister decides yeah that's what we're going to do what's next you call you make a call for a project that now executes this call so now you have one another six months or so you give it to them they make the call they tell the world please apply and they tell you and stop working on your thing because that's against the funding rules that we have you cannot start working on your thing it will break your application you won't get money. So you go pause and then you wait for the project to evaluate your work. You may be talking to dozens of smart people and eventually they pick you and they say, "Yep, it's you.

It takes another three months before you get the first monies and then three months later you have to send documentation on how you spent the money, right? We're talking years before you have the money and the administrative administrative burden that you have is I don't know what 20 30% if you're a profit it's probably 80% to do all this documentation it's crazy and this is money to fund these people to make them faster. So we said okay speed right speed simplicity how can you achieve that? Instead of mistrusting our innovators that they buy the wrong pencils and go at the wrong conferences we say you do what you want as long as you achieve the goal you promise to us.

We were looking for those goals in shorter cycles. We don't give you money for three or five years but maybe for six months or a year or so. And maybe we not only give you money, but maybe some of your competitors out there that may use different technologies. We don't tell you what technologies you're using. We want to you to achieve the goals, right? And if you don't deliver the results and you cannot explain to us why, then you're not getting the additional funding and that's the end of the story, right? And we don't count pencils because we don't care. If you're drinking champagne all the time and you're reaching this goal, have our money.

Doesn't work that way. Unfortunately, I tried but it not for long at least, right? You feel good for a while and then well then you die. So we do this in two ways. What one is what we call challenges where we say hey we need a solution to one of those big questions of our time. So how about getting rid of virus infections? Guess when we started that was our first challenge in 21 in the middle of corona with this whole everything corona happened because we did not have a medication against this virus right which well when you talk to experts many said well it's very complicated they mutate all the time and what have you yeah things are complicated I mean HIV virus is complicated and we thought we were never going to solve it and we solved it right so that's what innovators do hey right impossible stuff that's what that's that's what I want to do.

So we started the broadband antiviral challenge and I give this one example then I tell you a few more that we made which said find a solution one that is quickly adoptable to mutations time to market is critical for these things best is broadband works for all viruses that's probably unrealistic without modification but anyway something that's faster than what we have now because we had calculated corona pandem if you had had a a a vaccine or a medication against the virus within the first 100 days days more or less nothing would have happened because it hadn't spread at the time very much. So we would have saved trillions and a lot of people wouldn't have died and we wouldn't had to lock our children away our teenagers at home which is even worse I think than most of the other effects that this had.

So you would say, oh that's a goal to have, right? We put down 35 million, which is not a whole lot for a government funding program. We made the call. The call lasts for for eight weeks. Gives you time to write down your proposal. You send it in 5 minutes before midnight, everyone, which is true by the way. We had like two applications until uh 11:30 at night and then we had 47 at midnight. Right? And we were sitting there like nobody cares, right? Which wouldn't have been great. Then we looked at these things for for two weeks with our jury which was external experts, you know, very diverse from venture capital, from science, you know, from from all sort of walks of life to evaluate the best proposals.

We invited 20 teams. They showed up for two days, crunchtime pitches, ding, ding, ding, ding, ding. And we selected nine teams. and they got 700K for the first year each. They send an invoice. The contract is 14 pages. The contract is published with a call so you know what you're signing. It's called pre-commercial procurement. You know, it's an order. So, we place an order. If you're a startup, it's revenues. We don't show up on your cap table, right? It's revenues. Investors like revenues. You send the bill, you get the money 3 days later. So the time to money from you applying at 5 before midnight to your bank account is something like 14 or 21 days for this.

Right. The second thing that we do is project proposals. Anytime 24/7 365 you send us your pitch deck and documentation about your project. needs to have sort of a TRL of at least three meaning no basic science because that's well funded in Germany right with Maxplank and all these guys and universities I think we're doing well if you look at the money we're putting in there but now we want to make it applied science we want to get it out of the world so if you're at this stage where science funding ends or your garage budget you know it doesn't need to be science it's can be in your garage whatever you want it can be you single person doing that we don't care you don't have to be further F.

Everyone is fur with a good idea, right? Compound nouns, I tell you, lots of them. Now, we send the proposal. If you like what we see, we ask you a few questions and give you money for answering them. We call that validation contract. It's up to 200k. It's not super much money, but it carries you a couple of weeks. So, opposite of what the other funding instruments do is we say keep on running, right? but help us evaluate more. So this is a paid due diligence if you like that helps you continue the work on your project and also to get to know each others to help you build a team to connect you to you know important people from government science what have you because that's one of the perks of being a a federal government agency everybody talks to you and even feels good about it because we're not the IRS right we're not so we're kind of liked which is good because we need all these experts all of you out there

to help us evaluate the project and help the project push push forward. So now if we like what we see, we now have thanks to SPRIND Liberty Law several means. We can if you if you're founding a new company, it's not older than 5 years, we can give you up to 1 million in form of a SU vendor. That's that's a grant. Now, if that makes you bankable, we can co-invest with private sector up to 70% of the round only limited by our budget. So there is no number that limits us. It's not 1 million and we can do five, we can do 50 and we've done 50. So that whatever it takes. Now if you're not bankable yet we can continue with grants and that can be if it's early technology up to 35 million from us but you need a small percentage 25% sort of private contribution which can be many things can be money but it can be work whatever right that carries you across this 25% bump of course with the

goal to make you bankable make you financeable and then we can come in with these instruments. So eventually our money also gets a share of the company and potentially generates a return but we're not focusing on returns. We're we're focusing on return on another compound noun folks shaft on you know what the overall economy economic impact is of this thing that we're doing because we're we're spending tax money. We've built the agency now from zero in 2020 to 250 million in this year. We've even been growing in this year which not a lot of government agency can say because as you might know we didn't have a government for quite some time. We didn't have a budget for this year.

This is now all changing. And the third thing that we do is because we're being exposed to all these technologies. We're running all these projects. We've seen two and a half thousand projects. We've launched 11 challenges. One call is open by the way. It's the autonomous flight challenge. It's the second time we try this. First time we were in airing. We had a an obstacle course, you know, fog and rain and all these complicated things and flying things mostly called drones had to find their way to a box and bring it back without global navigation systems, right? So, edge AI on the drone. Nobody made it. I mean, one team made it back actually, but without the box.

So, we said, "Okay, give it a year." And then we try again. And that's what we're doing. This call is open. If you go to our website, sprind.org, call runs for 6 weeks or so and then you know what the process is. You know, two weeks later you will have your money and can go and then we'll try again and see see what happens. It's kind of the real fun stuff because that's a real competition at the very end. You can imagine we funded another 200 projects from the project funding. The amounts are sometimes very low and then they get private funding and we go out. We're crowding in not out. So if there's private money enough, the private sector gets to finance these projects and some we had to put down like 45 million for laser or 80 million for a clinical phase 2 for an Alzheimer's medication and things like that.

You you find all these uh on our website in our podcast this where the innovators talk which is also very interesting. Have have a listen to that. Now comes AI and now what should government do to bring AI forward? Of course, we're lucky enough to talk to many of you and many others to develop a strategy and we put down these strategies on paper and we give it to the governments to our government. And if you look at the coalition agreement, you will find a lot of topics that we're working on like fusion energy or AI obviously and all these other things in the coalition agreement. Now the big question is will we operate the same way we've done before sort of the slow way or do we do it the fast way this time do we leapfrog or do we follow so for example if you read our AI strategy which you find on our website just Google SPRIND and AI and you find the document it will say well G GPUs are great

and we need them now but we should use them to build new model architectures and support new data modalities. We should invent compound compute so to speak. How can we distribute? How can we keep data locally but still train a model? Johannes launched this. Johannes, he's the man with the money, right? He's Mr. AI at SPRIND. Um the compound learning challenge is running already. So people already have their money. We have seven very promising teams uh that will free us from the shackles of having to have 100,000 GPUs in one data center. Even having to have GPUs to begin with because the other thing that we want to in invest in from aside from other algorithms is new computing architectures.

You know, we're we're doing compute like John Fonoyman invented 80 years ago and building nuclear power plants next to data centers is just not a very good idea. Especially in Germany because we don't build them anymore will take 15 years or so to build a fusion power plant. But that is crazy. I mean to build one brain that consumes the energy of a power of a nuclear power plant is probably not the pinnacle of this technology. We were very early and of course one way to fix it is better algorithms. That's one challenge and the other way to fix it is build better hardware post forman and we will be launching such a challenge. We already did an architectural challenge for this.

We will be opening a challenge where we want teams that build new algorithms on the one side and that build new hardware on the other and we want to buy systems at the end of that. It it will include include quantum computers but that's just one postman architecture. It's one special analog computer. There's so many others out there. Now there's analog computers which we like a lot uh including quantum computers. I'm not sure they will work eventually but that's how it is. But then there is neomorphic computing in memory computing optical computing even reservoir computing which is biological computing and these things have a lot of advantages when it for example comes comes to power consumption right and so we need to bring down the power that these things need.

We need to bring down the size of the data center. We probably need to get it down to a fingernail. DAPA brought the GPS receivers from boxes like this down to a fingernail that's now in your phone. We wouldn't have smartphones without DAPA because they just sized these things down. And we have to do the same. We have to bring the power that's now running in a data center with 100,000 GH100s onto a fingernail. Now, that sounds crazy, right? And it's probably not next year, but it will be in a couple of years. And I'm sure we can get there. That's the future that I can imagine. And then we build solutions for the people that make our lives better because that's much more satisfying than building things that make life worse or make very few people very rich.

That's not what our humanistic foundation is. So begin with creating this narrative in your head. What's the what's what's the positive narrative for a future that I want? And second, what can I do to get to this future? And if it's a cool idea, please come to us. Thank you very much. I think you spoke for them. I found it very interesting. We said who did government funding who tried to accept government funding. All the hands were had a little bit of group therapy session. So, so everyone in the quiet. Yeah, definitely. Yeah. Yeah. talk to my therapeutic about it. Um I I want to ask you one thing. I think what is really I think also important in your job.

I heard what really helped me in life is that happiness is the right management of expectations. Yeah. So how do you manage expectation especially in something like where perhaps the expectations are very very high in in in doing something revolutionary and and and things like that. How is there a right management of expectation for the things that you or the companies develop in Europe? I think we simply have to be honest with what we do. So I say a lot I don't know. So I was in a in a in a in an audition how is it called in zitsong bundark. So when are we going to have our first fusion power plant? Mhm. And my answer was, you know, if you would have asked the experts, the AI experts, most of them, not all of them, I know there's one or two that would have said differently, when is when when do we have really useful AI tools for everyone in the world?

Yeah. You would have gotten very different answers. Yeah. Right. Between say two years and never, right? That is what the experts would have said, right? So now you're asking an expert on fusion. Mhm. And I can tell you this, it's not going to be two weeks. So I said 15 years in my presentation plus minus 5 years because that sounds kind of right but if it's 25 years or 10 so what you know the thing is free energy in abundance too cheap to meter everywhere available fully you know green will solve a lot of mankind's problems. We will have you know climate change will be much more controllable. We have clean water. We even have food and many other things.

we will defossilize in no time if we have that. So the impact of this is big enough and this is what we're making high hopes for and if we get there 10 years later so what yeah yeah I get that but don't you think that all this predicting I get you you always also get the the question how will AI be in 10 years and in five years and and stuff like that. My response is I'm a general manager not an astrologer. Yeah, this is this is my answer to that. How do you manage this kind of prediction economy? Well, I'm saying like what I said at the intro, you know, if if we don't try it, it will never come.

Yeah. If if if it would be sort of putable in a spreadsheet where a number would come out and it would say 42, right? Then then well, yeah, it probably wouldn't need me, right? Because then anybody can finance this. There are lots of uncertainties in these things, right? And we have to live with those. We we have what we call a serendipity culture. We just allow for these happy accidents, so to speak. And a lot of us are only here because of happy accidents. You know that, right? Okay. Well, it was not a happy accident that he was on stage. Thank you so much, Rafael. Pleasure. Thank you, my friend. Thank you.

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.