Mistral AI faces lawmakers: Arthur Mensch sounds the alarm on digital sovereignty

I swear it.

Very well, Madam President, Mr. Rapporteur who is not here, so Madam Rapporteur, or the rapporteur, I don't know anymore. Ladies and gentlemen deputies, well, I want to thank you for your invitation to discuss this key subject, which is the autonomy of Europe in terms of artificial intelligence, in terms of digital services in general. That is to say, I think we shouldn't make a distinction between today's digital services and especially tomorrow's, which is essentially artificial intelligence. You have storage, you have a bit of execution of agents on it, etc. But fundamentally, digital services, the cloud, it's artificial intelligence. There is no distinction to be made on this subject. And so that may answer your question. I noted your two questions, I will answer them later. Perhaps to give a little context. So as you know, artificial intelligence is redefining economic and geopolitical balances. The question is not so much whether Europe can compete, but rather how it can do it. Because I think that if it doesn't compete, in reality, it completely gives up any form of presence and voice in the concert of nations. So the question is, can it do it without depending on extra-European actors? The answer is obviously no. The second question is, can it do it by gaining leverage over its suppliers? By gaining leverage as a technology exporter and by managing to rebalance the economic imbalances that already exist and will increase? I think the answer is yes. A few elements on Mistral AI. So it's a company that we founded a little over 3 years ago. So on April 28, 2023, with the intuition in 2023 that generative artificial intelligence was going to significantly shake up the cards in the field of digital services, was going to change quite profoundly the way the economy works because fundamentally it moves intelligence from labor to capital and to machines. We saw in 2023 because we had, me and my co-founders Guillaume and Timothée, had a certain experience with the big American players that the world was heading towards a fairly clear oligopoly and we obviously wanted to counter this oligopoly and so we created Mistral in 2023 with about fifteen people in France and England at the start, essentially researchers. What we did was first demonstrate to the world that we knew how to train language models, which is what generative artificial intelligence is about. And then we demonstrated, we put them in, we made an open model, that is to say we gave them to the world to show that we could also take them, modify them, deploy them where we wanted, etc. So it was a technique of decentralization and displacement of established actors. And then from there, we continued to invest in models. Today, we have invested 1 billion this year in R&D. So we invest very significantly, we build models, we are one of the leaders in open models that we can modify, that we can deploy where we want. And we have a whole economic model that is both above, that is to say we will provide a platform for deploying agents, delegating tasks to artificial intelligence. We will create business applications in contact with our clients who are companies in general, large companies. And then we have an economic model that is below, that is to say we also host the models, we put them on machines and we make them generate what we call tokens, which is the economic unit of artificial intelligence. That is to say we train models that generate sequences. From these sequence generations, we will take actions, we will execute things, we will think. These sequences are made of what we call tokens, which are numbers, between 0 and 65,000. And so we generate these tokens for our clients, hosting them, we transform electrons into tokens. There, today, a few figures, we are 1,000 employees, a valuation of 12 billion euros, and 1 billion euros in revenue targeted by the end of the year. Our clients are enterprise clients, clients in the public sector. To name a few in France, we have the DINUM, the Caisse des Dépôts, France Travail. In the private sector, CMA CGM, Stellantis, TotalEnergies, BNP Paribas, etc. We do about 30% of our turnover in France, 70% abroad, and about three-quarters of our turnover in Europe. To give you an idea of where we are, we are fundamentally a technology exporter to both the United States and Asia. So what you need to understand about generative artificial intelligence and fundamentally about what we call artificial intelligence today, is that it's a model that consists of transforming energy into intelligence, into token generation which are then used to delegate tasks, to do mathematical research, to operate logistics chains, to improve tools, to deploy systems embedded in robots, in drones, etc. So we are in a business where we transform electricity into token generation. And it's fundamentally something that needs to be treated more or less like a natural resource. That is to say, we need to think about intelligence as we think about energy. So our goal is to provide intelligence that is cheap, affordable, we'll say in English, which is secure in terms of supply, that is to say we don't need to go get it from Americans for example, and which is also sustainable because it transforms energy that is less carbon-intensive than elsewhere because we use in particular French energy. I think we don't think enough in this way. Intelligence is like electricity, it's like energy. And besides, we have this tendency, which in my opinion is a fairly false tendency, to separate digital services from artificial intelligence. Today, the cloud is artificial intelligence. The growth of the cloud is artificial intelligence. High value-added services, with high margins that allow for R&D, it's artificial intelligence. So in fact, everything else, from the moment we gain in artificial intelligence, we can build everything else which are essentially cloud services that are commodities today and which essentially consist of knowing how to operate open-source software because it has become the global infrastructure. Our goal is obviously to create this value chain. So this, I would say, our chain, where we are in the value chain is between the operation of clusters and the operation of servers up to the creation of business applications. And I think it's an atomic unit of the value chain that goes from the construction of semiconductors, so from ASML for example, this company which is one of our clients which does lithography, to deployment in companies. So that means that the cloud and AI are really the same thing. What I can say is that in fact, we must see sovereignty. We must stop thinking about sovereignty as isolationism. We must think about it as business people think about it. That is to say, we must think about sovereignty as a leverage issue. That is to say, in a world where you import all your digital services from the United States, you have no leverage over the United States. In a world where you create part of your service, possibly export it, you have additional leverage. In a world where you sell goods to the United States, you have leverage over them. We see it in European Union negotiations, we export a certain number of things. That gives us arguments and cards to play when we are at the negotiating table. If we don't have this capacity to not import all our digital services from the United States, we have a real problem. The problem that arises, in fact, is that what we are talking about, I think the total market we are talking about, we must compare it to the payroll. Today, at Mistral, our consumption of artificial intelligence for our employees is 10% of our payroll. Now, we make this technology, so we adopt it a little faster than elsewhere. And if you extrapolate, in 3 to 4 years, all of Europe uses artificial intelligence for 10% of its payroll. 10% of Europe's payroll is about 1 trillion. 1 trillion, so it's digital services because you have to generate the tokens, etc. It's a trillion if we use non-European technology, it's a trillion trade deficit to add to our existing trade deficit on digital services. It's a trillion that goes elsewhere to be invested in R&D and is not reinvested in R&D in France, so in France or in Europe. So that, I think, is a very important point, is that we must think about this sovereignty issue from a macroeconomic point of view. An essential macroeconomic issue and I think we don't realize the extent of it. So that's a first issue. There is obviously an issue of economic security, that is to say when you know how to deploy artificial intelligence, you want to, in essential services, you want to ensure that access to the service cannot be cut off, which is actually quite theoretical. We've seen it in different places. So that means it's important for all sovereign, governmental aspects. It's obviously important for defense because artificial intelligence is now at the heart of operational centers of the armed forces and therefore becomes the platform that allows making operational decisions on the battlefield, that allows for intelligence, etc. So if we don't have access to this technology, we will have to import it from elsewhere and that means that we can be cut off from this access. So I think it's particularly critical. Finally, there is an additional issue which is not an issue that we really deal with directly because we don't have a significant consumer activity, is that artificial intelligence that generates content shapes cultural representations, shapes the language, also shapes ethics. That is to say, the models we make, they have a certain policy, they activate certain things, they will write a certain type of code, etc. And so all of this becomes a mediation of information and it becomes a mediation of action. And so if we have to import models from elsewhere and if we don't have an alternative solution in Europe, well, we will be forced to use the biases and choices made by others, the United States and China in particular. So a few recommendations to finish. You say 'Artificial intelligence offers the hope of reshuffling the cards.' That's true, it comes back to your first question. One of the assets of Europe and France in particular is its talent base. It's an asset, it's an entry into the system. That is to say, if you don't have large companies, innovative companies that can attract these talents and convince them to work for us, well, they will simply go elsewhere and bring their value elsewhere. That's what we've seen a lot, and I think it's a thing that is starting to reverse a bit. We must benefit from it. Beyond talent, I mean, Europe doesn't have many assets. We have heavier regulation, we have a fragmented market. Obviously, a fragmented market means more difficulties to sell, but it's also longer-term partnerships. So that, we can say it's an asset. There is a very strong cultural diversity asset in language, etc., which is a real diversity for artificial intelligence. What I can say is that fundamentally we only exist if there is a market. And the problem we see is that this market materializes a little too slowly compared to what it should be. That is to say, companies don't yet fully realize the amount of artificial intelligence they will use in the coming years. And in fact, the consequence of this is that those who can invest before demand are rather the Americans. And so for example, there is an energy surplus in France of about 9 GW on average, well, this energy surplus will be captured rather by the companies that are the most aggressive from a capitalist point of view and who will be able to put money on the table, transform these electrons into tokens before others. And so for us, our issue, we manage to be quite aggressive but we would like to be more aggressive and for that we need visibility on the market. This market is obviously the private market, so companies that choose to work with us for 5 years, for 7 years and we have plenty. But it's also public demand which forms nearly 50% of Europe's GDP. And so being able to ensure that this public demand trickles down the entire value chain starting from where we are strong, where we have value-added which is artificial intelligence, and doing it in a concentrated way, that is to say not deploying it with subsidies, simply making demand at the end of the chain on digital services. It's an essential asset. It's an asset that our American and Chinese partners use massively. It's an asset that we have always been afraid to use in Europe. It must absolutely stop. And finally, in general, we seek to structure markets and ensure a relatively unified demand. So unified markets are useful. Unified capital markets are useful. We never do enough. And then finally, from a purely infrastructural point of view, France indeed has this chance of having surplus electricity. We must continue to do so because artificial intelligence, we can consider that it's about 1 kW per person in 5 years. 1 kW per person in 50 years, that means 40 GW to build in France, that means about 400 GW to build in Europe. That means about 2 trillion in investments to be made. So we really need to, in fact I made a mistake in my figures, sorry, 20 trillion, it's 20 trillion in deployment. So that means we need to be aware of it and we need to build electricity but above all we need to ensure that electrons are transformed into tokens with European companies. And then to build 400 GW in Europe, we need to accelerate projects. So that means we need permits more quickly. We also need a notion of reservation of energy capacities. There. And then on your second question, I think I've addressed it a bit. Beyond AI, digital services. So yes, you can do database managers, you can do hosting, you can do virtual machines, etc. Those are services with extremely low margins. You can't build a business at scale starting from a place where the margin is low. If you want to build a business at scale, you start from the place where the margin is important and then once you are big enough, you use a part of your R&D but you don't even need to be big to go create all the services you use. If you start from the place where the margin is low and you don't have volume, if you have neither volume nor margin, you have no chance of reaching scale. So either you are in a place with low volume, high margin, or you have low margin but high volume. That was the Chinese strategy and you move up the value chain. But if you are in a place where you have a high margin and not much volume at the beginning, you can only move down the value chain. So I think that's really important to take into account. You must start from the high-level services, go back down to low-level services because those have less margin and that's how we approach the problem from a strategic point of view.

It's the economic unit of artificial intelligence. It's something that is billed in millions of tokens. 1 million of tokens on one of our models for example, it will be 1 euro. So you send me 1 million tokens, I generate 1 million tokens for you. You will pay me 1 million euros for the input. 3, well 1 euro for the input, 3 euros for the output. So it's... And a token, what is a token? It's a few letters in a word. It's a number, essentially. It can also be a few pixels in an image when it's input. So we manage all multimodal data. It can be a piece of sound as well. So that's the unit. Our economic model is partially a hosted model, so consumption-based. That is to say you use technology, you host technology with us and you pay per token, you pay per unit of time in CPU. So when you do something other than artificial intelligence, you pay for storage. We also have a model that is more of a platform model where we bring our software, deploy it on client infrastructure which then pays for its infrastructure, its GPUs, its CPUs, its storage, and on that it's more of a license model on which we have a service layer to build value and transform companies.

The token is the client unit of the digital service. So when you buy a digital service, you buy tokens. The unit for us and the way we finance our projects, etc., is rather the MW. The MW is the electricity we will transfer into intelligence. 1 MW, so I'll give you in GW because I know the figures well in gigawatts. When you invest and create a data center, you put graphics cards that will allow generating tokens when you host the model. To create 1 GW, you will invest 50 billion dollars over 5 years. So it costs you about 10 billion per year. With that, well, I'll give you a bit of the margins. With 10 billion in data center per year, you can make about 20 billion in revenue in tokens. You can sell 20 billion tokens if you do things correctly. If you are about at scale, 1 GW is about at scale. So that's it, that is to say you have 50% gross margin between the digital service and the investments that the digital service provider makes. So it's a significant margin, it's a margin that allows investing and doing R&D. In particular, it allows financing what costs extremely much in artificial intelligence which is the training of models because we use ourselves GPUs, we use today tens of MW to train models. So those are costs only for us that we bear. So that's about what we can keep in mind, 1 GW. I mentioned that if you have even just 10% of the European payroll which becomes artificial intelligence, we're talking about approximately 400 MW because we're talking about 10,000 euros per year and 10,000 euros per year rents you about 1 kW of GPU. A GPU is 2 kW. So you rent half a GPU with that. So finally you are in a situation where to build this, you need to make 20 trillion in investments. It's 20 trillion in investment, it will bring back, it's about 8 trillion per year if you want of what can go back to, no, sorry, you have 20 trillion in investment and it's about, yes, 8 trillion per year in revenues that can go back to the United States. So we are on scales that are quite significant if you want. And what you need to take into account is that, there, there is this 50% gross margin between the service provider and the one who makes the service. In fact, in the chain that transforms the electron into token, there is about 10% that goes to the electron. So if your only role, the only role of Europe is to be the energy provider, that means 90% of the value is elsewhere. And so it leaves Europe and will be invested, recycled in R&D, used as a vector of power by other actors. So it must be taken into account.

Very well, I will take your questions in order. So on the ethical subject, well, we make systems that are systems that help humans be more efficient, that will automate a certain number of processes while keeping humans in the loop on important decisions. And so, we have a whole team of designers, about ten designers with us, who will reflect on human-machine interactions, ensuring that when artificial intelligence summarizes information to help with decision-making, it provides enough elements so that the human remains informed. Now, I'm not telling you it's an easy subject because in fact, you still have the issue that humans tend to be a bit lazy and so if the thing does it for them, they think less and less. So in fact, there is still this issue of managing to keep them in the loop, you have to stimulate the human when they are facing the machine. And so, these are obviously subjects we reflect on. Then, you have a sub-part of ethics which is the subject of engagement in defense. So we work with the Ministry of the Armed Forces, we work also with certain allies of France and it happens that this technology is considered a dual-use technology and therefore fundamentally it is regulated in terms of exports, in terms of export control and so we have the whole framework of the regulation on export control that applies to us and so that allows us to have, we don't claim to have an ethical framework, we don't claim to have the legitimacy to have a sufficient ethical framework, we don't have democratic legitimacy as a company. On the other hand, we fit into frameworks that have democratic legitimacy and that allows us to work correctly and to not do what we don't have the right to do. Then, when we work with the Ministry of the Armed Forces in France, we don't claim, unlike other actors, to have a right of oversight over end use. Because we don't claim to have the legitimacy to explain to the French armed forces what they can do, what they cannot do with the technology. We can tell them and we can do it by project, we can explain the reliability of things, etc. So, but then, who is sovereign finally in the use of technology? The army has a democratic legitimacy that we don't have. And so in that capacity, we have a duty to advise, we explain how it works and I can tell you that today it is used in a perfectly reasonable manner and within the limits of what artificial intelligence can do. We must be aware that artificial intelligence allows for defensive capabilities, it allows...

It allows having much more operational armies, obviously that improves operations and so on. It allows doing defensive work but it's also absolutely critical in deterrence. That is to say, you have armies like the Russian army that use artificial intelligence in their drones massively. If you're not capable of having counters that are themselves activated by artificial intelligence, well you don't have sufficient deterrence. And so the deployment of artificial intelligence in autonomous systems in particular, I think you absolutely must not cut back on that. That is to say, it's absolutely indispensable for having conventional deterrence today. That also means you need to master artificial intelligence end to end. That's on the first subject.

Then on the subject of cyber defense and cyber attack, we have one of our American competitors who knows very well how to do fear marketing if you will, but the subject that models are excellent programmers and are absolutely capable of orchestrating attacks, discovering vulnerabilities and proposing exploits, these are things that have been ramping up for six months, it's rising linearly, predictably, it's rising everywhere at the same time. That is to say, you have another American competitor who has a model that does exactly the same thing. You have Chinese actors who do it too and our models are capable of discovering all the vulnerabilities that are reported by Mythos for example. So for us, that means we work with our clients to help them on these cyber subjects. It's obviously a sovereign subject and so it's yet another argument for saying that you need to have control over this technology. You can't have the French army's code bases scanned by Mythos. I think there's an irremediable dependency being created that absolutely must find solutions.

Then, on the subject of time. On the subject of time, I would say we don't have time. I think notably because it's linked to physical constraints. We're talking about natural resources. Here we're talking about using natural resources to transform electricity into tokens. There isn't infinite electricity in the world. The United States are saturating their electricity to transform it into artificial intelligence. They're deploying a trillion dollars next year. When they deploy a trillion dollars, that means they want two trillion dollars in return. So they can see there's an immense market being created and they can see that whoever possesses the resource, whoever supplies the chips, whoever possesses the electrons, whoever possesses the power plants or who has access to energy on a massive scale, that's who wins.

So that means today you have, well it's 9 GW, it's 90 TW per year that are available in France. Well, there are lots of things we can do with it, partly we'll do artificial intelligence. You have actors with enormous balance sheets who have plenty of money to put on the table before seeing demand to transform electrons into intelligence. But they're going to do it in the next two years. Once they've done it, well you don't have electricity anymore. You have to wait for power plants, create new ones and so on. So there's a form of monopolization of the European energy resource that is underway, that is accelerating at a crazy speed, and of which we're not aware, and it's truly irremediable. That is to say, we already have a demand problem, because the Americans arrived before on the cloud, they provide all the European industry, they provide the French state, obviously the cloud market share, and artificial intelligence tends to compress for non-American actors. Not only do they grow with the market but on top of that their share increases.

So I think this is something that can be reversed with artificial intelligence and what we're starting to see with our clients, with our revenue, with our growth. But the problem is that even if you solve the demand problem, you have a supply problem because once the supply is monopolized by American actors, well suddenly we no longer have supply and we can no longer transform electrons into tokens. And this is playing out in the next two years. That is to say, what needs to be put on the table today in terms of investment is tens, hundreds of billions of dollars. And to do that, you need visibility on the market. And market visibility must be indicated by European actors, by European states. Today, it's not indicated massively enough for us to take the risk of deploying 100 billion dollars. So I think we need to become aware of this urgency and unfortunately, I don't see very well how we'll become aware of it. So we're going to manage to build, let's say 1 GW by 2029. Those are our objectives. It's not enough. We should build more. But for that, we need a form of market visibility and demand visibility that isn't what we have today.

And finally perhaps on the last question, you asked a question about distillation. What is distillation? You take a large model and you use that model to make a smaller one that will be more efficient. So we use it massively because we train large models, it costs us a lot. Then we use our large models to train small models that we then serve to our clients because it costs them less to serve. So what this means is that it's absolutely indispensable. If you want to know what's in your models and control them, you need to be capable of building the largest models. So that means you need to know how to train them, which means you need to have a lot of computing capacity of your own, which is pure R&D that allows training them. So does it allow catching up? No, it's a technology that essentially reduces costs internally and so we take large models and use them internally to make small ones, and we need the large models, which means we need to invest a lot in them.

So I'm going to answer, well I can explain a bit our role in Campus IA which is very minor, that is to say we have a very small participation. We took a participation because in the value chain we're capable of building clusters, so we buy the servers, we plug them in. That is to say, it looks like physical racks that will be put in buildings. But we don't build the buildings. We don't build the buildings, we don't build the transformers, we don't build the cooling, we'll rather seek partnerships to be able to do that. So in that capacity, we need data center developers. So we took a very, very minority position in this project and above all it's a supplier, that is to say it will become a supplier for us for part of the construction. Another part will be used to supply hyperscalers because you start from a situation where 80% of European digital services are actually imported from hyperscalers. And so naturally the starting point, if you will, if you start building infrastructure and you look at infinite demand, you end up with 80% hyperscaler with I think the objective that France and Europe must have, which is to ensure that this proportion decreases over time. So that's why Campus IA today will serve us in part and will certainly serve hyperscalers in part. So that's what I can say. As for the name, I didn't choose it. On a number of technical aspects, I don't know. In any case what's sure is that a portion of the revenue is indeed redirected to public research.

Then on the subject of capital, you mentioned, well that's the real infrastructure, that is to say you can build a data center, you lease it, that gives you an equity return and it's essentially a pretty good business. If you have a lot of funds, deploying them in data centers today, knowing that everyone wants data centers, gives you a fairly significant return, and so there are plenty of actors who want to do that. When you have a site of that size, you need actors with solid balance sheets, that is to say they can deploy tens of billions of euros. And so on that, naturally, because we forgot to create pension funds in France for example, we turn to foreign financiers. Because there's no capital market. European capital markets lack providers of solutions for this type of investment. So we go look elsewhere and I think we should benefit from it. That is to say, foreign investments, if they're under control, if the governance is good on the investments being made, and here I think today on this subject, BPI France, I believe you may have questioned them, plays a role and has a seat on the board and so on. Well these foreign investments are beneficial because they allow us to accelerate because obviously we lose a certain number of economies because when foreign capital arrives, they take a return, so we lose the return but on the other hand we build something so we can derive something from it. What we propose is to capture the margin from operations of servers up to digital services. So that's as much margin as we can reinvest in R&D. We're not capable today of a completely integrated model where we operate from the ground up to digital services. Maybe that will be the case one day. But there's a sharing of value. We're capable of creating quite a bit of value with this. We'll be able to reinvest it in R&D. Obviously, there are foreign investors who will recoup their investment somewhere. But on the other hand, without these foreign investors, we couldn't build. So you always need to think of sovereignty as a path. That is to say, you start from a state of the economy and balances in the value chain and you say this state isn't satisfactory. You need to change this state, it will take time. And how do you change this state? Well you need to put resources in the right place and then it will change gradually.

On environmental subjects, well obviously all of this costs money. There are externalities. If you can trace back the value chain to semiconductors, you need a lot of inputs, you need rare earths, you need land, although I think that's not the main issue because it's actually super dense. So 1 GW fits on about a hundred hectares. So we don't really have a land issue, I'd say from a point of view, well of course if you look at the city level, if you look at the European level, there's no issue. So it's not the main issue in any case. Then you have an electron issue. That's a real issue. The interest of building in France for us in particular is that 70% of electrons are nuclear. So that means the carbon footprint of the electron is largely reduced. And so building in France is fundamentally better than building elsewhere because tokens, people need them. So if someone builds 1 GW of capacity in France, they won't build 1 GW of capacity in Texas and ultimately the atmosphere will warm a little less. So I think you need to think about it from that perspective.

Then you have other inputs, you have water in particular. There has been a lot of technological progress to avoid waste heat and ensure that cooling is more efficient and that the majority of electrons are used in computation rather than in everything around it. And then you have biodiversity issues obviously, which seem, from having observed the sites, everything has been done and implemented to preserve those aspects. So for us, it's important. Well obviously it's sure that from an ecological point of view, it would be better to have a forest but from an economic point of view it still has a role. So it has an ecological effect. You need to think about artificial intelligence within ecology if you will. Once again, we transform energy into electrons. It's a sort of natural resource of electrons into tokens. Consider the token as a sort of natural resource. Natural resource, you need to sort of build a mine. The mine is the data center. And what I think we need to do, because if we don't, we have additional issues, and I think we need to do it. The more we do it ourselves, the more we have a voice in the matter on how we think about externalities. That is to say, if we don't do it ourselves, we'll just import what's proposed to us from outside. And I can tell you that outside, they think about biodiversity and climate change quite differently from how we think about it. So if we want to have a voice in the matter, in any case, we need to be capable of internalizing production. And that's absolutely essential. And so there are obviously compromises to be made. In the same way that we accepted building lithium mines in France, well we need to accept building data centers. But what would be a shame is building data centers where the bulk of the economic value completely escapes France. That's the issue. So rather than taking on these externalities, let's ensure they can be recycled into R&D and beneficial aspects for France.

I'll just add that today our new generation of models, we train them in France and I think we have the largest computing cluster in France. So it's a source of pride for us because from a carbon footprint perspective in particular, it's largely reduced compared to what we could do before when we weren't in France. So that's what I can say. Finally, on the subject of calculating carbon footprints and so on. Unlike some of our competitors, we think you need a life cycle approach and not a compensation approach because obviously it's a bit too easy when you have very large balance sheets to do compensation, when in fact you need to think in an integrated way about everything that goes into projects and have transparency per project rather than having global transparency balanced through quotas.

Of course. On the subject, well I think I'll take the first question which is about regulatory heaviness and fragmented market subjects which are somewhat the same thing. One of the regulatory burdens, well one of the regulatory problems in Europe, is that fundamentally every time you enter a European country, you need to open a new entity, understand a new stock option regime, understand a labor regulation that is different. And so that means you'll quickly need a team on the ground. That is to say, I've signed hundreds of documents to open entities in about ten European countries. I've opened dozens of bank accounts because generally you need an associated bank account. You have a whole set of regulations that are national regulations that aren't unified. You have taxation obviously which isn't unified when it absolutely should be if you want to have employee incentive systems that make sense. So we try to patch things, we try to do things as fairly as possible but we end up with fiscal situations that are frankly disparate between countries. So that's relatively catastrophic. So you can't have a unified market if you don't have a form of unified taxation, a form of unified social law and so on. So that's the first thing, I'd say that's the main burden.

Then you have a whole stack of regulations that are more or less consistent with each other and that force us to document a lot of things. So you have the GDPR, you have all the laws on copyright and text and data mining. You have the AI Act which isn't yet implemented but will be from August. All these things touch on fairly similar things generally. It touches on data, personal data, what's on the internet and it doesn't say exactly the same thing. It requires different documentation. It's implemented by 27 different implementation bodies in Europe that are more or less zealous depending on the countries. And so we end up, I end up with a compliance team of five people, which is today manageable for us because we're big enough, but if you're a new entrepreneur in France and you look at all this and then you look at the atmosphere and the fact that people rejoice about having put regulation back on the AI Act, that was early 2024, and you say I'm going to go to the United States instead, and that's what's happening. They go to the United States.

So we've overcome this aspect because we're big enough. It goes into overhead costs, it remains very painful, but it's especially that you lose lots of entrepreneurs who go elsewhere because they're afraid. You have investors who say, 'No but Europe has already lost, too much regulation.' And you have a whole narrative that is widely used by the United States to say Europe loses because anyway it's bureaucrats who are regulating in Brussels because they don't know how to do anything, they don't know how to innovate. And in fact, it's a destructive narrative because it's also internalized by Europeans. So it's a form of colonialism if you will. But the issue is that you need to reverse this narrative. To reverse this narrative, you also need to think about having regulations that are simpler, more unified, that allow us to go faster. So that's on that subject. I spoke of market fragmentation as an asset, it's not an asset, it's rather a hindrance. That means you need different teams selling differently everywhere. That means you also have about 60 telecom operators in Europe. You have three in the United States. That means they have 20 times more budget globally. So when they make an artificial intelligence investment, they can put a lot of money on the table. And that's how you end up with American technology. Startups, the first thing they do, they sell to other startups because that recycles venture capital money, it circulates a bit to see who will win, and then they sell to American companies that are enormous and buy very quickly, and that's how they scale. Once they've scaled, they export to Europe where since nobody emerged in time, well they take the whole market, they buy up the few companies that know how to do things and they end up consolidated, they have the whole market. That's how it works.

In Europe, we don't have that because you have much greater viscosity because companies are smaller and on top of that you need to speak 20 different languages, manage 27 different regulations. So it's rather not an opportunity. However, once you've done it, well it's true that European companies are more partnership-oriented, they have approaches that are more constructive. Sometimes they're champions, we're lucky to have lots of industrial champions in particular, and so those are good partners for us and since these are partnership approaches, it's difficult for Americans to replicate them. And so in a way, it's revenue that is relatively defensible. So the viscosity plays against you at the beginning, but then it plays in your favor when someone attacks you, when a competitor tries to take the market if you will. But there you have it, it's rather a disadvantage. So you'd prefer the market to be much less fragmented.

Then, last point on regulation as defense and as a way to counter the Americans. I think often the intention is good, that is to say to ensure it favors our European startups. I think if we look at the whole history of what we've done, that has never been the case. And a very simple reason is that regulation fundamentally puts an overhead on you that to overcome, you need to be big enough. And so if you don't have companies big enough with a regulatory burden that big, you end up being carried by American actors. So regulation favors the big ones and that's all the more true in technology. And so any discourse about how regulation can defend us, I think it's quite, it's a bit like the devil, the road to hell is paved with good intentions. I'd say there is one place though where we should be much better, and that's in ensuring public demand is recycled on European territory. That is to say, used to do R&D on European territory. That applies to defense, it applies to the entire public sector. I think there's an awareness that isn't necessarily a regulatory awareness but rather a political awareness that this is the thing to do, that this is what the United States have been doing since the 1940s and it worked very well for them. And so, should we regulate that? Maybe we can manage this aspect a bit. You need to do it while maintaining very high competition in Europe. You need to be a planner on what the state spends but not a planner on how things are used. And that way we'll get there. But regulating to defend generally doesn't work.

Yes, but on the other hand, adding additional rules...

respecter en se disant que les Européens vont s'en sortir mieux, c'est faux. C'est que vous avez des règles, vous avez forcément du lobby à mettre en œuvre pour que l'implémentation des règles soit jouable. Et en général, bah vous avez les acteurs américains qui sont énormes mais ils ont beaucoup plus de lobbyistes à Bruxelles que nous quoi. Donc c'est eux qui finissent par mettre à les rendre les plus favorables pour moi quoi.