Shaping the EU Cloud and AI Development Act | Paradigms

Description

How can we create responsible regulation when the AI race has just begun and there's still little evidence of its benefits for the European market and society? As the European Commission seeks public input on its upcoming Cloud & AI Development Act and its potential impacts, it aims to build more European data centers to match the economic power of the US and China in AI & cloud computing. But is it feasible and what do we do about the missing data?

In this episode, Zuzana and Max for the first time welcome Leitmotiv's co-founder Christiaan van Veen who is a human rights lawyer with extensive experience at the UN, EU, and national level (NL). Together, Max and Christiaan share details about the 40-page analysis and policy recommendations they submitted to the Commission's call for evidence on the Act's possible impact, in collaboration with cultural anthropologist Corinne Cath.

Through their rigorous analysis, solid conceptual framework, and six regulatory interventions, the authors point out how EU policy remains fixated on expanding cloud as a centralized “shared” or “infinity” factory – creating dependence on a few owners, weak competition, questionable social benefits, and guaranteed environmental costs.

And their ultimate proposal for the EU Commission? Fix market failures first, then invest where public value is clear.

(Correction (5:29): Christiaan mentions Article 9, but the correct version is Article 19.)

Paradigms is a podcast produced by Leitmotiv – an independent think tank focused on operationalizing values in our digital future.

Spotify: https://open.spotify.com/episode/6JlonWocmeNx803GCJNWs8?si=d1b97c75000d4fb7
Apple Podcasts: https://podcasts.apple.com/us/podcast/shaping-the-eu-cloud-and-ai-development-act/id1842815462?i=1000732605907
YouTube: https://www.youtube.com/watch?v=HrojS7WBLF8

Hosted on Ausha. See ausha.co/privacy-policy for more information.

Transcription

Speaker #0
This is the next chapter.
Speaker #1
Welcome to Paradigms, a podcast by Leitmotiv, the independent digital policy think tank that exists to operationalize values in our digital future.
Speaker #2
In each episode, we unpack the latest developments and debates in our digital world. Together, we try to make sense. of what is happening to our digital future, who is shaping it, and what we can do to change its path.
Speaker #1
My name is Susanna Ludwikova and I'm here to ask the difficult questions.
Speaker #2
And my name is Max Schulze and I'm here to give simple answers. At Leibmotiv, I think about the policies that shape our digital future and how to operationalize our societal values within them.
Speaker #1
And my job is to bring our thinking to your ears, eyes and minds. In our podcast, we invite you to think with us as we unpack the building blocks of our digital world and work towards a better future that amplifies the best of our core values and our humanity.
Speaker #3
Today, Max and I have a guest in our virtual studio, who is Christian. Can you guys introduce each other since you are both colleagues and friends?
Speaker #2
Of course. I will go first. So Christian is joining us and I think it's really nice to have him here today because he's also joining us as a co-founder for Leitmotiv, our digital policy think tank. And I'm really happy to have him here, especially because he brings a background in law, but also a background in human rights to the conversation and has been working on data centers for a year now and knows more about them than I do apparently now. So Christian, maybe you want to add something that I missed.
Speaker #4
Thanks so much, Max. Well, I know about law and about human rights and about human rights law as well. So that may be a small addition. And also to say that I don't know more about data centers than you, obviously, since you've worked on this for what, the last 15 years of your career on digital infrastructure in general in countries around the globe. and also as a serial entrepreneur, of course, book and business and in... in civil society and so I'm very much looking forward to this conversation about one of our joint projects.
Speaker #2
Me too.
Speaker #3
Then Christian, welcome to Paradigms. It's really nice to have you as our first guest.
Speaker #4
Thank you.
Speaker #3
And today we are here to talk about your most recent publication as co-authors which is a 40-page big analysis and policy recommendations that you prepared as a response to European Commission's call for public input. before they develop the upcoming AI and Cloud Development Act. And in this response, you are both critical, claiming that Europe does not like computational infrastructure, challenging previous institutional documents. And you're also analytical as you outline six regulatory and policy interventions on the development of this Cloud and AI Development Act and the markets in Europe. So before we dive into the whole document, Let's first talk about how it came to be. And so during the process of the whole writing up the analysis, how did the collaboration between you two go? Who brought which strengths to the table?
Speaker #2
I think I can maybe start with a little bit of the process. So with our previous or other organization, the SDA, we've often been involved in these European policy processes. And there's, of course, a more informal... involvement talking to the policymakers directly also attending their events there's a lot of research being done on this act right now in the background that i i think will also come out soon and then there's the more formal response and our goal i think was to first of all not just challenge the assumptions but i think to also think it through what it actually could look like if Europe wants to develop on that front. Yeah, I think. That was the main overarching goal. And I think the strength that we were lacking in the past when we did this was always that we cannot really translate it into concrete policy. And we're really lacking sometimes the ability to frame it in this global context and really account for all the externalities from environmental concerns, but also from societal and human rights concerns. And maybe Christian will comment on. what strength you brought to this, but I hope I presented it on a platter.
Speaker #4
Well, I think it was a good combination of experiences here. So Max has obviously been involved in these discussions at the European level for a much longer time than I have. But I think the combined experience where I focused in my career on environmental and climatic impacts, plus on human rights, brought new dimensions to the table, which we integrated into the submission. I think we should also add that Max and I... received contributions from Corinne Kath at Article 9, which were also very helpful for us to think this through in a more comprehensive way. The other thing I think I should add is, because you said at first, Susanna, that the submission was critical and analytical, I would add to that that it was also constructive. And I think that we very deliberately focused, as Max also said just now, not just on what's wrong, according to us in the current proposal, but also how it could be fixed and what alternative options the European Commission, the EU legislator has. And I think that's very important that we do more than just criticize and analyze, but we also constructively propose alternatives. And that's very much what was on our mind when we were writing this together.
Speaker #3
And was it a very smooth... smooth
Speaker #2
process or how long did it take you what were the hurdles i'll let you want to go first max um i it's very it is very difficult um to write such a comprehensive analysis um i think what makes it even more difficult and it's something you you hear me talk about a lot is the lack of really transparent data on how much infrastructure do we have already how much data centers there are, how well are they utilized, how much energy do they actually use, not just estimations. And I think if that data would be there, this analysis would be very easy. But given that data is absent, I think what we spend most of the time on is proving something to the commission. It's very hard to show, especially with our very strong value set on fact-based recommendations. And I think that made it... more time-consuming than it could have been. Because I'm also curious to hear your take on it.
Speaker #4
I think that's a very relevant dimension. I think what also struck me is that, you know, if you look at the title of the proposed act, it seems very technical and it seems like a very sort of separate domain. But once you dive into this, we see that this is linked to various natural resources inputs. It's about downstream digital economy and how that functions. It's about competition on these markets. It's about their climate impact. It's about other environmental impacts. It's about downstream human rights impacts. So it's actually quite a multidimensional field. And so I think part of the process here was also for us to come together and sort of find a language in which we could capture all of that. which, as I just said, spans a number of fields of expertise. So it was quite a challenge, but I think it was worth going through that exercise to come up with a submission that was really in-depth and strong on multiple levels and 40 pages long, which might be a reflection of that process.
Speaker #2
That's the short version, right? Yeah,
Speaker #4
exactly. That's the exact summary.
Speaker #2
I also don't envy the commission having to weave this all together. I think it's a really difficult task. And I want to give a shout out also to Corinne because we spent a lot of time trying to build a framework or a model on how you can think about all of this as one big thing. And we call it the shared factory and the agile production machine. And we hope that, you know, even if you don't agree with our assumptions or you don't believe in our recommendations, I think. the model that especially came a lot out of Corinne's head really helps to think this through. And that's all we are asking, really. I think that's our fundamental ask. It's like, before we put billions of taxpayer money into developing gigafactories and all this kind of thing, please just think it through. Let's have a democratic process and be clear on what are the inputs we need and what are the outputs we are trying to accomplish with those inputs. And that's, I think, at the core is our message.
Speaker #4
I'm going to put on the head of interviewer now for a moment, Susanna, if you permit me. And I think it's helpful, Max, to explain what you, what we mean when we say shared factory. And actually, this is not just a sort of a performative action where I ask you a question so you can talk about it. But I think we actually spend a lot of time thinking through what is this exactly? And I think it's very important to take listeners through that process because it's not that straightforward.
Speaker #2
Yeah. I hope I can do it justice to explain it quickly. I think what we try to do is to say, look, cloud as a concept is a factory. And in order to be efficient, we basically say like instead of, you know, 4,000 companies using each building their own factory, of course, it makes sense to build one shared factory where we can all use it together. That is very efficient. and In that factory, we have equipment to manufacture things, to manufacture digital products and services. And those machines, you used to bring them yourself. And now you can also share those. And those are the cloud services that a lot of digital products and services are built upon. So not only is now the factory shared, but also the machinery is shared. But all of that is owned by a handful of companies. Yes, you gain an efficiency improvement, but you're also massively dependent on a single shared factory owner who happens to own both the building, the land, the resources, the machinery. And it's all integrated, which, of course, makes it really easy and desirable to produce digital products and services very quickly. Because I don't have to pay for the machinery. I don't have to pay for the factory. It's all there. And I can just focus on. you know, making AI services, building a startup. And I can do that really, really fast. But we don't ask the question at what cost. But in the pure pursuit of efficiency, it feels right, but it leads to massive dependence. And I think a good comparison is, of course, we have done similar things in the physical world with outsourcing to shared factory platforms abroad. And we are now slowly, and during Corona, we saw it even more. see the consequence of that. And I'm not opposed to globalization, but the beauty of the shared factory in cloud is that it's not per se a physical thing. So there we have a lot more opportunity to make it more distributed, to create lots of shared factories, to create European ones and things like this. So I feel like we have a lot more power in the cloud space than we have maybe in the physical production space. I hope I did it justice, Christian. I'm open to criticism.
Speaker #4
No, I think you did it justice. I think you highlighted the fact that there are on paper a lot of efficiencies in doing it this way. But what we also focus on, of course, in the submission is a lot of the inefficiencies here, among other things that, you know, prices are higher than you would expect in a very efficient market. That is very difficult to switch other than what you would expect in a very well-functioning market. And of course, going beyond market economics, there are a lot of inefficiencies in terms of dumping the effects on the rest of society and on our collective environment. And those were some of the what we called externalities that should be should be taken into account also when making decisions in this in this area. and I think that also goes back to what you just said about lack of information in order to make good policy in this area we need to have a good overall picture of just how it functions at present um this market or these markets what its inefficiencies are exactly what its externalities are exactly and then to have a democratic debate about that and make uh and make decisions and a lot of that is uh is currently missing and uh and so i think we are not pretending to have all the answers, but we are very clear on the fact that the... The current debate cannot be a rational one and an evidence-based one because we lack so much information. And there's so many aspects to this which haven't really been explored either in the original proposal by the commission or in the debate that surrounds it so far.
Speaker #2
To add to that, I would say especially the concept of the infinity factory, which is what we currently are pursuing. is not sustainable, not just in an environmental sense, but also an economic sense. You mentioned it's more expensive. It's more expensive because it's a factory where I can walk in and say I want to manufacture something and there's always a machine available to me. It is infinite production capacity for digital products or digital services that can't continue. And now the taxpayer is asked to fund the continuation of an infinity factory because we need even more. But we've never asked the question, for what and how much for what? And do we actually want to produce these products? But we are just expanding the factory to just make more and more and more. And the only person that really benefits is whoever owns the factory.
Speaker #4
Yeah. I mean, this is something that struck me as a relative outsider into these specific debates on digital infrastructure. I've worked on AI policy and human rights for a longer time, but not so much on this particular aspect of these markets. One thing that struck me is that, and it's in the name of the act as well, it's very much focused on the factories and on expanding the number of factories. There's remarkably little discussion about what's actually being made in these factories. So it's as if everyone is talking about making more factories without actually talking about the products and the services that will ultimately be produced. And it strikes me that that part of the debate is vague, to say the least. So what are we doing here exactly? What kind of problems are being resolved? What kind of products are being made? We need to ask critical questions there because all of these investments and all of these impacts, including environmental impacts, of course, need to be worth it if you're going to go in that direction. But are they? And it's fun to read all these brochures by AI companies on all the world's problems that will be solved in the future using AI, among other things. but I think Max you recently wrote to me about how many question marks there are in those documents. Like, wouldn't it be great if AI would solve climate change? Question mark. Question mark.
Speaker #2
They just want to work your cancer. Also, the maybe is always great.
Speaker #4
Yeah.
Speaker #2
Yeah, I think you're worrying too much. I really think you're really too concerned because as long as we keep on digitalizing, productivity growth will follow. Everything will get better. and just use the shared factory, build more digital services, build, build, build. That's the mantra. And venture capital, taxpayers and companies will pay for it anyways. And I'm being sarcastic. I hope the listeners know it's a gigantic extraction machine. The factory extracts money from government by government digitalization, from venture capital, which is partly funded by pension funds and by companies who are aiming to become more productive. and to be more efficient through digitalization. And everybody pays for the shared factory without being guaranteed any outcome.
Speaker #3
On this one, I also have another question that the EU Commission page also proposes that cloud investment will benefit startups and SMEs, as you've also hinted. But you warn against startup credit schemes. So why this pushback? Do you see a tension between supporting emerging firms and preventing... piggybacking on these hyperscalers?
Speaker #2
I mean, startups should exist, right? But I keep on posing the question, which is not mentioned in any of the policy documents or in any communication is, if I'm a startup, how do, first of all, how do I access the gigafactory? Right? First question, how do I get the compute that I need to build my startup from that? Second question, is it going to be cheaper than what I currently can get on the market? Because why would I go for that if it's not cheaper? Simple economics. And the third question is, if Google is giving me 300,000 euros worth of computation in their factory, am I going to get the equivalent from a government gigafactory? And if that's not solved, simple economic theory dictates I would choose the cheapest one. And if the cheapest one is free, I will go to the free one. So the policy intervention should be much more focused on... solving that equation, either making sure that these credits that are handed out can be spent anywhere, so making that more fluid, or not having the credits at all and having fair competition, or, for example, creating a European equivalent, which I think might end up being very expensive and raises the question for what are we spending this money again, what Christian raised before. But without solving that equation,
Speaker #4
why are you are you providing computes do you have anything to add well just slightly off topic but I think it's related as well What strikes me in this debate is that if you're being skeptical of initialists like this, I think you're easily put in the category of unrealist, non-realist. So it's simply good economics to go in this direction. So we should follow the path that we're currently already on and expand capacity. And so if you question that, you're basically being put in the category of the skeptic that really doesn't understand economics. But my counter question would be, is this really good economics? So what is actually the total productivity gain and the aggregate wealth that we're gaining from going into this direction? And then we try to look for evidence. so economic evidence of the benefits. There's actually a scarcity of that. And so I would like to put it on its head and say, how realistic is it to expect economic growth, if that's your goal, from heading in this direction, as the commission proposes here, if there's not much evidence that there will be? We cite a paper by Darren Acemoglu, who won the Nobel... Price and Economics last year or the year before that, who did the math and ends up saying that for the US in a 10-year period, the economic growth related to AI would be something in the order of one or one and a half percent. And that is not a lot in light of the massive investments made here and all the promises made on the benefit side, one percent. one and a half percent, if that's what you're aiming for, then perhaps we should tone down on all the expectations.
Speaker #2
And I think if you want economic growth, just build on that. I'm not advocating for that, but fundamentally you need cheap compute. AI economics are fundamentally driven by the cost of compute. Most companies today are not profitable because the compute is too expensive. And we know, Capitalist's playbook number one, competition. creates lower prices. Since when does the state flooding the market with cheap compute or, let's see if it's cheap at all, induce competition? It does not. It actually messes with the competitive aspects. And in an analogy, like with wind energy, we created the supply of wind energy as a consensus-based decision, especially in Germany, for society and society paid? for flooding the market with wind energy. We wanted that and we did that. And the outcome, though, was clear. We had more renewable energy, therefore we have less fossil fuels and we could shut down some of the coal-fired power plants. Very clear investment outcome. And wind energy wouldn't have worked if the state wouldn't have interfered with the economics of it. But in this case, I think we don't even have competition yet because it's such an oligopolistic market structure. So why interfere in a market when... it's already failing at competition. Shouldn't you then first focus on introducing competition on the pure compute trading layer? And I think we propose a lot of ideas on how to induce marketplaces and competition. But I think the commission is like three steps ahead of solving a problem that is a pure, maybe pure competitive problem for now.
Speaker #4
I mean, it goes even a bit further than that, I would say. And it's that they don't really want to interfere in the... competitive functioning. So it's actually, they want to stay away from that. So they're not actively interested in interfering there. So competition law is sort of brushed aside. Then economic benefits are unclear. Societal benefits, question mark, and environmental impacts are guaranteed. But yet somehow those environmental... those guaranteed environmental impacts, however big they are exactly, because that we don't know either, should be weighed against some kind of benefit. But again, question mark. So the whole totality of this is almost like a textbook case of something that's not been thought through completely. And that's not to sound arrogant, but I think there's such a rush and such a push to be ahead of this among policymakers, not just in the EU, but that happens globally. that asking these very simple critical questions are almost overlooked. And it's easy to ask questions, and then we need to have answers to those questions. And I think we're trying to do both in the submission, to both ask the simple questions. What is the benefit here exactly? But then also to try to think it through. Again, that's not something we can relay completely in a podcast, but anyone interested should actually spend some time and read what we wrote.
Speaker #0
We will find out the answer to those questions.
Speaker #3
Yes, I think asking these critical questions is very important for the crucial bodies to trip in this AI race that's happening currently. And you think then this narrative is driven by industry lobbying or just bad economics? You might have already talked about this, but a simple answer.
Speaker #4
One thing on that. So next, I'll let you answer. You're better placed to answer. But one thing to add is that I think for a lot of people, whether they're policymakers, whether they're in civil society or elsewhere, whether they're journalists, it's very hard to admit that you might not fully understand where this is going or where it's presently at, where everyone involved gives you the impression that they know. And so I think collectively what's happening here is that a lot of probably pretty intelligent actors in this debate are afraid to say that they also don't understand. But they don't want to be the one who asks the dumb question in class. I think that's the underlying sort of mass psychology that's going on here as well.
Speaker #2
And if anybody listening to this is like, I want to help. have some help ask dumb questions we are more than willing to come and ask stupid questions everywhere that's part of what we do um i think on the lobbying versus you know bad economics question in in their defense i think the entire ai mega sector is very open that the economics don't work right so nvidia just announced that they're going to pay for open ai's infrastructure oracle got the big contract for cloud computing with OpenAI as well, and 10 minutes later invested in OpenAI. So they're paying their own cloud bill. And I think they're very, and Sam Altman also openly says that at the current price, this is not sustainable. It should be costing 20,000 euros a month, not 20 euros a month. And whenever you have a business sector or market where the economics don't work, what do you do? You go to the state and you ask them to fix it, right? And I think that's literally what's happening. Whenever the NVIDIA CEO, OpenAI CEO, and Tropic CEO walks out of a commission building, 10 minutes later, I feel like we get an announcement that the EU is going to invest billions. And I think that's not the job here. It is not the job of the state to fix market economics. If something doesn't work, let it fail. Because that's what we created capitalism for, to make sure... Good ideas survive in a market and bad ideas fail. And right now, I think with AI, we're not saying it's a bad idea, just saying as it currently stands, it's maybe better placed in universities for more research and not ready for the market. Because apparently in the market, it doesn't work.
Speaker #4
There's also a bit of an irony here, right? Where the big winners in capitalism come to governments. to help them out while at the same time professing their government should stay out of their business as much as possible because it will ruin the economics of it.
Speaker #2
Please build more data centers and more energy infrastructure but please do not look closely at how we're making money.
Speaker #3
Very nice retention based in a relationship to have to navigate the state and tech. And then back to the content of your work, your paper suggests key interventions. like European Digital Resource Exchange or migration loans and tighter takeover rules. Which of these do you see as the hook that could make it to the final regulation or which would benefit the most?
Speaker #2
I think there are the things that are more the macro topics that if they would make it, they would have to be the whole policy. For example, creating a marketplace to trade compute would require complete rewiring of the market and really... solving a lot of competition law problems and making sure this is compatible with the World Trade Organization. It's the holy grail, I would say.
Speaker #0
But then we also suggested some very simple, well, simple, but more tangible interventions that could be part of any other policy as well. Migrations and other things like this. But maybe, Christian, you want to outline a few of those.
Speaker #1
Well, I think, first of all, at a bit more of an abstract level, I think what's very important to highlight is that if you perceive as... just any market as functioning, that market will give signals that indicate where other participants should be heading. This is a market where a lot of those signals are hidden. So in terms of price, what is the actual price of a cloud offering, for instance? What kind of cross subsidization has happened within vertically integrated firms, for instance? A lot of the aspects that you would expect in a well-functioning market. are missing here. And that also robs regulators, governments, and so at the end of the day, the public, democracy, robs it of the chance to steer it in a certain direction. So we don't have good signals. Therefore, we don't really know what's going on in these markets. And because we really don't know what's going on in these markets, we really can't regulate on outcomes and effect. So I think One of the things that the submission focuses on is how you can let that market function as a market shoot in order for the EU authorities or member states, and that same would apply for outside of the EU, to intervene where they see that as a fit option. So I think that's an important thing to say about our proposals. I think, as Max said... We presented a whole, and so ideally we would like to see the whole implemented, but you can still take parts of it and make it better. And I think Max already talked about startup credits. There are other forms of non-competitive behavior that I think could be addressed in the policy proposals that we have. So in other words, to make the market a little bit fairer. And even if some of those would be implemented, it would be a better situation than we're in currently.
Speaker #2
It sounds very convenient to make these interventions applicable not only here, but in other areas as well. But once the consultation process closes and the act proceeds, what is LightMontiv's next step? Is it advocacy, monitoring or coalition building? What do you expect will happen?
Speaker #1
Well, first, it's not over yet. And I think the latest information is that the process has been postponed slightly. So we're looking towards finalization by the first quarter of 2026, if I'm not mistaken. So there's still some time to go. And in a rapidly changing world in general and lightning speed developments in AI more specifically. I don't think we can truly predict what's going to happen in the next few months. So we'll be continuing our legislative advocacy until the time that the act is finally adopted, if it gets to that point. We haven't internally discussed what to do next because the outcome is still unclear. But even with adopted legislation, in my experience, there's a lot of work to be done on implementation and also, of course, on potential future changes and amendments to that legislation. So it's never over.
Speaker #0
I think there's nothing to add to that. I think advocacy is in our blood. So I think it's very difficult to not do it. So, so. I think we are continuously going to be involved in this process. I think in this case, it's very difficult to form a coalition in that sense, because the business interests here are very stark. Like you have the incumbents, so to say, the US cloud providers, US AI companies, and even some European actors in that. You have the whole Eurostack movement, which is about like, give us so much money that we can basically rebuild everything European. And I think we find ourselves in the middle, being the most pragmatic, the most practical, let's restore competition on compute actors. I think it's at the moment, at least it's better in my perspective, that we stay alone and independent and move our agenda forward than to try to build too many groups of interests where we have to balance each other's agenda.
Speaker #1
It's helpful, though, to add to that, to see that the landscape in Brussels is slightly changing, it seems. There are more civil society organizations interested in... debates around data centers because of environmental impacts primarily. So it's good to see that a number of civil society organizations working in that area have also made submissions to this process from Beyond Fossil Fuels and Green Screen Coalition, Client Earth and others. I think that helps put extra spotlight on the environmental dimension. So in the A bigger scheme of things, I think, that will help sort of balance the debate, which has been very unbalanced until this point in time. And I think ultimately, without us necessarily having to form that coalition, it's clear that interests align among a lot of different actors, whether they're corporations or NGOs, that all at least want things to be very different from what's in the current commission proposal. I think among civil society, among certain, especially smaller businesses in Europe, parts of the general public, there's an unease with the direction this is heading in. And I think in that sense, there's an alignment without those interests being fully similar and in coordination.
Speaker #2
Sounds promising. And let's see what happens. And let's hope for the best for our digital future. Thank you very much for joining me today for the conversation. Thank you for contributing with this important piece of work. And yeah, have a nice day.
Speaker #1
Thanks so much for the invitation.

About Paradigms

Unpacking the Leitmotiv of our digital future – where we are now, where we're going, and who is setting the direction? Each week, Zuzana and Max discuss the latest developments and public debates in the digital sphere to figure out our future and ask: Is this where we, as a society, want to go?

This podcast is part of our mission as an independent think-tank focused on operationalizing values in our digital future. Together with you, we want to imagine a Leitmotiv (’vision’) for a digital world that is grounded in societal values and mobilize our power to shape it. Follow us on LinkedIn, Bluesky or Instagram to stay on the pulse of digital policy or the latest tech talk and learn everything about how we’re building a values-centered digital future.

Apple Podcasts: https://ltmv.io/podcast-apple

Spotify: https://ltmv.io/podcast-spotify

YouTube: https://ltmv.io/podcast-yt

Hosted on Ausha. See ausha.co/privacy-policy for more information.

About Paradigms

Apple Podcasts: https://ltmv.io/podcast-apple

Spotify: https://ltmv.io/podcast-spotify

YouTube: https://ltmv.io/podcast-yt

Hosted on Ausha. See ausha.co/privacy-policy for more information.

Description

And their ultimate proposal for the EU Commission? Fix market failures first, then invest where public value is clear.

(Correction (5:29): Christiaan mentions Article 9, but the correct version is Article 19.)

Paradigms is a podcast produced by Leitmotiv – an independent think tank focused on operationalizing values in our digital future.

Hosted on Ausha. See ausha.co/privacy-policy for more information.

Transcription

Speaker #0
This is the next chapter.
Speaker #1
Welcome to Paradigms, a podcast by Leitmotiv, the independent digital policy think tank that exists to operationalize values in our digital future.
Speaker #2
In each episode, we unpack the latest developments and debates in our digital world. Together, we try to make sense. of what is happening to our digital future, who is shaping it, and what we can do to change its path.
Speaker #1
My name is Susanna Ludwikova and I'm here to ask the difficult questions.
Speaker #2
And my name is Max Schulze and I'm here to give simple answers. At Leibmotiv, I think about the policies that shape our digital future and how to operationalize our societal values within them.
Speaker #1
And my job is to bring our thinking to your ears, eyes and minds. In our podcast, we invite you to think with us as we unpack the building blocks of our digital world and work towards a better future that amplifies the best of our core values and our humanity.
Speaker #3
Today, Max and I have a guest in our virtual studio, who is Christian. Can you guys introduce each other since you are both colleagues and friends?
Speaker #2
Of course. I will go first. So Christian is joining us and I think it's really nice to have him here today because he's also joining us as a co-founder for Leitmotiv, our digital policy think tank. And I'm really happy to have him here, especially because he brings a background in law, but also a background in human rights to the conversation and has been working on data centers for a year now and knows more about them than I do apparently now. So Christian, maybe you want to add something that I missed.
Speaker #4
Thanks so much, Max. Well, I know about law and about human rights and about human rights law as well. So that may be a small addition. And also to say that I don't know more about data centers than you, obviously, since you've worked on this for what, the last 15 years of your career on digital infrastructure in general in countries around the globe. and also as a serial entrepreneur, of course, book and business and in... in civil society and so I'm very much looking forward to this conversation about one of our joint projects.
Speaker #2
Me too.
Speaker #3
Then Christian, welcome to Paradigms. It's really nice to have you as our first guest.
Speaker #4
Thank you.
Speaker #3
And today we are here to talk about your most recent publication as co-authors which is a 40-page big analysis and policy recommendations that you prepared as a response to European Commission's call for public input. before they develop the upcoming AI and Cloud Development Act. And in this response, you are both critical, claiming that Europe does not like computational infrastructure, challenging previous institutional documents. And you're also analytical as you outline six regulatory and policy interventions on the development of this Cloud and AI Development Act and the markets in Europe. So before we dive into the whole document, Let's first talk about how it came to be. And so during the process of the whole writing up the analysis, how did the collaboration between you two go? Who brought which strengths to the table?
Speaker #2
I think I can maybe start with a little bit of the process. So with our previous or other organization, the SDA, we've often been involved in these European policy processes. And there's, of course, a more informal... involvement talking to the policymakers directly also attending their events there's a lot of research being done on this act right now in the background that i i think will also come out soon and then there's the more formal response and our goal i think was to first of all not just challenge the assumptions but i think to also think it through what it actually could look like if Europe wants to develop on that front. Yeah, I think. That was the main overarching goal. And I think the strength that we were lacking in the past when we did this was always that we cannot really translate it into concrete policy. And we're really lacking sometimes the ability to frame it in this global context and really account for all the externalities from environmental concerns, but also from societal and human rights concerns. And maybe Christian will comment on. what strength you brought to this, but I hope I presented it on a platter.
Speaker #4
Well, I think it was a good combination of experiences here. So Max has obviously been involved in these discussions at the European level for a much longer time than I have. But I think the combined experience where I focused in my career on environmental and climatic impacts, plus on human rights, brought new dimensions to the table, which we integrated into the submission. I think we should also add that Max and I... received contributions from Corinne Kath at Article 9, which were also very helpful for us to think this through in a more comprehensive way. The other thing I think I should add is, because you said at first, Susanna, that the submission was critical and analytical, I would add to that that it was also constructive. And I think that we very deliberately focused, as Max also said just now, not just on what's wrong, according to us in the current proposal, but also how it could be fixed and what alternative options the European Commission, the EU legislator has. And I think that's very important that we do more than just criticize and analyze, but we also constructively propose alternatives. And that's very much what was on our mind when we were writing this together.
Speaker #3
And was it a very smooth... smooth
Speaker #2
process or how long did it take you what were the hurdles i'll let you want to go first max um i it's very it is very difficult um to write such a comprehensive analysis um i think what makes it even more difficult and it's something you you hear me talk about a lot is the lack of really transparent data on how much infrastructure do we have already how much data centers there are, how well are they utilized, how much energy do they actually use, not just estimations. And I think if that data would be there, this analysis would be very easy. But given that data is absent, I think what we spend most of the time on is proving something to the commission. It's very hard to show, especially with our very strong value set on fact-based recommendations. And I think that made it... more time-consuming than it could have been. Because I'm also curious to hear your take on it.
Speaker #4
I think that's a very relevant dimension. I think what also struck me is that, you know, if you look at the title of the proposed act, it seems very technical and it seems like a very sort of separate domain. But once you dive into this, we see that this is linked to various natural resources inputs. It's about downstream digital economy and how that functions. It's about competition on these markets. It's about their climate impact. It's about other environmental impacts. It's about downstream human rights impacts. So it's actually quite a multidimensional field. And so I think part of the process here was also for us to come together and sort of find a language in which we could capture all of that. which, as I just said, spans a number of fields of expertise. So it was quite a challenge, but I think it was worth going through that exercise to come up with a submission that was really in-depth and strong on multiple levels and 40 pages long, which might be a reflection of that process.
Speaker #2
That's the short version, right? Yeah,
Speaker #4
exactly. That's the exact summary.
Speaker #2
I also don't envy the commission having to weave this all together. I think it's a really difficult task. And I want to give a shout out also to Corinne because we spent a lot of time trying to build a framework or a model on how you can think about all of this as one big thing. And we call it the shared factory and the agile production machine. And we hope that, you know, even if you don't agree with our assumptions or you don't believe in our recommendations, I think. the model that especially came a lot out of Corinne's head really helps to think this through. And that's all we are asking, really. I think that's our fundamental ask. It's like, before we put billions of taxpayer money into developing gigafactories and all this kind of thing, please just think it through. Let's have a democratic process and be clear on what are the inputs we need and what are the outputs we are trying to accomplish with those inputs. And that's, I think, at the core is our message.
Speaker #4
I'm going to put on the head of interviewer now for a moment, Susanna, if you permit me. And I think it's helpful, Max, to explain what you, what we mean when we say shared factory. And actually, this is not just a sort of a performative action where I ask you a question so you can talk about it. But I think we actually spend a lot of time thinking through what is this exactly? And I think it's very important to take listeners through that process because it's not that straightforward.
Speaker #2
Yeah. I hope I can do it justice to explain it quickly. I think what we try to do is to say, look, cloud as a concept is a factory. And in order to be efficient, we basically say like instead of, you know, 4,000 companies using each building their own factory, of course, it makes sense to build one shared factory where we can all use it together. That is very efficient. and In that factory, we have equipment to manufacture things, to manufacture digital products and services. And those machines, you used to bring them yourself. And now you can also share those. And those are the cloud services that a lot of digital products and services are built upon. So not only is now the factory shared, but also the machinery is shared. But all of that is owned by a handful of companies. Yes, you gain an efficiency improvement, but you're also massively dependent on a single shared factory owner who happens to own both the building, the land, the resources, the machinery. And it's all integrated, which, of course, makes it really easy and desirable to produce digital products and services very quickly. Because I don't have to pay for the machinery. I don't have to pay for the factory. It's all there. And I can just focus on. you know, making AI services, building a startup. And I can do that really, really fast. But we don't ask the question at what cost. But in the pure pursuit of efficiency, it feels right, but it leads to massive dependence. And I think a good comparison is, of course, we have done similar things in the physical world with outsourcing to shared factory platforms abroad. And we are now slowly, and during Corona, we saw it even more. see the consequence of that. And I'm not opposed to globalization, but the beauty of the shared factory in cloud is that it's not per se a physical thing. So there we have a lot more opportunity to make it more distributed, to create lots of shared factories, to create European ones and things like this. So I feel like we have a lot more power in the cloud space than we have maybe in the physical production space. I hope I did it justice, Christian. I'm open to criticism.
Speaker #4
No, I think you did it justice. I think you highlighted the fact that there are on paper a lot of efficiencies in doing it this way. But what we also focus on, of course, in the submission is a lot of the inefficiencies here, among other things that, you know, prices are higher than you would expect in a very efficient market. That is very difficult to switch other than what you would expect in a very well-functioning market. And of course, going beyond market economics, there are a lot of inefficiencies in terms of dumping the effects on the rest of society and on our collective environment. And those were some of the what we called externalities that should be should be taken into account also when making decisions in this in this area. and I think that also goes back to what you just said about lack of information in order to make good policy in this area we need to have a good overall picture of just how it functions at present um this market or these markets what its inefficiencies are exactly what its externalities are exactly and then to have a democratic debate about that and make uh and make decisions and a lot of that is uh is currently missing and uh and so i think we are not pretending to have all the answers, but we are very clear on the fact that the... The current debate cannot be a rational one and an evidence-based one because we lack so much information. And there's so many aspects to this which haven't really been explored either in the original proposal by the commission or in the debate that surrounds it so far.
Speaker #2
To add to that, I would say especially the concept of the infinity factory, which is what we currently are pursuing. is not sustainable, not just in an environmental sense, but also an economic sense. You mentioned it's more expensive. It's more expensive because it's a factory where I can walk in and say I want to manufacture something and there's always a machine available to me. It is infinite production capacity for digital products or digital services that can't continue. And now the taxpayer is asked to fund the continuation of an infinity factory because we need even more. But we've never asked the question, for what and how much for what? And do we actually want to produce these products? But we are just expanding the factory to just make more and more and more. And the only person that really benefits is whoever owns the factory.
Speaker #4
Yeah. I mean, this is something that struck me as a relative outsider into these specific debates on digital infrastructure. I've worked on AI policy and human rights for a longer time, but not so much on this particular aspect of these markets. One thing that struck me is that, and it's in the name of the act as well, it's very much focused on the factories and on expanding the number of factories. There's remarkably little discussion about what's actually being made in these factories. So it's as if everyone is talking about making more factories without actually talking about the products and the services that will ultimately be produced. And it strikes me that that part of the debate is vague, to say the least. So what are we doing here exactly? What kind of problems are being resolved? What kind of products are being made? We need to ask critical questions there because all of these investments and all of these impacts, including environmental impacts, of course, need to be worth it if you're going to go in that direction. But are they? And it's fun to read all these brochures by AI companies on all the world's problems that will be solved in the future using AI, among other things. but I think Max you recently wrote to me about how many question marks there are in those documents. Like, wouldn't it be great if AI would solve climate change? Question mark. Question mark.
Speaker #2
They just want to work your cancer. Also, the maybe is always great.
Speaker #4
Yeah.
Speaker #2
Yeah, I think you're worrying too much. I really think you're really too concerned because as long as we keep on digitalizing, productivity growth will follow. Everything will get better. and just use the shared factory, build more digital services, build, build, build. That's the mantra. And venture capital, taxpayers and companies will pay for it anyways. And I'm being sarcastic. I hope the listeners know it's a gigantic extraction machine. The factory extracts money from government by government digitalization, from venture capital, which is partly funded by pension funds and by companies who are aiming to become more productive. and to be more efficient through digitalization. And everybody pays for the shared factory without being guaranteed any outcome.
Speaker #3
On this one, I also have another question that the EU Commission page also proposes that cloud investment will benefit startups and SMEs, as you've also hinted. But you warn against startup credit schemes. So why this pushback? Do you see a tension between supporting emerging firms and preventing... piggybacking on these hyperscalers?
Speaker #2
I mean, startups should exist, right? But I keep on posing the question, which is not mentioned in any of the policy documents or in any communication is, if I'm a startup, how do, first of all, how do I access the gigafactory? Right? First question, how do I get the compute that I need to build my startup from that? Second question, is it going to be cheaper than what I currently can get on the market? Because why would I go for that if it's not cheaper? Simple economics. And the third question is, if Google is giving me 300,000 euros worth of computation in their factory, am I going to get the equivalent from a government gigafactory? And if that's not solved, simple economic theory dictates I would choose the cheapest one. And if the cheapest one is free, I will go to the free one. So the policy intervention should be much more focused on... solving that equation, either making sure that these credits that are handed out can be spent anywhere, so making that more fluid, or not having the credits at all and having fair competition, or, for example, creating a European equivalent, which I think might end up being very expensive and raises the question for what are we spending this money again, what Christian raised before. But without solving that equation,
Speaker #4
why are you are you providing computes do you have anything to add well just slightly off topic but I think it's related as well What strikes me in this debate is that if you're being skeptical of initialists like this, I think you're easily put in the category of unrealist, non-realist. So it's simply good economics to go in this direction. So we should follow the path that we're currently already on and expand capacity. And so if you question that, you're basically being put in the category of the skeptic that really doesn't understand economics. But my counter question would be, is this really good economics? So what is actually the total productivity gain and the aggregate wealth that we're gaining from going into this direction? And then we try to look for evidence. so economic evidence of the benefits. There's actually a scarcity of that. And so I would like to put it on its head and say, how realistic is it to expect economic growth, if that's your goal, from heading in this direction, as the commission proposes here, if there's not much evidence that there will be? We cite a paper by Darren Acemoglu, who won the Nobel... Price and Economics last year or the year before that, who did the math and ends up saying that for the US in a 10-year period, the economic growth related to AI would be something in the order of one or one and a half percent. And that is not a lot in light of the massive investments made here and all the promises made on the benefit side, one percent. one and a half percent, if that's what you're aiming for, then perhaps we should tone down on all the expectations.
Speaker #2
And I think if you want economic growth, just build on that. I'm not advocating for that, but fundamentally you need cheap compute. AI economics are fundamentally driven by the cost of compute. Most companies today are not profitable because the compute is too expensive. And we know, Capitalist's playbook number one, competition. creates lower prices. Since when does the state flooding the market with cheap compute or, let's see if it's cheap at all, induce competition? It does not. It actually messes with the competitive aspects. And in an analogy, like with wind energy, we created the supply of wind energy as a consensus-based decision, especially in Germany, for society and society paid? for flooding the market with wind energy. We wanted that and we did that. And the outcome, though, was clear. We had more renewable energy, therefore we have less fossil fuels and we could shut down some of the coal-fired power plants. Very clear investment outcome. And wind energy wouldn't have worked if the state wouldn't have interfered with the economics of it. But in this case, I think we don't even have competition yet because it's such an oligopolistic market structure. So why interfere in a market when... it's already failing at competition. Shouldn't you then first focus on introducing competition on the pure compute trading layer? And I think we propose a lot of ideas on how to induce marketplaces and competition. But I think the commission is like three steps ahead of solving a problem that is a pure, maybe pure competitive problem for now.
Speaker #4
I mean, it goes even a bit further than that, I would say. And it's that they don't really want to interfere in the... competitive functioning. So it's actually, they want to stay away from that. So they're not actively interested in interfering there. So competition law is sort of brushed aside. Then economic benefits are unclear. Societal benefits, question mark, and environmental impacts are guaranteed. But yet somehow those environmental... those guaranteed environmental impacts, however big they are exactly, because that we don't know either, should be weighed against some kind of benefit. But again, question mark. So the whole totality of this is almost like a textbook case of something that's not been thought through completely. And that's not to sound arrogant, but I think there's such a rush and such a push to be ahead of this among policymakers, not just in the EU, but that happens globally. that asking these very simple critical questions are almost overlooked. And it's easy to ask questions, and then we need to have answers to those questions. And I think we're trying to do both in the submission, to both ask the simple questions. What is the benefit here exactly? But then also to try to think it through. Again, that's not something we can relay completely in a podcast, but anyone interested should actually spend some time and read what we wrote.
Speaker #0
We will find out the answer to those questions.
Speaker #3
Yes, I think asking these critical questions is very important for the crucial bodies to trip in this AI race that's happening currently. And you think then this narrative is driven by industry lobbying or just bad economics? You might have already talked about this, but a simple answer.
Speaker #4
One thing on that. So next, I'll let you answer. You're better placed to answer. But one thing to add is that I think for a lot of people, whether they're policymakers, whether they're in civil society or elsewhere, whether they're journalists, it's very hard to admit that you might not fully understand where this is going or where it's presently at, where everyone involved gives you the impression that they know. And so I think collectively what's happening here is that a lot of probably pretty intelligent actors in this debate are afraid to say that they also don't understand. But they don't want to be the one who asks the dumb question in class. I think that's the underlying sort of mass psychology that's going on here as well.
Speaker #2
And if anybody listening to this is like, I want to help. have some help ask dumb questions we are more than willing to come and ask stupid questions everywhere that's part of what we do um i think on the lobbying versus you know bad economics question in in their defense i think the entire ai mega sector is very open that the economics don't work right so nvidia just announced that they're going to pay for open ai's infrastructure oracle got the big contract for cloud computing with OpenAI as well, and 10 minutes later invested in OpenAI. So they're paying their own cloud bill. And I think they're very, and Sam Altman also openly says that at the current price, this is not sustainable. It should be costing 20,000 euros a month, not 20 euros a month. And whenever you have a business sector or market where the economics don't work, what do you do? You go to the state and you ask them to fix it, right? And I think that's literally what's happening. Whenever the NVIDIA CEO, OpenAI CEO, and Tropic CEO walks out of a commission building, 10 minutes later, I feel like we get an announcement that the EU is going to invest billions. And I think that's not the job here. It is not the job of the state to fix market economics. If something doesn't work, let it fail. Because that's what we created capitalism for, to make sure... Good ideas survive in a market and bad ideas fail. And right now, I think with AI, we're not saying it's a bad idea, just saying as it currently stands, it's maybe better placed in universities for more research and not ready for the market. Because apparently in the market, it doesn't work.
Speaker #4
There's also a bit of an irony here, right? Where the big winners in capitalism come to governments. to help them out while at the same time professing their government should stay out of their business as much as possible because it will ruin the economics of it.
Speaker #2
Please build more data centers and more energy infrastructure but please do not look closely at how we're making money.
Speaker #3
Very nice retention based in a relationship to have to navigate the state and tech. And then back to the content of your work, your paper suggests key interventions. like European Digital Resource Exchange or migration loans and tighter takeover rules. Which of these do you see as the hook that could make it to the final regulation or which would benefit the most?
Speaker #2
I think there are the things that are more the macro topics that if they would make it, they would have to be the whole policy. For example, creating a marketplace to trade compute would require complete rewiring of the market and really... solving a lot of competition law problems and making sure this is compatible with the World Trade Organization. It's the holy grail, I would say.
Speaker #0
But then we also suggested some very simple, well, simple, but more tangible interventions that could be part of any other policy as well. Migrations and other things like this. But maybe, Christian, you want to outline a few of those.
Speaker #1
Well, I think, first of all, at a bit more of an abstract level, I think what's very important to highlight is that if you perceive as... just any market as functioning, that market will give signals that indicate where other participants should be heading. This is a market where a lot of those signals are hidden. So in terms of price, what is the actual price of a cloud offering, for instance? What kind of cross subsidization has happened within vertically integrated firms, for instance? A lot of the aspects that you would expect in a well-functioning market. are missing here. And that also robs regulators, governments, and so at the end of the day, the public, democracy, robs it of the chance to steer it in a certain direction. So we don't have good signals. Therefore, we don't really know what's going on in these markets. And because we really don't know what's going on in these markets, we really can't regulate on outcomes and effect. So I think One of the things that the submission focuses on is how you can let that market function as a market shoot in order for the EU authorities or member states, and that same would apply for outside of the EU, to intervene where they see that as a fit option. So I think that's an important thing to say about our proposals. I think, as Max said... We presented a whole, and so ideally we would like to see the whole implemented, but you can still take parts of it and make it better. And I think Max already talked about startup credits. There are other forms of non-competitive behavior that I think could be addressed in the policy proposals that we have. So in other words, to make the market a little bit fairer. And even if some of those would be implemented, it would be a better situation than we're in currently.
Speaker #2
It sounds very convenient to make these interventions applicable not only here, but in other areas as well. But once the consultation process closes and the act proceeds, what is LightMontiv's next step? Is it advocacy, monitoring or coalition building? What do you expect will happen?
Speaker #1
Well, first, it's not over yet. And I think the latest information is that the process has been postponed slightly. So we're looking towards finalization by the first quarter of 2026, if I'm not mistaken. So there's still some time to go. And in a rapidly changing world in general and lightning speed developments in AI more specifically. I don't think we can truly predict what's going to happen in the next few months. So we'll be continuing our legislative advocacy until the time that the act is finally adopted, if it gets to that point. We haven't internally discussed what to do next because the outcome is still unclear. But even with adopted legislation, in my experience, there's a lot of work to be done on implementation and also, of course, on potential future changes and amendments to that legislation. So it's never over.
Speaker #0
I think there's nothing to add to that. I think advocacy is in our blood. So I think it's very difficult to not do it. So, so. I think we are continuously going to be involved in this process. I think in this case, it's very difficult to form a coalition in that sense, because the business interests here are very stark. Like you have the incumbents, so to say, the US cloud providers, US AI companies, and even some European actors in that. You have the whole Eurostack movement, which is about like, give us so much money that we can basically rebuild everything European. And I think we find ourselves in the middle, being the most pragmatic, the most practical, let's restore competition on compute actors. I think it's at the moment, at least it's better in my perspective, that we stay alone and independent and move our agenda forward than to try to build too many groups of interests where we have to balance each other's agenda.
Speaker #1
It's helpful, though, to add to that, to see that the landscape in Brussels is slightly changing, it seems. There are more civil society organizations interested in... debates around data centers because of environmental impacts primarily. So it's good to see that a number of civil society organizations working in that area have also made submissions to this process from Beyond Fossil Fuels and Green Screen Coalition, Client Earth and others. I think that helps put extra spotlight on the environmental dimension. So in the A bigger scheme of things, I think, that will help sort of balance the debate, which has been very unbalanced until this point in time. And I think ultimately, without us necessarily having to form that coalition, it's clear that interests align among a lot of different actors, whether they're corporations or NGOs, that all at least want things to be very different from what's in the current commission proposal. I think among civil society, among certain, especially smaller businesses in Europe, parts of the general public, there's an unease with the direction this is heading in. And I think in that sense, there's an alignment without those interests being fully similar and in coordination.
Speaker #2
Sounds promising. And let's see what happens. And let's hope for the best for our digital future. Thank you very much for joining me today for the conversation. Thank you for contributing with this important piece of work. And yeah, have a nice day.
Speaker #1
Thanks so much for the invitation.

About Paradigms

Apple Podcasts: https://ltmv.io/podcast-apple

Spotify: https://ltmv.io/podcast-spotify

YouTube: https://ltmv.io/podcast-yt

Hosted on Ausha. See ausha.co/privacy-policy for more information.

About Paradigms

Apple Podcasts: https://ltmv.io/podcast-apple

Spotify: https://ltmv.io/podcast-spotify

YouTube: https://ltmv.io/podcast-yt

Hosted on Ausha. See ausha.co/privacy-policy for more information.

Embed

You may also like

Description

And their ultimate proposal for the EU Commission? Fix market failures first, then invest where public value is clear.

(Correction (5:29): Christiaan mentions Article 9, but the correct version is Article 19.)

Paradigms is a podcast produced by Leitmotiv – an independent think tank focused on operationalizing values in our digital future.

Hosted on Ausha. See ausha.co/privacy-policy for more information.

Transcription

Speaker #0
This is the next chapter.
Speaker #1
Welcome to Paradigms, a podcast by Leitmotiv, the independent digital policy think tank that exists to operationalize values in our digital future.
Speaker #2
In each episode, we unpack the latest developments and debates in our digital world. Together, we try to make sense. of what is happening to our digital future, who is shaping it, and what we can do to change its path.
Speaker #1
My name is Susanna Ludwikova and I'm here to ask the difficult questions.
Speaker #2
And my name is Max Schulze and I'm here to give simple answers. At Leibmotiv, I think about the policies that shape our digital future and how to operationalize our societal values within them.
Speaker #1
And my job is to bring our thinking to your ears, eyes and minds. In our podcast, we invite you to think with us as we unpack the building blocks of our digital world and work towards a better future that amplifies the best of our core values and our humanity.
Speaker #3
Today, Max and I have a guest in our virtual studio, who is Christian. Can you guys introduce each other since you are both colleagues and friends?
Speaker #2
Of course. I will go first. So Christian is joining us and I think it's really nice to have him here today because he's also joining us as a co-founder for Leitmotiv, our digital policy think tank. And I'm really happy to have him here, especially because he brings a background in law, but also a background in human rights to the conversation and has been working on data centers for a year now and knows more about them than I do apparently now. So Christian, maybe you want to add something that I missed.
Speaker #4
Thanks so much, Max. Well, I know about law and about human rights and about human rights law as well. So that may be a small addition. And also to say that I don't know more about data centers than you, obviously, since you've worked on this for what, the last 15 years of your career on digital infrastructure in general in countries around the globe. and also as a serial entrepreneur, of course, book and business and in... in civil society and so I'm very much looking forward to this conversation about one of our joint projects.
Speaker #2
Me too.
Speaker #3
Then Christian, welcome to Paradigms. It's really nice to have you as our first guest.
Speaker #4
Thank you.
Speaker #3
And today we are here to talk about your most recent publication as co-authors which is a 40-page big analysis and policy recommendations that you prepared as a response to European Commission's call for public input. before they develop the upcoming AI and Cloud Development Act. And in this response, you are both critical, claiming that Europe does not like computational infrastructure, challenging previous institutional documents. And you're also analytical as you outline six regulatory and policy interventions on the development of this Cloud and AI Development Act and the markets in Europe. So before we dive into the whole document, Let's first talk about how it came to be. And so during the process of the whole writing up the analysis, how did the collaboration between you two go? Who brought which strengths to the table?
Speaker #2
I think I can maybe start with a little bit of the process. So with our previous or other organization, the SDA, we've often been involved in these European policy processes. And there's, of course, a more informal... involvement talking to the policymakers directly also attending their events there's a lot of research being done on this act right now in the background that i i think will also come out soon and then there's the more formal response and our goal i think was to first of all not just challenge the assumptions but i think to also think it through what it actually could look like if Europe wants to develop on that front. Yeah, I think. That was the main overarching goal. And I think the strength that we were lacking in the past when we did this was always that we cannot really translate it into concrete policy. And we're really lacking sometimes the ability to frame it in this global context and really account for all the externalities from environmental concerns, but also from societal and human rights concerns. And maybe Christian will comment on. what strength you brought to this, but I hope I presented it on a platter.
Speaker #4
Well, I think it was a good combination of experiences here. So Max has obviously been involved in these discussions at the European level for a much longer time than I have. But I think the combined experience where I focused in my career on environmental and climatic impacts, plus on human rights, brought new dimensions to the table, which we integrated into the submission. I think we should also add that Max and I... received contributions from Corinne Kath at Article 9, which were also very helpful for us to think this through in a more comprehensive way. The other thing I think I should add is, because you said at first, Susanna, that the submission was critical and analytical, I would add to that that it was also constructive. And I think that we very deliberately focused, as Max also said just now, not just on what's wrong, according to us in the current proposal, but also how it could be fixed and what alternative options the European Commission, the EU legislator has. And I think that's very important that we do more than just criticize and analyze, but we also constructively propose alternatives. And that's very much what was on our mind when we were writing this together.
Speaker #3
And was it a very smooth... smooth
Speaker #2
process or how long did it take you what were the hurdles i'll let you want to go first max um i it's very it is very difficult um to write such a comprehensive analysis um i think what makes it even more difficult and it's something you you hear me talk about a lot is the lack of really transparent data on how much infrastructure do we have already how much data centers there are, how well are they utilized, how much energy do they actually use, not just estimations. And I think if that data would be there, this analysis would be very easy. But given that data is absent, I think what we spend most of the time on is proving something to the commission. It's very hard to show, especially with our very strong value set on fact-based recommendations. And I think that made it... more time-consuming than it could have been. Because I'm also curious to hear your take on it.
Speaker #4
I think that's a very relevant dimension. I think what also struck me is that, you know, if you look at the title of the proposed act, it seems very technical and it seems like a very sort of separate domain. But once you dive into this, we see that this is linked to various natural resources inputs. It's about downstream digital economy and how that functions. It's about competition on these markets. It's about their climate impact. It's about other environmental impacts. It's about downstream human rights impacts. So it's actually quite a multidimensional field. And so I think part of the process here was also for us to come together and sort of find a language in which we could capture all of that. which, as I just said, spans a number of fields of expertise. So it was quite a challenge, but I think it was worth going through that exercise to come up with a submission that was really in-depth and strong on multiple levels and 40 pages long, which might be a reflection of that process.
Speaker #2
That's the short version, right? Yeah,
Speaker #4
exactly. That's the exact summary.
Speaker #2
I also don't envy the commission having to weave this all together. I think it's a really difficult task. And I want to give a shout out also to Corinne because we spent a lot of time trying to build a framework or a model on how you can think about all of this as one big thing. And we call it the shared factory and the agile production machine. And we hope that, you know, even if you don't agree with our assumptions or you don't believe in our recommendations, I think. the model that especially came a lot out of Corinne's head really helps to think this through. And that's all we are asking, really. I think that's our fundamental ask. It's like, before we put billions of taxpayer money into developing gigafactories and all this kind of thing, please just think it through. Let's have a democratic process and be clear on what are the inputs we need and what are the outputs we are trying to accomplish with those inputs. And that's, I think, at the core is our message.
Speaker #4
I'm going to put on the head of interviewer now for a moment, Susanna, if you permit me. And I think it's helpful, Max, to explain what you, what we mean when we say shared factory. And actually, this is not just a sort of a performative action where I ask you a question so you can talk about it. But I think we actually spend a lot of time thinking through what is this exactly? And I think it's very important to take listeners through that process because it's not that straightforward.
Speaker #2
Yeah. I hope I can do it justice to explain it quickly. I think what we try to do is to say, look, cloud as a concept is a factory. And in order to be efficient, we basically say like instead of, you know, 4,000 companies using each building their own factory, of course, it makes sense to build one shared factory where we can all use it together. That is very efficient. and In that factory, we have equipment to manufacture things, to manufacture digital products and services. And those machines, you used to bring them yourself. And now you can also share those. And those are the cloud services that a lot of digital products and services are built upon. So not only is now the factory shared, but also the machinery is shared. But all of that is owned by a handful of companies. Yes, you gain an efficiency improvement, but you're also massively dependent on a single shared factory owner who happens to own both the building, the land, the resources, the machinery. And it's all integrated, which, of course, makes it really easy and desirable to produce digital products and services very quickly. Because I don't have to pay for the machinery. I don't have to pay for the factory. It's all there. And I can just focus on. you know, making AI services, building a startup. And I can do that really, really fast. But we don't ask the question at what cost. But in the pure pursuit of efficiency, it feels right, but it leads to massive dependence. And I think a good comparison is, of course, we have done similar things in the physical world with outsourcing to shared factory platforms abroad. And we are now slowly, and during Corona, we saw it even more. see the consequence of that. And I'm not opposed to globalization, but the beauty of the shared factory in cloud is that it's not per se a physical thing. So there we have a lot more opportunity to make it more distributed, to create lots of shared factories, to create European ones and things like this. So I feel like we have a lot more power in the cloud space than we have maybe in the physical production space. I hope I did it justice, Christian. I'm open to criticism.
Speaker #4
No, I think you did it justice. I think you highlighted the fact that there are on paper a lot of efficiencies in doing it this way. But what we also focus on, of course, in the submission is a lot of the inefficiencies here, among other things that, you know, prices are higher than you would expect in a very efficient market. That is very difficult to switch other than what you would expect in a very well-functioning market. And of course, going beyond market economics, there are a lot of inefficiencies in terms of dumping the effects on the rest of society and on our collective environment. And those were some of the what we called externalities that should be should be taken into account also when making decisions in this in this area. and I think that also goes back to what you just said about lack of information in order to make good policy in this area we need to have a good overall picture of just how it functions at present um this market or these markets what its inefficiencies are exactly what its externalities are exactly and then to have a democratic debate about that and make uh and make decisions and a lot of that is uh is currently missing and uh and so i think we are not pretending to have all the answers, but we are very clear on the fact that the... The current debate cannot be a rational one and an evidence-based one because we lack so much information. And there's so many aspects to this which haven't really been explored either in the original proposal by the commission or in the debate that surrounds it so far.
Speaker #2
To add to that, I would say especially the concept of the infinity factory, which is what we currently are pursuing. is not sustainable, not just in an environmental sense, but also an economic sense. You mentioned it's more expensive. It's more expensive because it's a factory where I can walk in and say I want to manufacture something and there's always a machine available to me. It is infinite production capacity for digital products or digital services that can't continue. And now the taxpayer is asked to fund the continuation of an infinity factory because we need even more. But we've never asked the question, for what and how much for what? And do we actually want to produce these products? But we are just expanding the factory to just make more and more and more. And the only person that really benefits is whoever owns the factory.
Speaker #4
Yeah. I mean, this is something that struck me as a relative outsider into these specific debates on digital infrastructure. I've worked on AI policy and human rights for a longer time, but not so much on this particular aspect of these markets. One thing that struck me is that, and it's in the name of the act as well, it's very much focused on the factories and on expanding the number of factories. There's remarkably little discussion about what's actually being made in these factories. So it's as if everyone is talking about making more factories without actually talking about the products and the services that will ultimately be produced. And it strikes me that that part of the debate is vague, to say the least. So what are we doing here exactly? What kind of problems are being resolved? What kind of products are being made? We need to ask critical questions there because all of these investments and all of these impacts, including environmental impacts, of course, need to be worth it if you're going to go in that direction. But are they? And it's fun to read all these brochures by AI companies on all the world's problems that will be solved in the future using AI, among other things. but I think Max you recently wrote to me about how many question marks there are in those documents. Like, wouldn't it be great if AI would solve climate change? Question mark. Question mark.
Speaker #2
They just want to work your cancer. Also, the maybe is always great.
Speaker #4
Yeah.
Speaker #2
Yeah, I think you're worrying too much. I really think you're really too concerned because as long as we keep on digitalizing, productivity growth will follow. Everything will get better. and just use the shared factory, build more digital services, build, build, build. That's the mantra. And venture capital, taxpayers and companies will pay for it anyways. And I'm being sarcastic. I hope the listeners know it's a gigantic extraction machine. The factory extracts money from government by government digitalization, from venture capital, which is partly funded by pension funds and by companies who are aiming to become more productive. and to be more efficient through digitalization. And everybody pays for the shared factory without being guaranteed any outcome.
Speaker #3
On this one, I also have another question that the EU Commission page also proposes that cloud investment will benefit startups and SMEs, as you've also hinted. But you warn against startup credit schemes. So why this pushback? Do you see a tension between supporting emerging firms and preventing... piggybacking on these hyperscalers?
Speaker #2
I mean, startups should exist, right? But I keep on posing the question, which is not mentioned in any of the policy documents or in any communication is, if I'm a startup, how do, first of all, how do I access the gigafactory? Right? First question, how do I get the compute that I need to build my startup from that? Second question, is it going to be cheaper than what I currently can get on the market? Because why would I go for that if it's not cheaper? Simple economics. And the third question is, if Google is giving me 300,000 euros worth of computation in their factory, am I going to get the equivalent from a government gigafactory? And if that's not solved, simple economic theory dictates I would choose the cheapest one. And if the cheapest one is free, I will go to the free one. So the policy intervention should be much more focused on... solving that equation, either making sure that these credits that are handed out can be spent anywhere, so making that more fluid, or not having the credits at all and having fair competition, or, for example, creating a European equivalent, which I think might end up being very expensive and raises the question for what are we spending this money again, what Christian raised before. But without solving that equation,
Speaker #4
why are you are you providing computes do you have anything to add well just slightly off topic but I think it's related as well What strikes me in this debate is that if you're being skeptical of initialists like this, I think you're easily put in the category of unrealist, non-realist. So it's simply good economics to go in this direction. So we should follow the path that we're currently already on and expand capacity. And so if you question that, you're basically being put in the category of the skeptic that really doesn't understand economics. But my counter question would be, is this really good economics? So what is actually the total productivity gain and the aggregate wealth that we're gaining from going into this direction? And then we try to look for evidence. so economic evidence of the benefits. There's actually a scarcity of that. And so I would like to put it on its head and say, how realistic is it to expect economic growth, if that's your goal, from heading in this direction, as the commission proposes here, if there's not much evidence that there will be? We cite a paper by Darren Acemoglu, who won the Nobel... Price and Economics last year or the year before that, who did the math and ends up saying that for the US in a 10-year period, the economic growth related to AI would be something in the order of one or one and a half percent. And that is not a lot in light of the massive investments made here and all the promises made on the benefit side, one percent. one and a half percent, if that's what you're aiming for, then perhaps we should tone down on all the expectations.
Speaker #2
And I think if you want economic growth, just build on that. I'm not advocating for that, but fundamentally you need cheap compute. AI economics are fundamentally driven by the cost of compute. Most companies today are not profitable because the compute is too expensive. And we know, Capitalist's playbook number one, competition. creates lower prices. Since when does the state flooding the market with cheap compute or, let's see if it's cheap at all, induce competition? It does not. It actually messes with the competitive aspects. And in an analogy, like with wind energy, we created the supply of wind energy as a consensus-based decision, especially in Germany, for society and society paid? for flooding the market with wind energy. We wanted that and we did that. And the outcome, though, was clear. We had more renewable energy, therefore we have less fossil fuels and we could shut down some of the coal-fired power plants. Very clear investment outcome. And wind energy wouldn't have worked if the state wouldn't have interfered with the economics of it. But in this case, I think we don't even have competition yet because it's such an oligopolistic market structure. So why interfere in a market when... it's already failing at competition. Shouldn't you then first focus on introducing competition on the pure compute trading layer? And I think we propose a lot of ideas on how to induce marketplaces and competition. But I think the commission is like three steps ahead of solving a problem that is a pure, maybe pure competitive problem for now.
Speaker #4
I mean, it goes even a bit further than that, I would say. And it's that they don't really want to interfere in the... competitive functioning. So it's actually, they want to stay away from that. So they're not actively interested in interfering there. So competition law is sort of brushed aside. Then economic benefits are unclear. Societal benefits, question mark, and environmental impacts are guaranteed. But yet somehow those environmental... those guaranteed environmental impacts, however big they are exactly, because that we don't know either, should be weighed against some kind of benefit. But again, question mark. So the whole totality of this is almost like a textbook case of something that's not been thought through completely. And that's not to sound arrogant, but I think there's such a rush and such a push to be ahead of this among policymakers, not just in the EU, but that happens globally. that asking these very simple critical questions are almost overlooked. And it's easy to ask questions, and then we need to have answers to those questions. And I think we're trying to do both in the submission, to both ask the simple questions. What is the benefit here exactly? But then also to try to think it through. Again, that's not something we can relay completely in a podcast, but anyone interested should actually spend some time and read what we wrote.
Speaker #0
We will find out the answer to those questions.
Speaker #3
Yes, I think asking these critical questions is very important for the crucial bodies to trip in this AI race that's happening currently. And you think then this narrative is driven by industry lobbying or just bad economics? You might have already talked about this, but a simple answer.
Speaker #4
One thing on that. So next, I'll let you answer. You're better placed to answer. But one thing to add is that I think for a lot of people, whether they're policymakers, whether they're in civil society or elsewhere, whether they're journalists, it's very hard to admit that you might not fully understand where this is going or where it's presently at, where everyone involved gives you the impression that they know. And so I think collectively what's happening here is that a lot of probably pretty intelligent actors in this debate are afraid to say that they also don't understand. But they don't want to be the one who asks the dumb question in class. I think that's the underlying sort of mass psychology that's going on here as well.
Speaker #2
And if anybody listening to this is like, I want to help. have some help ask dumb questions we are more than willing to come and ask stupid questions everywhere that's part of what we do um i think on the lobbying versus you know bad economics question in in their defense i think the entire ai mega sector is very open that the economics don't work right so nvidia just announced that they're going to pay for open ai's infrastructure oracle got the big contract for cloud computing with OpenAI as well, and 10 minutes later invested in OpenAI. So they're paying their own cloud bill. And I think they're very, and Sam Altman also openly says that at the current price, this is not sustainable. It should be costing 20,000 euros a month, not 20 euros a month. And whenever you have a business sector or market where the economics don't work, what do you do? You go to the state and you ask them to fix it, right? And I think that's literally what's happening. Whenever the NVIDIA CEO, OpenAI CEO, and Tropic CEO walks out of a commission building, 10 minutes later, I feel like we get an announcement that the EU is going to invest billions. And I think that's not the job here. It is not the job of the state to fix market economics. If something doesn't work, let it fail. Because that's what we created capitalism for, to make sure... Good ideas survive in a market and bad ideas fail. And right now, I think with AI, we're not saying it's a bad idea, just saying as it currently stands, it's maybe better placed in universities for more research and not ready for the market. Because apparently in the market, it doesn't work.
Speaker #4
There's also a bit of an irony here, right? Where the big winners in capitalism come to governments. to help them out while at the same time professing their government should stay out of their business as much as possible because it will ruin the economics of it.
Speaker #2
Please build more data centers and more energy infrastructure but please do not look closely at how we're making money.
Speaker #3
Very nice retention based in a relationship to have to navigate the state and tech. And then back to the content of your work, your paper suggests key interventions. like European Digital Resource Exchange or migration loans and tighter takeover rules. Which of these do you see as the hook that could make it to the final regulation or which would benefit the most?
Speaker #2
I think there are the things that are more the macro topics that if they would make it, they would have to be the whole policy. For example, creating a marketplace to trade compute would require complete rewiring of the market and really... solving a lot of competition law problems and making sure this is compatible with the World Trade Organization. It's the holy grail, I would say.
Speaker #0
But then we also suggested some very simple, well, simple, but more tangible interventions that could be part of any other policy as well. Migrations and other things like this. But maybe, Christian, you want to outline a few of those.
Speaker #1
Well, I think, first of all, at a bit more of an abstract level, I think what's very important to highlight is that if you perceive as... just any market as functioning, that market will give signals that indicate where other participants should be heading. This is a market where a lot of those signals are hidden. So in terms of price, what is the actual price of a cloud offering, for instance? What kind of cross subsidization has happened within vertically integrated firms, for instance? A lot of the aspects that you would expect in a well-functioning market. are missing here. And that also robs regulators, governments, and so at the end of the day, the public, democracy, robs it of the chance to steer it in a certain direction. So we don't have good signals. Therefore, we don't really know what's going on in these markets. And because we really don't know what's going on in these markets, we really can't regulate on outcomes and effect. So I think One of the things that the submission focuses on is how you can let that market function as a market shoot in order for the EU authorities or member states, and that same would apply for outside of the EU, to intervene where they see that as a fit option. So I think that's an important thing to say about our proposals. I think, as Max said... We presented a whole, and so ideally we would like to see the whole implemented, but you can still take parts of it and make it better. And I think Max already talked about startup credits. There are other forms of non-competitive behavior that I think could be addressed in the policy proposals that we have. So in other words, to make the market a little bit fairer. And even if some of those would be implemented, it would be a better situation than we're in currently.
Speaker #2
It sounds very convenient to make these interventions applicable not only here, but in other areas as well. But once the consultation process closes and the act proceeds, what is LightMontiv's next step? Is it advocacy, monitoring or coalition building? What do you expect will happen?
Speaker #1
Well, first, it's not over yet. And I think the latest information is that the process has been postponed slightly. So we're looking towards finalization by the first quarter of 2026, if I'm not mistaken. So there's still some time to go. And in a rapidly changing world in general and lightning speed developments in AI more specifically. I don't think we can truly predict what's going to happen in the next few months. So we'll be continuing our legislative advocacy until the time that the act is finally adopted, if it gets to that point. We haven't internally discussed what to do next because the outcome is still unclear. But even with adopted legislation, in my experience, there's a lot of work to be done on implementation and also, of course, on potential future changes and amendments to that legislation. So it's never over.
Speaker #0
I think there's nothing to add to that. I think advocacy is in our blood. So I think it's very difficult to not do it. So, so. I think we are continuously going to be involved in this process. I think in this case, it's very difficult to form a coalition in that sense, because the business interests here are very stark. Like you have the incumbents, so to say, the US cloud providers, US AI companies, and even some European actors in that. You have the whole Eurostack movement, which is about like, give us so much money that we can basically rebuild everything European. And I think we find ourselves in the middle, being the most pragmatic, the most practical, let's restore competition on compute actors. I think it's at the moment, at least it's better in my perspective, that we stay alone and independent and move our agenda forward than to try to build too many groups of interests where we have to balance each other's agenda.
Speaker #1
It's helpful, though, to add to that, to see that the landscape in Brussels is slightly changing, it seems. There are more civil society organizations interested in... debates around data centers because of environmental impacts primarily. So it's good to see that a number of civil society organizations working in that area have also made submissions to this process from Beyond Fossil Fuels and Green Screen Coalition, Client Earth and others. I think that helps put extra spotlight on the environmental dimension. So in the A bigger scheme of things, I think, that will help sort of balance the debate, which has been very unbalanced until this point in time. And I think ultimately, without us necessarily having to form that coalition, it's clear that interests align among a lot of different actors, whether they're corporations or NGOs, that all at least want things to be very different from what's in the current commission proposal. I think among civil society, among certain, especially smaller businesses in Europe, parts of the general public, there's an unease with the direction this is heading in. And I think in that sense, there's an alignment without those interests being fully similar and in coordination.
Speaker #2
Sounds promising. And let's see what happens. And let's hope for the best for our digital future. Thank you very much for joining me today for the conversation. Thank you for contributing with this important piece of work. And yeah, have a nice day.
Speaker #1
Thanks so much for the invitation.

About Paradigms

Apple Podcasts: https://ltmv.io/podcast-apple

Spotify: https://ltmv.io/podcast-spotify

YouTube: https://ltmv.io/podcast-yt

Hosted on Ausha. See ausha.co/privacy-policy for more information.

About Paradigms

Apple Podcasts: https://ltmv.io/podcast-apple

Spotify: https://ltmv.io/podcast-spotify

YouTube: https://ltmv.io/podcast-yt

Hosted on Ausha. See ausha.co/privacy-policy for more information.

Description

And their ultimate proposal for the EU Commission? Fix market failures first, then invest where public value is clear.

(Correction (5:29): Christiaan mentions Article 9, but the correct version is Article 19.)

Paradigms is a podcast produced by Leitmotiv – an independent think tank focused on operationalizing values in our digital future.

Hosted on Ausha. See ausha.co/privacy-policy for more information.

Transcription

Speaker #0
This is the next chapter.
Speaker #1
Welcome to Paradigms, a podcast by Leitmotiv, the independent digital policy think tank that exists to operationalize values in our digital future.
Speaker #2
In each episode, we unpack the latest developments and debates in our digital world. Together, we try to make sense. of what is happening to our digital future, who is shaping it, and what we can do to change its path.
Speaker #1
My name is Susanna Ludwikova and I'm here to ask the difficult questions.
Speaker #2
And my name is Max Schulze and I'm here to give simple answers. At Leibmotiv, I think about the policies that shape our digital future and how to operationalize our societal values within them.
Speaker #1
And my job is to bring our thinking to your ears, eyes and minds. In our podcast, we invite you to think with us as we unpack the building blocks of our digital world and work towards a better future that amplifies the best of our core values and our humanity.
Speaker #3
Today, Max and I have a guest in our virtual studio, who is Christian. Can you guys introduce each other since you are both colleagues and friends?
Speaker #2
Of course. I will go first. So Christian is joining us and I think it's really nice to have him here today because he's also joining us as a co-founder for Leitmotiv, our digital policy think tank. And I'm really happy to have him here, especially because he brings a background in law, but also a background in human rights to the conversation and has been working on data centers for a year now and knows more about them than I do apparently now. So Christian, maybe you want to add something that I missed.
Speaker #4
Thanks so much, Max. Well, I know about law and about human rights and about human rights law as well. So that may be a small addition. And also to say that I don't know more about data centers than you, obviously, since you've worked on this for what, the last 15 years of your career on digital infrastructure in general in countries around the globe. and also as a serial entrepreneur, of course, book and business and in... in civil society and so I'm very much looking forward to this conversation about one of our joint projects.
Speaker #2
Me too.
Speaker #3
Then Christian, welcome to Paradigms. It's really nice to have you as our first guest.
Speaker #4
Thank you.
Speaker #3
And today we are here to talk about your most recent publication as co-authors which is a 40-page big analysis and policy recommendations that you prepared as a response to European Commission's call for public input. before they develop the upcoming AI and Cloud Development Act. And in this response, you are both critical, claiming that Europe does not like computational infrastructure, challenging previous institutional documents. And you're also analytical as you outline six regulatory and policy interventions on the development of this Cloud and AI Development Act and the markets in Europe. So before we dive into the whole document, Let's first talk about how it came to be. And so during the process of the whole writing up the analysis, how did the collaboration between you two go? Who brought which strengths to the table?
Speaker #2
I think I can maybe start with a little bit of the process. So with our previous or other organization, the SDA, we've often been involved in these European policy processes. And there's, of course, a more informal... involvement talking to the policymakers directly also attending their events there's a lot of research being done on this act right now in the background that i i think will also come out soon and then there's the more formal response and our goal i think was to first of all not just challenge the assumptions but i think to also think it through what it actually could look like if Europe wants to develop on that front. Yeah, I think. That was the main overarching goal. And I think the strength that we were lacking in the past when we did this was always that we cannot really translate it into concrete policy. And we're really lacking sometimes the ability to frame it in this global context and really account for all the externalities from environmental concerns, but also from societal and human rights concerns. And maybe Christian will comment on. what strength you brought to this, but I hope I presented it on a platter.
Speaker #4
Well, I think it was a good combination of experiences here. So Max has obviously been involved in these discussions at the European level for a much longer time than I have. But I think the combined experience where I focused in my career on environmental and climatic impacts, plus on human rights, brought new dimensions to the table, which we integrated into the submission. I think we should also add that Max and I... received contributions from Corinne Kath at Article 9, which were also very helpful for us to think this through in a more comprehensive way. The other thing I think I should add is, because you said at first, Susanna, that the submission was critical and analytical, I would add to that that it was also constructive. And I think that we very deliberately focused, as Max also said just now, not just on what's wrong, according to us in the current proposal, but also how it could be fixed and what alternative options the European Commission, the EU legislator has. And I think that's very important that we do more than just criticize and analyze, but we also constructively propose alternatives. And that's very much what was on our mind when we were writing this together.
Speaker #3
And was it a very smooth... smooth
Speaker #2
process or how long did it take you what were the hurdles i'll let you want to go first max um i it's very it is very difficult um to write such a comprehensive analysis um i think what makes it even more difficult and it's something you you hear me talk about a lot is the lack of really transparent data on how much infrastructure do we have already how much data centers there are, how well are they utilized, how much energy do they actually use, not just estimations. And I think if that data would be there, this analysis would be very easy. But given that data is absent, I think what we spend most of the time on is proving something to the commission. It's very hard to show, especially with our very strong value set on fact-based recommendations. And I think that made it... more time-consuming than it could have been. Because I'm also curious to hear your take on it.
Speaker #4
I think that's a very relevant dimension. I think what also struck me is that, you know, if you look at the title of the proposed act, it seems very technical and it seems like a very sort of separate domain. But once you dive into this, we see that this is linked to various natural resources inputs. It's about downstream digital economy and how that functions. It's about competition on these markets. It's about their climate impact. It's about other environmental impacts. It's about downstream human rights impacts. So it's actually quite a multidimensional field. And so I think part of the process here was also for us to come together and sort of find a language in which we could capture all of that. which, as I just said, spans a number of fields of expertise. So it was quite a challenge, but I think it was worth going through that exercise to come up with a submission that was really in-depth and strong on multiple levels and 40 pages long, which might be a reflection of that process.
Speaker #2
That's the short version, right? Yeah,
Speaker #4
exactly. That's the exact summary.
Speaker #2
I also don't envy the commission having to weave this all together. I think it's a really difficult task. And I want to give a shout out also to Corinne because we spent a lot of time trying to build a framework or a model on how you can think about all of this as one big thing. And we call it the shared factory and the agile production machine. And we hope that, you know, even if you don't agree with our assumptions or you don't believe in our recommendations, I think. the model that especially came a lot out of Corinne's head really helps to think this through. And that's all we are asking, really. I think that's our fundamental ask. It's like, before we put billions of taxpayer money into developing gigafactories and all this kind of thing, please just think it through. Let's have a democratic process and be clear on what are the inputs we need and what are the outputs we are trying to accomplish with those inputs. And that's, I think, at the core is our message.
Speaker #4
I'm going to put on the head of interviewer now for a moment, Susanna, if you permit me. And I think it's helpful, Max, to explain what you, what we mean when we say shared factory. And actually, this is not just a sort of a performative action where I ask you a question so you can talk about it. But I think we actually spend a lot of time thinking through what is this exactly? And I think it's very important to take listeners through that process because it's not that straightforward.
Speaker #2
Yeah. I hope I can do it justice to explain it quickly. I think what we try to do is to say, look, cloud as a concept is a factory. And in order to be efficient, we basically say like instead of, you know, 4,000 companies using each building their own factory, of course, it makes sense to build one shared factory where we can all use it together. That is very efficient. and In that factory, we have equipment to manufacture things, to manufacture digital products and services. And those machines, you used to bring them yourself. And now you can also share those. And those are the cloud services that a lot of digital products and services are built upon. So not only is now the factory shared, but also the machinery is shared. But all of that is owned by a handful of companies. Yes, you gain an efficiency improvement, but you're also massively dependent on a single shared factory owner who happens to own both the building, the land, the resources, the machinery. And it's all integrated, which, of course, makes it really easy and desirable to produce digital products and services very quickly. Because I don't have to pay for the machinery. I don't have to pay for the factory. It's all there. And I can just focus on. you know, making AI services, building a startup. And I can do that really, really fast. But we don't ask the question at what cost. But in the pure pursuit of efficiency, it feels right, but it leads to massive dependence. And I think a good comparison is, of course, we have done similar things in the physical world with outsourcing to shared factory platforms abroad. And we are now slowly, and during Corona, we saw it even more. see the consequence of that. And I'm not opposed to globalization, but the beauty of the shared factory in cloud is that it's not per se a physical thing. So there we have a lot more opportunity to make it more distributed, to create lots of shared factories, to create European ones and things like this. So I feel like we have a lot more power in the cloud space than we have maybe in the physical production space. I hope I did it justice, Christian. I'm open to criticism.
Speaker #4
No, I think you did it justice. I think you highlighted the fact that there are on paper a lot of efficiencies in doing it this way. But what we also focus on, of course, in the submission is a lot of the inefficiencies here, among other things that, you know, prices are higher than you would expect in a very efficient market. That is very difficult to switch other than what you would expect in a very well-functioning market. And of course, going beyond market economics, there are a lot of inefficiencies in terms of dumping the effects on the rest of society and on our collective environment. And those were some of the what we called externalities that should be should be taken into account also when making decisions in this in this area. and I think that also goes back to what you just said about lack of information in order to make good policy in this area we need to have a good overall picture of just how it functions at present um this market or these markets what its inefficiencies are exactly what its externalities are exactly and then to have a democratic debate about that and make uh and make decisions and a lot of that is uh is currently missing and uh and so i think we are not pretending to have all the answers, but we are very clear on the fact that the... The current debate cannot be a rational one and an evidence-based one because we lack so much information. And there's so many aspects to this which haven't really been explored either in the original proposal by the commission or in the debate that surrounds it so far.
Speaker #2
To add to that, I would say especially the concept of the infinity factory, which is what we currently are pursuing. is not sustainable, not just in an environmental sense, but also an economic sense. You mentioned it's more expensive. It's more expensive because it's a factory where I can walk in and say I want to manufacture something and there's always a machine available to me. It is infinite production capacity for digital products or digital services that can't continue. And now the taxpayer is asked to fund the continuation of an infinity factory because we need even more. But we've never asked the question, for what and how much for what? And do we actually want to produce these products? But we are just expanding the factory to just make more and more and more. And the only person that really benefits is whoever owns the factory.
Speaker #4
Yeah. I mean, this is something that struck me as a relative outsider into these specific debates on digital infrastructure. I've worked on AI policy and human rights for a longer time, but not so much on this particular aspect of these markets. One thing that struck me is that, and it's in the name of the act as well, it's very much focused on the factories and on expanding the number of factories. There's remarkably little discussion about what's actually being made in these factories. So it's as if everyone is talking about making more factories without actually talking about the products and the services that will ultimately be produced. And it strikes me that that part of the debate is vague, to say the least. So what are we doing here exactly? What kind of problems are being resolved? What kind of products are being made? We need to ask critical questions there because all of these investments and all of these impacts, including environmental impacts, of course, need to be worth it if you're going to go in that direction. But are they? And it's fun to read all these brochures by AI companies on all the world's problems that will be solved in the future using AI, among other things. but I think Max you recently wrote to me about how many question marks there are in those documents. Like, wouldn't it be great if AI would solve climate change? Question mark. Question mark.
Speaker #2
They just want to work your cancer. Also, the maybe is always great.
Speaker #4
Yeah.
Speaker #2
Yeah, I think you're worrying too much. I really think you're really too concerned because as long as we keep on digitalizing, productivity growth will follow. Everything will get better. and just use the shared factory, build more digital services, build, build, build. That's the mantra. And venture capital, taxpayers and companies will pay for it anyways. And I'm being sarcastic. I hope the listeners know it's a gigantic extraction machine. The factory extracts money from government by government digitalization, from venture capital, which is partly funded by pension funds and by companies who are aiming to become more productive. and to be more efficient through digitalization. And everybody pays for the shared factory without being guaranteed any outcome.
Speaker #3
On this one, I also have another question that the EU Commission page also proposes that cloud investment will benefit startups and SMEs, as you've also hinted. But you warn against startup credit schemes. So why this pushback? Do you see a tension between supporting emerging firms and preventing... piggybacking on these hyperscalers?
Speaker #2
I mean, startups should exist, right? But I keep on posing the question, which is not mentioned in any of the policy documents or in any communication is, if I'm a startup, how do, first of all, how do I access the gigafactory? Right? First question, how do I get the compute that I need to build my startup from that? Second question, is it going to be cheaper than what I currently can get on the market? Because why would I go for that if it's not cheaper? Simple economics. And the third question is, if Google is giving me 300,000 euros worth of computation in their factory, am I going to get the equivalent from a government gigafactory? And if that's not solved, simple economic theory dictates I would choose the cheapest one. And if the cheapest one is free, I will go to the free one. So the policy intervention should be much more focused on... solving that equation, either making sure that these credits that are handed out can be spent anywhere, so making that more fluid, or not having the credits at all and having fair competition, or, for example, creating a European equivalent, which I think might end up being very expensive and raises the question for what are we spending this money again, what Christian raised before. But without solving that equation,
Speaker #4
why are you are you providing computes do you have anything to add well just slightly off topic but I think it's related as well What strikes me in this debate is that if you're being skeptical of initialists like this, I think you're easily put in the category of unrealist, non-realist. So it's simply good economics to go in this direction. So we should follow the path that we're currently already on and expand capacity. And so if you question that, you're basically being put in the category of the skeptic that really doesn't understand economics. But my counter question would be, is this really good economics? So what is actually the total productivity gain and the aggregate wealth that we're gaining from going into this direction? And then we try to look for evidence. so economic evidence of the benefits. There's actually a scarcity of that. And so I would like to put it on its head and say, how realistic is it to expect economic growth, if that's your goal, from heading in this direction, as the commission proposes here, if there's not much evidence that there will be? We cite a paper by Darren Acemoglu, who won the Nobel... Price and Economics last year or the year before that, who did the math and ends up saying that for the US in a 10-year period, the economic growth related to AI would be something in the order of one or one and a half percent. And that is not a lot in light of the massive investments made here and all the promises made on the benefit side, one percent. one and a half percent, if that's what you're aiming for, then perhaps we should tone down on all the expectations.
Speaker #2
And I think if you want economic growth, just build on that. I'm not advocating for that, but fundamentally you need cheap compute. AI economics are fundamentally driven by the cost of compute. Most companies today are not profitable because the compute is too expensive. And we know, Capitalist's playbook number one, competition. creates lower prices. Since when does the state flooding the market with cheap compute or, let's see if it's cheap at all, induce competition? It does not. It actually messes with the competitive aspects. And in an analogy, like with wind energy, we created the supply of wind energy as a consensus-based decision, especially in Germany, for society and society paid? for flooding the market with wind energy. We wanted that and we did that. And the outcome, though, was clear. We had more renewable energy, therefore we have less fossil fuels and we could shut down some of the coal-fired power plants. Very clear investment outcome. And wind energy wouldn't have worked if the state wouldn't have interfered with the economics of it. But in this case, I think we don't even have competition yet because it's such an oligopolistic market structure. So why interfere in a market when... it's already failing at competition. Shouldn't you then first focus on introducing competition on the pure compute trading layer? And I think we propose a lot of ideas on how to induce marketplaces and competition. But I think the commission is like three steps ahead of solving a problem that is a pure, maybe pure competitive problem for now.
Speaker #4
I mean, it goes even a bit further than that, I would say. And it's that they don't really want to interfere in the... competitive functioning. So it's actually, they want to stay away from that. So they're not actively interested in interfering there. So competition law is sort of brushed aside. Then economic benefits are unclear. Societal benefits, question mark, and environmental impacts are guaranteed. But yet somehow those environmental... those guaranteed environmental impacts, however big they are exactly, because that we don't know either, should be weighed against some kind of benefit. But again, question mark. So the whole totality of this is almost like a textbook case of something that's not been thought through completely. And that's not to sound arrogant, but I think there's such a rush and such a push to be ahead of this among policymakers, not just in the EU, but that happens globally. that asking these very simple critical questions are almost overlooked. And it's easy to ask questions, and then we need to have answers to those questions. And I think we're trying to do both in the submission, to both ask the simple questions. What is the benefit here exactly? But then also to try to think it through. Again, that's not something we can relay completely in a podcast, but anyone interested should actually spend some time and read what we wrote.
Speaker #0
We will find out the answer to those questions.
Speaker #3
Yes, I think asking these critical questions is very important for the crucial bodies to trip in this AI race that's happening currently. And you think then this narrative is driven by industry lobbying or just bad economics? You might have already talked about this, but a simple answer.
Speaker #4
One thing on that. So next, I'll let you answer. You're better placed to answer. But one thing to add is that I think for a lot of people, whether they're policymakers, whether they're in civil society or elsewhere, whether they're journalists, it's very hard to admit that you might not fully understand where this is going or where it's presently at, where everyone involved gives you the impression that they know. And so I think collectively what's happening here is that a lot of probably pretty intelligent actors in this debate are afraid to say that they also don't understand. But they don't want to be the one who asks the dumb question in class. I think that's the underlying sort of mass psychology that's going on here as well.
Speaker #2
And if anybody listening to this is like, I want to help. have some help ask dumb questions we are more than willing to come and ask stupid questions everywhere that's part of what we do um i think on the lobbying versus you know bad economics question in in their defense i think the entire ai mega sector is very open that the economics don't work right so nvidia just announced that they're going to pay for open ai's infrastructure oracle got the big contract for cloud computing with OpenAI as well, and 10 minutes later invested in OpenAI. So they're paying their own cloud bill. And I think they're very, and Sam Altman also openly says that at the current price, this is not sustainable. It should be costing 20,000 euros a month, not 20 euros a month. And whenever you have a business sector or market where the economics don't work, what do you do? You go to the state and you ask them to fix it, right? And I think that's literally what's happening. Whenever the NVIDIA CEO, OpenAI CEO, and Tropic CEO walks out of a commission building, 10 minutes later, I feel like we get an announcement that the EU is going to invest billions. And I think that's not the job here. It is not the job of the state to fix market economics. If something doesn't work, let it fail. Because that's what we created capitalism for, to make sure... Good ideas survive in a market and bad ideas fail. And right now, I think with AI, we're not saying it's a bad idea, just saying as it currently stands, it's maybe better placed in universities for more research and not ready for the market. Because apparently in the market, it doesn't work.
Speaker #4
There's also a bit of an irony here, right? Where the big winners in capitalism come to governments. to help them out while at the same time professing their government should stay out of their business as much as possible because it will ruin the economics of it.
Speaker #2
Please build more data centers and more energy infrastructure but please do not look closely at how we're making money.
Speaker #3
Very nice retention based in a relationship to have to navigate the state and tech. And then back to the content of your work, your paper suggests key interventions. like European Digital Resource Exchange or migration loans and tighter takeover rules. Which of these do you see as the hook that could make it to the final regulation or which would benefit the most?
Speaker #2
I think there are the things that are more the macro topics that if they would make it, they would have to be the whole policy. For example, creating a marketplace to trade compute would require complete rewiring of the market and really... solving a lot of competition law problems and making sure this is compatible with the World Trade Organization. It's the holy grail, I would say.
Speaker #0
But then we also suggested some very simple, well, simple, but more tangible interventions that could be part of any other policy as well. Migrations and other things like this. But maybe, Christian, you want to outline a few of those.
Speaker #1
Well, I think, first of all, at a bit more of an abstract level, I think what's very important to highlight is that if you perceive as... just any market as functioning, that market will give signals that indicate where other participants should be heading. This is a market where a lot of those signals are hidden. So in terms of price, what is the actual price of a cloud offering, for instance? What kind of cross subsidization has happened within vertically integrated firms, for instance? A lot of the aspects that you would expect in a well-functioning market. are missing here. And that also robs regulators, governments, and so at the end of the day, the public, democracy, robs it of the chance to steer it in a certain direction. So we don't have good signals. Therefore, we don't really know what's going on in these markets. And because we really don't know what's going on in these markets, we really can't regulate on outcomes and effect. So I think One of the things that the submission focuses on is how you can let that market function as a market shoot in order for the EU authorities or member states, and that same would apply for outside of the EU, to intervene where they see that as a fit option. So I think that's an important thing to say about our proposals. I think, as Max said... We presented a whole, and so ideally we would like to see the whole implemented, but you can still take parts of it and make it better. And I think Max already talked about startup credits. There are other forms of non-competitive behavior that I think could be addressed in the policy proposals that we have. So in other words, to make the market a little bit fairer. And even if some of those would be implemented, it would be a better situation than we're in currently.
Speaker #2
It sounds very convenient to make these interventions applicable not only here, but in other areas as well. But once the consultation process closes and the act proceeds, what is LightMontiv's next step? Is it advocacy, monitoring or coalition building? What do you expect will happen?
Speaker #1
Well, first, it's not over yet. And I think the latest information is that the process has been postponed slightly. So we're looking towards finalization by the first quarter of 2026, if I'm not mistaken. So there's still some time to go. And in a rapidly changing world in general and lightning speed developments in AI more specifically. I don't think we can truly predict what's going to happen in the next few months. So we'll be continuing our legislative advocacy until the time that the act is finally adopted, if it gets to that point. We haven't internally discussed what to do next because the outcome is still unclear. But even with adopted legislation, in my experience, there's a lot of work to be done on implementation and also, of course, on potential future changes and amendments to that legislation. So it's never over.
Speaker #0
I think there's nothing to add to that. I think advocacy is in our blood. So I think it's very difficult to not do it. So, so. I think we are continuously going to be involved in this process. I think in this case, it's very difficult to form a coalition in that sense, because the business interests here are very stark. Like you have the incumbents, so to say, the US cloud providers, US AI companies, and even some European actors in that. You have the whole Eurostack movement, which is about like, give us so much money that we can basically rebuild everything European. And I think we find ourselves in the middle, being the most pragmatic, the most practical, let's restore competition on compute actors. I think it's at the moment, at least it's better in my perspective, that we stay alone and independent and move our agenda forward than to try to build too many groups of interests where we have to balance each other's agenda.
Speaker #1
It's helpful, though, to add to that, to see that the landscape in Brussels is slightly changing, it seems. There are more civil society organizations interested in... debates around data centers because of environmental impacts primarily. So it's good to see that a number of civil society organizations working in that area have also made submissions to this process from Beyond Fossil Fuels and Green Screen Coalition, Client Earth and others. I think that helps put extra spotlight on the environmental dimension. So in the A bigger scheme of things, I think, that will help sort of balance the debate, which has been very unbalanced until this point in time. And I think ultimately, without us necessarily having to form that coalition, it's clear that interests align among a lot of different actors, whether they're corporations or NGOs, that all at least want things to be very different from what's in the current commission proposal. I think among civil society, among certain, especially smaller businesses in Europe, parts of the general public, there's an unease with the direction this is heading in. And I think in that sense, there's an alignment without those interests being fully similar and in coordination.
Speaker #2
Sounds promising. And let's see what happens. And let's hope for the best for our digital future. Thank you very much for joining me today for the conversation. Thank you for contributing with this important piece of work. And yeah, have a nice day.
Speaker #1
Thanks so much for the invitation.

About Paradigms

Apple Podcasts: https://ltmv.io/podcast-apple

Spotify: https://ltmv.io/podcast-spotify

YouTube: https://ltmv.io/podcast-yt

Hosted on Ausha. See ausha.co/privacy-policy for more information.

About Paradigms

Apple Podcasts: https://ltmv.io/podcast-apple

Spotify: https://ltmv.io/podcast-spotify

YouTube: https://ltmv.io/podcast-yt

Hosted on Ausha. See ausha.co/privacy-policy for more information.

Embed