The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch

Mistral CEO Arthur Mensch on Sovereign AI, Inference Economics, and Why Energy Isn't the Bottleneck

50 min Listen to the original Episode page
Sovereign AIMistral AIInference MarketAI InfrastructureEuropean TechEnterprise AI

Summary

Mistral CEO Arthur Mensch argues that sovereign AI is critical for Europe's economy, positioning Mistral as a key player in model training and infrastructure. He claims energy costs are only ~10% of compute TCO, with chips being the primary bottleneck, predicting energy constraints will hit around 2029-2030. Mensch discusses Mistral's vertical integration strategy, noting 80% of token traffic comes from their own models, and highlights the 'data wall' challenge where sourcing high-quality data is harder than scaling compute. He also addresses the inference market's bifurcation between commodity and premium enterprise pricing, and Mistral's focus on four enterprise verticals to avoid consumer market risks.

Key takeaways

  1. 1
    Mistral expects significant growth in CPU-related tasks, sandboxes, and storage as AI agents perform more background, CPU-intensive work.

    Unidentified speaker (speaker_0) · 29:20

Episode sections

  1. 0:00 – 8:50

    Sovereign AI and Mistral's Role

    Arthur Mensch argues that sovereign AI is critical for Europe's economic growth, citing structural trade balance issues rather than just political factors. He positions Mistral as a key player in providing this sovereignty through model training, data control, and infrastructure.

  2. 8:50 – 14:41

    Safety of Open Source Models

    Mensch addresses concerns about Chinese open-source models, stating that while they offer choice, they lack guaranteed safety due to unknown data supply chains. He notes that 80% of Mistral's token traffic comes from their own models, which they can monitor and secure.

  3. 14:41 – 24:05

    Energy vs. Compute Bottlenecks

    Mensch claims energy costs are only ~10% of compute TCO, making chips and capital the primary bottlenecks. He predicts energy constraints will become critical around 2029-2030, requiring state intervention in nuclear and solar planning.

  4. 24:05 – 30:25

    Mistral's Infrastructure Strategy

    Mistral is vertically integrating by building its own infrastructure to serve tokens and GPUs, avoiding chip manufacturing. They are expanding into CPU-intensive runtime tasks, sandboxes, and storage as AI agents become more complex.

  5. 30:25 – 31:45

    Inference Market Dynamics

    The speaker describes the inference market as having infinite demand, behaving like a commodity for on-demand users while offering premium, guaranteed service levels for mature enterprise customers.

  6. 31:45 – 33:10

    Neo Lab Survival and Mistral's Strategy

    Success for new AI labs depends on balancing top-tier research with a customer-centric enterprise motion, focusing on scalable markets rather than just hiring researchers.

  7. 33:10 – 34:21

    Compute Bottleneck for New Entrants

    While Europe has abundant talent, the primary bottleneck for new AI labs is accessing GPUs at a reasonable price due to a severe supply crunch.

  8. 34:21 – 35:29

    Mistral's Model Success and Team Resilience

    Mistral's Large 4 model leads in open-weight coding and cyber defense, a result of team resilience and the ability to course-correct when model training fails initially.

  9. 35:29 – 36:20

    Data Walls and Training Run Realities

    The speaker disputes the immediacy of $100 billion training runs, citing a 'data wall' where sourcing high-quality, domain-specific data is harder than scaling compute.

  10. 36:20 – 37:30

    AI Stock Decline vs. Compute Demand

    Despite compute scarcity, Mistral remains financially stable due to vertical integration and revenue-backed collateral, avoiding the excessive debt leverage seen in other AI stocks.

  11. 37:30 – 41:22

    Strategic Focus on Enterprise Verticals

    Mistral avoided the consumer market and chip design to focus on four enterprise verticals: manufacturing, financial services, defense, and the public sector.

  12. 41:22 – 49:37

    Economic Growth via Physical Effects

    AI drives GDP growth by enabling better physical products (e.g., efficient planes, safer cars) rather than just providing intelligence, with the main friction being ambiguous enterprise problems.

Assets discussed

AssetKindStanceContext
OpenAIprivate companyneutralthink we'll end up in a world where if you want to do mathematics, you need to work at OpenAI. Luckily, there is enough people in the world that don't want this to occur.

“Stance” is how the speakers talked about the asset, not a recommendation.

Claims worth verifying

Mistral is a $21 billion company and is described as Europe's last hope for frontier AI.

Unidentified speaker (speaker_1) ·0:15

How to verify: Check recent press releases, funding announcements, or valuation reports from Mistral or financial news outlets. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Arthur Mensch stated that leaving frontier AI to the US and China will cost Europe its place in the world.

Unidentified speaker (speaker_1) ·4:09

How to verify: Search for transcripts or news articles quoting Arthur Mensch on this specific statement. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Mistral Large 4 leads open-weight models in agency coding and cybersecurity.

Unidentified speaker (speaker_0) ·7:19

How to verify: Check benchmark results from independent testing labs (e.g., LMSYS Chatbot Arena) or Mistral's official documentation. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

80% of Mistral's token traffic comes from their own models.

Unidentified speaker (UNKNOWN) ·11:02

How to verify: Request internal metrics from Mistral or analyze API usage data if publicly available. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Energy will become a bottleneck for AI infrastructure in Europe around 2029-2030.

Unidentified speaker (speaker_0) ·16:27

How to verify: Monitor European energy grid expansion plans and AI data center construction timelines. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Mistral's revenue base is 60% European, with the rest from the US and APAC.

Unidentified speaker (speaker_0) ·22:46

How to verify: Check Mistral's investor presentations or annual reports for geographic revenue breakdown. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Primary source

Mistral spent 18 months setting up its infrastructure to serve GPUs and tokens.

Unidentified speaker (speaker_0) ·24:39

How to verify: Review Mistral's technical blog posts or infrastructure announcements regarding their data sovereignty measures. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Mistral is partnering with Accenture and TCS for scaling AI implementations.

Unidentified speaker (speaker_0) ·26:40

How to verify: Check press releases or partnership announcements from Mistral, Accenture, and TCS. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Primary source

Mistral expects significant growth in CPU-related tasks, sandboxes, and storage due to AI agents.

Unidentified speaker (speaker_0) ·29:20

How to verify: Analyze Mistral's product roadmap or technical whitepapers on agent-based workloads. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Mistral has raised approximately $7.5BN, including debt, and reached a valuation above $24BN.

Unidentified speaker (speaker_1) ·30:25

How to verify: Check Mistral's official press releases or funding announcements from 2026-2026; verify valuation via private market data providers like Preqin or PitchBook. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

There are 102 new AI lab providers created, with another deep mind spin out every week.

Unidentified speaker (speaker_1) ·31:45

How to verify: This is a rhetorical exaggeration by the speaker; verify by searching for recent DeepMind spin-offs and new AI startup announcements in 2026-2026. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Mistral's Large 4 model is leading the pack in the open weight category for coding and cyber defense.

Unidentified speaker (speaker_1) ·34:21

How to verify: Check independent benchmarks such as Hugging Face Open LLM Leaderboard, SWE-bench, or cyber defense-specific evaluations published in 2026-2026. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Mistral posted a tech report for Large 4 recently.

Unidentified speaker (speaker_0) ·34:46

How to verify: Search Mistral's official blog or arXiv for 'Mistral Large 4 technical report' published in 2026-2026. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Nabius Nebius and CoreWeb stocks are down 40% due to the AI trade cracking.

Unidentified speaker (speaker_1) ·36:20

How to verify: Check stock performance data for Nebius (if publicly traded or via private market valuation changes) and CoreWeb (verify ticker and exchange) on financial platforms like Bloomberg or Yahoo Finance for late 2026-2026. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Mistral operates as a vertically integrated company, building compute for higher-level enterprise purposes.

Unidentified speaker (speaker_0) ·36:34

How to verify: Review Mistral's corporate structure, partnerships with cloud providers, and any announcements regarding in-house hardware or dedicated infrastructure. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Mistral focuses on four verticals: manufacturing, financial services, defense, and public sector.

Unidentified speaker (UNKNOWN) ·37:51

How to verify: Check Mistral's website, case studies, or press releases detailing their enterprise solutions and target industries. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Mistral acquired a team called Amy specialist in AI simulations for physics.

Unidentified speaker (speaker_0) ·45:12

How to verify: Search for press releases or news articles about Mistral's acquisition of a team or company named 'Amy' or similar, focusing on physics simulations. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Laradin is used by Gainsight, Vertiv, and SurveyMonkey.

Unidentified speaker (speaker_1) ·47:52

How to verify: Check Laradin's customer list or press releases; verify these companies' AI tooling vendors via their own procurement disclosures or public partnerships. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Shopify's Shop skill allows AI agents to search millions of stores and complete purchases.

Unidentified speaker (speaker_1) ·48:32

How to verify: Test the Shopify Shop skill integration with an AI agent or review Shopify's developer documentation and API capabilities for autonomous purchasing. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify Mistral's current valuation against recent funding rounds or press releases.
  • Check Mistral's public API documentation or traffic reports to validate the 80% self-traffic claim.
  • Review industry reports on AI infrastructure TCO to compare the 10% energy cost assertion against competitors.
  • Monitor European energy grid expansion plans and nuclear/solar project timelines for the 2029-2030 bottleneck prediction.
  • Analyze Mistral's geographic revenue breakdown in investor presentations or annual reports.
  • Check partnership announcements from Mistral, Accenture, and TCS to verify system integrator collaborations.
  • Review Mistral's product roadmap or technical whitepapers for evidence of CPU-intensive agent workload growth.
  • Investigate the 'data wall' claims by comparing public data availability with industry reports on high-quality domain-specific datasets.
Full transcriptexpand
0:00
Unidentified speaker (speaker_0): Sovereign AI is going to be the most important category. So the cost of energy, at the end of the day, is not the one thing that matters in the economic of a project. It's pretty clear that there is not enough compute. I mean, the inference market behaves very much like a commodity market, which is pretty high margin today.
0:15
Unidentified speaker (speaker_1): He's our last hope. He is the last hope in Europe for us to build a frontier AI lab. Yep, you guessed it. I have Arthur Mensch, co co-founder founder and CEO of Mistral, Mr. Al on the show today. It is now a $21 billion company. It is the company that is keeping Europe relevant in the race for frontier models. This is Arthur, like you've never seen him
0:35
Unidentified speaker (speaker_1): before. He's unleashed, he's funny, and he says some incredible things that I have to say I was not expecting. This was such a joy to do, and I'm so excited to welcome Arthur to the show. But before we dive into the show today, base 44 is one of those products I used, and I just thought straight away, man, I wish I'd had this when I started 20
0:55
Unidentified speaker (speaker_1): VC with $50 and no contacts in Venture. I wanted to see what it could do. So I built a guest research hub for the show. Upcoming guests, company notes, previous interviews, sponsor contacts, follow-ups, all in one amazing location. I described it in plain English. A few minutes later, I had the product open in front of me. Database log login,
1:15
Unidentified speaker (speaker_1): backend, and hosting already handled and just all in one. It made it so easy. And that's what I loved. I didn't write a long spec. I didn't stitch five different tools together. I built the thing and I started using it. For any founder listening, that's the point. The faster you turn an idea into something real, the faster you launch. Apps, dashboards, internal systems, AI agents,
1:35
Unidentified speaker (speaker_1): whatever you need to build, base 44 gets you there. Go to base 44. base44.com forward slash 20vc. That's base44.com slash 20vc. While base44 helps you build faster, Larridin helps you measure AI impact. Every company now runs two workforces: humans and AI agents.
1:55
Unidentified speaker (speaker_1): Larridin is the first unified dashboard across both workforces that measures their work and optimizes their impact. Larridin Larradin answered answers four questions. Who's using AI? Which tools? What is it costing? And was it actually being produced for every team across every vendor? AI ROI stops being a
2:14
Unidentified speaker (speaker_1): guess and becomes a number that you can consistently and confidently measure. Gainsight, Vertiv, SurveyMonkey, and many other enterprise organizations run their AI programs on Larridin. Laradin. Thousands of AI tools. and millions of dollars of AI spend are measured. You bet on AI, Lauradin proves that it's working. Visit Lauridin.com
2:34
Unidentified speaker (speaker_1): to learn more. That's L-A-R-R-I-D-I-N.com. While Lauradin measures AI impact, SHOP helps you buy smarter. My agent will summarize a board deck, researcher founder, and draft my notes. Ask it to buy something though, and it just falls apart. Wrong size, dead links, stuck at a login, and I end up doing
2:54
Unidentified speaker (speaker_1): it myself. That's why the shop skill from Shopify is supporting today's show. Head over to shop.md, copy the prompt, paste it into whatever agent you already use, and that's it. Your agent can now search millions of stores on Shopify, over a billion products. Sign in with Shop and it knows your preferences and order history. So the picks
3:14
Unidentified speaker (speaker_1): pics actually look like you, and nothing is bought until you approve it in shop. I've installed the shop skill with my agent, and I love what it found for me. So I needed protein powder in a specific flavor under a specific budget, I didn't want to spend too much, and I wanted it before the weekend. So I just typed that into my agent, exactly how I'd say it. Before it
3:34
Unidentified speaker (speaker_1): would have been stuck at checkout. This time though, it came back with exactly what I wanted. My details were already filled in. I hit approve and that was it. Turn your agent into a personal shopper with the shop skill. Visit shop.md slash 20 minute vc. That's shop.md slash 20 minute vc.
3:52
Unidentified speaker (speaker_2): You have now arrived at your destination.
3:55
Unidentified speaker (speaker_1): Arthur, dude, I am so excited for this. I have interviewed some of our favorite counterparts in the US. You are our remaining hope in Europe. No pressure, but thank you so much for joining me today, dude. Thank
4:06
Unidentified speaker (speaker_0): you, Harry, for having me today.
4:07
Unidentified speaker (speaker_1): Now
4:09
Unidentified speaker (speaker_1): I was doing some research and I say some quite opinionated things, Arthur, but it seems I'm not the only one because you the other day said that leaving frontier AI to the US and China will end up costing Europe what remains of its place in the world. I
4:25
Unidentified speaker (speaker_0): did say that, and I truly believe that this is such an important technology that it needs to be distributed in the world. We're talking about a technology that is comparable to electricity, a technology that is falling more and more under export control. And so AI is going to be a big enabler of economic growth. And if there's no
4:45
Unidentified speaker (speaker_0): AI being produced in Europe in particular, Europe will struggle to actually produce the growth it needs to produce.
4:51
Unidentified speaker (speaker_1): I'm going direct. To what extent does the success of Mistral depend on US overreach? Yeah, obviously, when we have a Trump administration, we see things like Fable being banned and it it causes a lot of questions and the need for sovereignty. In a world where there's not US overreach, does that threaten sovereign providers?
5:09
Unidentified speaker (speaker_0): I'd say this is structural. This is not really linked to any way of operating or of showing power. It's more linked to the nature of the technology itself. If you start as an enterprise or as as a state to use AI in all of your core processes, it means that you want to make sure that the money you're spending on those processes
5:28
Unidentified speaker (speaker_0): actually accrues pretty close to you. If you don't, it means you have a trade balance problem. It means you have enormous flows of cash that are going from one place in the world, it can be Europe, to another place. in the Europe in the world, which can be the US. And so there is a macroeconomic problem to not being able to produce artificial intelligence, to do RD in Europe territory.
5:48
Unidentified speaker (speaker_0): That is not linked to a certain way of saying that uh Europe is falling behind, or a certain way of using soft power, let's say. So it's structural. I won't say that uh some people are not making it a little bit emotional, but when we start started Mistral in 2023, we started with this conviction that this was a sovereign technology, in that
6:08
Unidentified speaker (speaker_0): as a company, as a state, wherever you are, whether in the US, in Europe, or in Asia, you need to own the technology. You need to be able to customize the systems, to run them very close to your data. You need to be able to control the cost. You need to work in true partnerships with companies that understand how to build AI systems. And sovereignty presents. precedes 2025
6:28
Unidentified speaker (speaker_0): election. So it turned out to be true what we said in 2023. It turned out to become quite emotional when we start to speak with Greenland, but it's mostly an economic problem. It's not really a political problem. It's just too important a technology.
6:41
Unidentified speaker (speaker_1): When I hear you speak, I think of Lynn at Fireworks, who came on the show and said actually very similar things. We need companies to have their own models to train them very close to their own data. How do you feel when you hear me say that? And is that the evolution of the company, allowing companies to own their own models and work with their own data?
6:59
Unidentified speaker (speaker_0): And what the category we define, and the reason why companies work with us is that we bring them sovereign AI technology. And so in sovereignty technology, you you need three things: you need to have a source of model that is secured. So you need a company that can train the models itself. So we have that model factor. factory. Mistral Large 4 is actually leading the open weight models in agency
7:19
Unidentified speaker (speaker_0): coding, it's leading in uh in cyber defense and cybersecurity.
7:23
Unidentified speaker (speaker_2): So
7:23
Unidentified speaker (speaker_0): that aspect of being able to constantly produce improvements on models and to be able to tell our customers that in two years from now you'll get better models because we are constantly training them and spending billions on it is super critical to anything sovereign. So it's the reason why Mr. Mistral has been created and it's it's the the main part where we're investing money. Now you also need
7:43
Unidentified speaker (speaker_0): to have control over your data. And so in general, you don't want that data to flow back to the AI providers. And so we've made sure at the technological level that our platform to run the agents, that our platform for customizing models actually sit on our customer tenant. That's super important in many cases. When work with
8:03
Unidentified speaker (speaker_0): manufacturing companies, they don't want their IP to flow back to a subprocessor. Now, that's not enough in general. Because AI is consuming a lot of energy, uh, and because when you move at scale, you actually need a lot of GPUs that as an enterprise you have difficulty to procure. You also need to have token generation services and you need to have token factories basically. And you need a
8:23
Unidentified speaker (speaker_0): combination of data that sits in your infrastructure and tokens that are produced in managed infrastructure at a lower cost. And so what we do as a company is do three things. We train the models, we deploy the platform and the data plane with our customers or on their on their tenant, and then we host what is energy, what is intense in energy, and
8:43
Unidentified speaker (speaker_0): that's the reason why we've been building our own infrastructure. So we have these three things, and those are the three ingredients that are necessary to serve sovereign AI workloads.
8:50
Unidentified speaker (speaker_1): Can I ask if we start on one? You said the model is secured. I have the world's best experts on the show, Arthur. I'm very lucky. But they tell me different things. Okay. Some people say, you know, the Chinese open source models, you download them, you run them locally. Hey, they're totally safe. And I'm like, wow, great. And then next week someone says, hey, but we have no idea whether there's prompt
9:09
Unidentified speaker (speaker_1): injections that you don't know, there's backdoors. back doors. Absolutely not. You cannot be sure that. they're safe. Help me out here, Arthur.
9:17
Unidentified speaker (UNKNOWN): Well,
9:17
Unidentified speaker (speaker_0): at Mistral, we we have a I mean, one of our businesses to really run the inference for open models. So we run ours, but we also run others. And
9:26
Unidentified speaker (speaker_1): why why why do you run others?
9:28
Unidentified speaker (speaker_0): Because customers want choice. And when you have infrastructure, you want to provide as much choice as possible to saturize the infrastructure. That is the reason why that brought us to actually serve more and more of uh open models. Because in different domains, you may want to pick a different model. You may want to pick smaller ones, you may want to pick bigger ones. So the more choice you have, the better usage you can make of the infrastructure,
9:48
Unidentified speaker (speaker_0): the lower your cost can be. And it's in a way it's a commodity market, so you want the cost to be as low as possible. So that's the reason why we serve those models. But what you say is true. Uh, models are compression of knowledge. Uh, if you don't control the data. supply chain, if you don't know what goes into the models, you cannot guarantee anything on the output. And
10:08
Unidentified speaker (speaker_0): so when we serve our models, we can give further guarantees to our customers than when we serve third-party models. Precisely because theoretically, you can make sure that in a certain environment with a certain context, the model behaves in a certain way. There's no way of excluding that, there's no way of measuring that. We can monitor that online, that's what we do. but you can't exclude that
10:28
Unidentified speaker (speaker_0): it's being used as a as a vector of attack. I think it is very theoretical because I don't think it's the case today. But on the other hand, this is part of the threat model. And so our customers they are asking us to be able to provide them with all of the open weight models that are available. We're helping them do it in a secured Old way where we can help them monitor what's going on. But in many cases,
10:48
Unidentified speaker (speaker_0): they are actually very keen to use our models. because we're able to tell them that the data supply chain, everything we've trained them on, is something that we have validated before.
10:57
Unidentified speaker (speaker_1): What percent of token traffic is your own models versus other open source? I
11:02
Unidentified speaker (UNKNOWN): would
11:02
Unidentified speaker (speaker_0): say around 80% comes from our own uh models.
11:05
Unidentified speaker (speaker_1): Wow. That's a lot more than I thought. Okay. On the 20%, everyone tells me that oh my God, American CIOs and CIOs would never use Chinese open source. models. And actually, that fear really is there. You have them as customers, you have amazing customers. Are they fearful of using Chinese open source models?
11:23
Unidentified speaker (speaker_0): It depends how critical their industry is in general to governments. If you work with defense, for instance, we work with defense customers in Europe. They would never use Chinese technology. If you work on the finance sector again, because it's critical infrastructure, they are very reluctant to do it. If you work in retail, uh if you work in certain domains of manufacturing, again it depends where
11:43
Unidentified speaker (speaker_0): the reluctance is different. Again, I think there is a common understanding that openweight models, if they are being deployed on sovereign infrastructure like ours, the only valid threat is that they go rogue and have weird behaviors in certain domains because you may have a supply chain attack on it. There There's is a common understanding that uh this is very theoretical and uh it's certainly not the case.
12:03
Unidentified speaker (speaker_0): today. But indeed it's part of the conversation. I'd say what is also part of the conversation is not this, is it a threat to deploy third-party models that I don't fully understand? What is part of the conversation is am I going to be able to get updates in six months from now? Because what's going on is These models have gotten so strong that
12:23
Unidentified speaker (speaker_0): they're able to do cyber attacks. That they're you can use that for. pretty much advanced purposes, let's say. And so the consequence is that they are more and more being treated as defense systems. It's being treated de facto by the US as a defense system because at some point in May they started to ban them from outside of the US. If you look at what China is doing today, they're
12:43
Unidentified speaker (speaker_0): also looking at ways of restricting exports beyond a certain threshold. The reason for that is that effectively this is a dual-use dual use technology. And so as any dual use technology, security of supply matters even more. And so the the companies we work with realise that they need security of supply not only on the infrastructure side, but also on the model IP side.
13:03
Unidentified speaker (speaker_1): With that in mind, do you think your open analyze and neuroanthropics withholding their best models is good or bad for society? And is that just an inherent feature that will be prevalent within this industry?
13:14
Unidentified speaker (speaker_0): Is it good or bad for society? I don't know. What is clear is that there is immense leverage to centralized AI providers in a world where there's no providers that is allowing decentralized deployment that is allowing sovereign deployment. So what we are doing is extremely important for companies. If we end up in a world where all companies need to
13:34
Unidentified speaker (speaker_0): rely on two or three providers that are centralized AI providers that give them enormous leverage. They will use the data that is being produced by the systems to make better systems and to go after the business of their customers. Because at the end, this is the end game. This is pretty pretty clearly what is being priced in today. So the approach we're taking is to make sure that this world in which you have
13:54
Unidentified speaker (speaker_0): an oligopoly of AI providers that are leveraging all data being produced by enterprises to make better systems that can deintermediate those enterprises is not the world we live in. And we're very aligned in that with all of our customers that of course don't want to live in such a world. So I would say the fact that there are interval models that are being used to do certain research, for instance,
14:14
Unidentified speaker (speaker_0): is an early sign of what can happen in a world where you have massive concentration of power in artificial intelligence. Luckily, it's not that complex to build models, it's about deploying capital, it's about having the right team that is training the right thing. And so we're we've been able to do it. The open weight frontier is catching up with the closed frontier. In many domains, it's actually above. And so that means that I don't
14:34
Unidentified speaker (speaker_0): think we'll end up in a world where if you want to do mathematics, you need to work at OpenAI. Luckily, there is enough people in the world that don't want this to occur.
14:41
Unidentified speaker (speaker_1): You've said the word energy a couple of times. And one of the core underlying tenants of being successful is having the energy and the infrastructure to power European energy. AI. We have a very different energy system to the US and we have a very different energy system to China. In China they'll move three million people to put some form of energy infrastructure and in the UK we won't move a
15:01
Unidentified speaker (speaker_1): birdhouse. It's quite concerning despite the joke. How do you think about ruminate on the energy supply challenge? Well
15:08
Unidentified speaker (speaker_0): I think the first thing to realize is that energy is around ten 10% percent of the bill in the total cost of ownership of a uh compute. So we're not talking about the place where we can make most of the optimization. So the cost of energy at the end of the day is not the one thing that matters in the economic of a project. What matters is the cost of the chips and what matters is the cost of capital. This is dominating the the bill of material.
15:28
Unidentified speaker (speaker_0): So it's a problem. It will become more and more of a problem as we deploy more compute in Europe because we'll run into supply constraints. Again, it's not evenly distributed. So if you look at France, for instance, there are excessive energy, so you can build capacity. If you look at Nordics, it's also the case. In the UK it's a little less the case. Germany it's a little less the case. But
15:48
Unidentified speaker (speaker_0): there is a realisation that AI is going to drive energy consumption forward. I would say in the next two or three years, the true uh fight, the true challenge is to actually make sure that for Mistral, make sure that we build as fast as possible. to use the excess excess of energy that is already there. And so that is what we are at at work doing
16:08
Unidentified speaker (speaker_0): raising capital, building more, building the team to build all of this capacity. There is excess of energy. It's not the bottleneck today. It will become the bottleneck in twenty nine, twenty thirty. But right now the question is how can we build Built on existing energy? Now we need to think ahead of time in starting in twenty 2029-2030, nine,
16:27
Unidentified speaker (speaker_0): as multiple gigawatt of capacity are being deployed every year on AI, we do need to ramp up the production. What we are hopeful for is that in France, for instance, there is an enormous EPR program where uh nuclear energy is going to be produced much more. You have large uh efforts on the solar side as well. So it's going to be a
16:47
Unidentified speaker (speaker_0): problem, it's mostly a coordination problem because if you have demonstration. demand coming from AI on energy, the business case for building new nuclear plants is pretty well made.
16:56
Unidentified speaker (speaker_1): It totally is and I agree with you. I am a big believer in Adam Smith's invisible hand, but the lead time is very real. And so that is just like the 29-2030 is a challenge.
17:06
Unidentified speaker (UNKNOWN): Yeah.
17:06
Unidentified speaker (speaker_0): Nuclear plants take longer than that to build. Solar plants can go faster. And say I do believe in that invisible hand. The states have to play a role uh in facilitating the build-out, but also in planning ahead of time. They they have a longer term vision than enterprises in general. And so we are spending some time, I'm spending some time
17:26
Unidentified speaker (speaker_0): explaining that the energy demand is going to grow and that it needs to be addressed somehow. It needs to be part of the business plan of uh of any energy supplier.
17:35
Unidentified speaker (speaker_1): You said about the cost of chips being important. What do you know now?
17:39
Unidentified speaker (speaker_1): about compute buying, forecasting demand of compute chips that you didn't know? I
17:45
Unidentified speaker (speaker_0): mean it's quite a process. So we've learned to run that process to match supply and demand. It's a complex industrial process. It has nothing to do with software. You need to understand the latency in the supply chain. You need to understand that the supply chain comes from multiple I mean, the supply chain is fairly complex because you have GPUs, but you also have CPUs. you have storage, uh, you have electrical
18:05
Unidentified speaker (speaker_0): parts, uh, you need to make sure that your own suppliers have no problem with their own suppliers. So that aspect is something we actually learned, but it's uh it's a fairly complex one where you need to manage resources, you need to have multiple forecasts, you need to be able to forecast your demand for multiple buckets. So enterprises, which are our main ICP, but
18:25
Unidentified speaker (speaker_0): also AI natives that can take larger pieces of GPUs but want them for a shorter period of time. You have companies like Microsoft that are off taking capacity from us that we can use to finance more capacity. So at the end, the problem of building a cloud is about matching supply and demand. It's about managing the supply
18:44
Unidentified speaker (speaker_0): chain ahead of time, giving visibility to your suppliers. And it's about matching all of this with the finances. You need project financing, you need people that are strong at understanding credit. It's actually quite interesting to build such a business. What we've learned is that at the end, if you if you're able to front capacity ahead of demand, you get something that is pretty
19:04
Unidentified speaker (speaker_0): easy to sell. And it's going to be the case for the next few years. And so we're very grateful that we anticipated the fact that we would run out of supply fairly quickly, that our customers would have supply chain. problems for their own GPUs. Because now that we have those GPUs, we can serve their needs.
19:19
Unidentified speaker (speaker_1): That was one, which is cost of chips. Cost of capital was another one. If you were resource unconstrained completely, what would you do today that you're not doing?
19:29
Unidentified speaker (speaker_0): We would build much more capacity. I think that's the big bottleneck in building capacity, is of course capital. And we see the demand and we know that our customers
19:39
Unidentified speaker (speaker_0): are uh running out of compute. As we take some of their use cases and we help them scale it, suddenly they realize they don't need eight GPUs, but they need thousand GPUs. And that's for one use case. And so we see that demand ahead of our customers. So we have pretty high conviction. That's why we put equity at work to build capacity. And so what we are working on, what we have always been working
19:59
Unidentified speaker (speaker_0): on, is to make sure we have more capacity under management. even by by building ahead of demand. And today if we just look at what Europe needs, it's pretty clear that there is not enough compute. You may argue that there is maybe a little too much compute being built in the US, but what is clear is that there is a lot of slack to build capacity in Europe in a way where you are
20:19
Unidentified speaker (speaker_0): 100% confident that it's going to be allocated. So that's uh that's what we're working on.
20:23
Unidentified speaker (speaker_1): Well, you said hey, there's a role to be played by the state as well. What would you like to see the state, be it France, be it the UK, do that they're not doing today?
20:33
Unidentified speaker (UNKNOWN): I
20:33
Unidentified speaker (speaker_0): mean we've been quite vocal about the kind of ways in which they can intervene in a way which is market neutral and that just would facilitate investment. Uh the first thing is of course to facilitate permitting and to make sure that there are dedicated sites where things are accelerated. where uh all stakeholders, whether regional
20:53
Unidentified speaker (speaker_0): or state level, are aligned. And I would say France, Sweden are making a great job at that, for instance. So that's mostly on facilitation of aligning stakeholders because these are physical assets, uh, they are built within regional communities. It's super important to make sure that everybody is benefiting from it. And so the help of states
21:13
Unidentified speaker (speaker_0): happens there. What they can do at the financial level, and we've said that publicly, is that state guarantees on leasing data centers, for instance, on reserving space for 10 or 15 years, are super useful for companies like us that have a relatively small balance sheet and that knows how to build and who to build it for. So that's
21:33
Unidentified speaker (speaker_0): a role they can play by giving some guarantees, uh putting some guarantee on the table to build further. and ahead of time. Then of course, public off take. Uh we see public sector is is an important customer for us from a lot of different states in Europe and outside. And again, we bring them value, they bring value to the citizens, it creates
21:53
Unidentified speaker (speaker_0): off-take for companies like us, and that allows us to build further. And if you build further and faster, you build cheaper uh in this day of supply supply constraint.
22:02
Unidentified speaker (speaker_1): You said you sell to states that and kind of going back to that element of kind of sovereignty. How far down does sovereignty need to go? And what I mean by that is like we obviously have cohere in Canada. Does like every country need their own model? Is it blocks that need their how do we think about like the granularity of sovereignty?
22:19
Unidentified speaker (speaker_0): I think the granularity that is allowed load is the you need a fairly large market to build uh relevant models. And to today Mistral is able to build relevant models because we operate not at national level uh we're not a French company but we serve Europe we serve the US we serve Asia and what we need is alternatives but there will not be a
22:39
Unidentified speaker (speaker_0): single sovereign AI provider and model trainer at every state level this is not going to happen.
22:44
Unidentified speaker (speaker_1): What percent of your customers are European?
22:46
Unidentified speaker (speaker_0): If we look at our revenue base it's around sixty 60%. percent and the rest comes from the US and comes from APAC.
22:53
Unidentified speaker (speaker_1): And they use you to do post-training harnesses for their own models?
22:57
Unidentified speaker (speaker_0): Yeah, European ones, they use us for multiple reasons, but we help them build applications on either things that are related to the efficiency of their businesses or related to accelerating the RD. We work with manufacturing companies, with financial services, we work with defense companies, and we have a pretty strong verticalization approach. to all of these three domains.
23:16
Unidentified speaker (speaker_0): And the fourth being public sector. The way we work with manufacturing companies is in general we are with their RD teams and we help them build better products. So we help ISML build better machines, operate better machines, we help Airbus build better planes, and we bring reasoning models, but we also bring the physics models, the models that have a quantitative understanding of the world. In financial services,
23:36
Unidentified speaker (speaker_0): we build models and harnesses that understand. markets. We also bring models that understand quantitative time series to help do predictions. So we come with a set of models that allow us to quickly build applications that bring value to our customers. So that's how we serve them. And I guess that back to your point on sovereignty, all of these companies have
23:56
Unidentified speaker (speaker_0): a sovereign I AI strategy. They need deployment of the technology in a way where they have business continuity guarantees. affordability constraints, etc.
24:04
Unidentified speaker (speaker_1): We see OpenAI, Anthropic, DeepSeek move into application layer and also move into chip layer with Jalapeno, with Anthropics Rumer, DeepSeeker doing their own. Will we see a Mysterile chip? How do you think about the importance of verticalization and owning every layer?
24:19
Unidentified speaker (UNKNOWN): No,
24:20
Unidentified speaker (speaker_0): it's not planned. We have gone down the stack, or I would say upstream if you look at it from an industrial perspective. We have gone from training models to serving the models we train on infrastructure that we own. So that's that's already a pretty significant move. That's a move that took around 18 months to to set up, and that is not pretty successful because we can serve the
24:39
Unidentified speaker (speaker_0): GPUs, we can serve tokens, uh, we have tokens that are purely Europe generated in that it escapes any form of US extraterritorial law. It matters a lot for our customers. So that took us some effort. We are doubling down on this effort to scale that further. And I think that's going to keep us busy for a while. We are pretty
24:59
Unidentified speaker (speaker_0): vertically integrated in that we we have cloud services infrastructure and we go all the way to building the business applications for our customers. You mentioned the application layer. In AI, you bring the platform and you build custom applications for customers. That's how it works. That is what AI allows you to do.
25:14
Unidentified speaker (UNKNOWN): And
25:14
Unidentified speaker (speaker_1): sorry, this is this is through an FDE army. Is
25:17
Unidentified speaker (UNKNOWN): it going?
25:18
Unidentified speaker (speaker_0): Yeah, it's FDE but building on top of common primitives. So uh workflow management, agent management, management of sandboxes, observability of the deployment of different business applications. So at the end, we are building cloud services that are adapted to run and build with AI agents.
25:35
Unidentified speaker (speaker_1): How do you think about doing it through Mistral FDEs versus partnerships economy, which we've traditionally had?
25:41
Unidentified speaker (speaker_0): It has been super helpful to have our FDEs front and center at the beginning of our journey because it allowed us to understand exactly how where enterprises were struggling. They started with individual productivity, it was not enough. Chatbots were not delivering any kind of value creation for them. And so we went and spent some time with the managers, the business unit
26:01
Unidentified speaker (speaker_0): owners, and understood the kind. of core processes they needed to automate with artificial intelligence. We did it in a way where we built business applications, custom applications that are managing state, uh applications where agents are operating the process end-to-end. And it was super helpful for us to understand the primitives to build. So we built durable
26:21
Unidentified speaker (speaker_0): process execution, for instance. We understood how to orchestrate agents in a way that was very efficient. effective. So having this FDEs, uh our FDE sit with customers directly has been super useful. Now, what we have done since last year is to start propagating the tools to some of our partners, like Accenture, like
26:40
Unidentified speaker (speaker_0): TCS. So system integrators that are great at scaling things, and that once you give them the right tools, once you give them the right training, they can help scale because they have just more people. But having people actually sitting hands on keyboard from Mistral with our customers is super useful for our customers and it's in high
27:00
Unidentified speaker (speaker_0): demand. It's super useful for us. It's not something we can scale indefinitely because we can't have tens of thousands of FDEs working for us. But it's something that now we are scaling through partners, and the two motions coexist. So we are in many accounts, we have FDEs that are working directly with customers. They're operating the projects, they're making sure that the
27:20
Unidentified speaker (speaker_0): delivery is occurring. And we work hands in hands with system integrators to do the work, to then do the maintenance and these kind of things.
27:28
Unidentified speaker (speaker_1): What line of revenue do you not have today that you think will be most significant in five years' time? When I was listening to you there, I was thinking, I bet you wouldn't have necessarily predicted these different motions when you started. when we first met Arthur where I don't know if you remember, but you were sitting on a park bench just chatting to me about life.
27:45
Unidentified speaker (speaker_0): I was. Yeah. You did not invest, you should have. Dude,
27:48
Unidentified speaker (UNKNOWN): I
27:48
Unidentified speaker (speaker_1): fucking regret it so much.
27:52
Unidentified speaker (UNKNOWN): I
27:52
Unidentified speaker (speaker_2): saw Jean-Charles
27:53
Unidentified speaker (speaker_1): yesterday. And every time I see
27:55
Unidentified speaker (speaker_2): Jean-Charles, we talk
27:56
Unidentified speaker (speaker_1): about you. Yeah, it's a brilliant reminder of just you know how much money I lost. It's or all missed out on. Sorry about that.
28:01
Unidentified speaker (UNKNOWN): Now
28:02
Unidentified speaker (speaker_0): what line of revenue? I mean, we have these two lines of revenue. We have one which where we deploy the platform that is truly licensed revenue uh for building business applications, and then we have the consumption revenue coming from the infrastructure business that we built. And the two things go hands in hands because we are able to our customers they they come in general for the
28:21
Unidentified speaker (speaker_0): license motion, but then they realize that they don't have enough GPUs, uh, and then we're trying to be helpful to them uh by providing them with either GPUs or or token services. So the two things are very well combined together.
28:34
Unidentified speaker (speaker_1): Do you think you'll have an inference business, a big inference business in three to five years' time? Do you think you'll have a sandbox business? Uh
28:40
Unidentified speaker (speaker_0): yeah. If you look at this granularity, I'd say the GPU side, like everything sitting on GPU is going to continue to grow because the demand is enormous. You have what we call anciliary service. services, like running soundboxes, running scaling up and down soundboxes very easily when you're running uh reinforcement learning algorithms, for instance, or when you're
29:00
Unidentified speaker (speaker_0): deploying many agents on the same tasks. That is a growing business for us because it comes from the demand of our customers. And in turn, that brings us to buy more and more CPUs. So I do expect what we call the runtime side of things. So where agents execute tasks, call tools, compile code, and run the code to
29:20
Unidentified speaker (speaker_0): grow. It's pretty clear that when you are in synchronous tasks where you were only using GPUs to answer conversations, you didn't have that. When when you start to have managed agents that are running on the background, doing a lot of tasks that are CPU intensive, you need more CPUs, you need more sandboxes, and you need more storage. as well because they
29:40
Unidentified speaker (speaker_0): produce a lot of data. So those two things, I would say CPU-related tasks, data-related tasks are going to be growing line of businesses for us.
29:47
Unidentified speaker (speaker_1): A big part of the business is selling to governments. What does no one know about selling to governments that everyone should know?
29:53
Unidentified speaker (speaker_0): The first thing is that uh there's no such thing as selling to governments in general. All governments are very are very different, I'd say, depending on.
30:00
Unidentified speaker (UNKNOWN): the
30:00
Unidentified speaker (speaker_1): Which is the best? Some
30:02
Unidentified speaker (speaker_0): governments. are small enough so that it looks more or less like an enterprise sales motion. It's the case of uh governments where everybody in ministries know each other, for instance. All the governments in larger countries are slightly more complex. So again, the complexity of finding where to create value, of understanding what the
30:21
Unidentified speaker (speaker_0): customer needs grow with the size of the country, I should say.
30:25
Unidentified speaker (speaker_1): Can you help me on the inference more? market? It's something that I just totally didn't know enough about. What I know now that I wish I'd known 12 months ago was that there is just infinite demand for inference. It's just the depth is astonishing. What happens with the inference market?
30:40
Unidentified speaker (UNKNOWN): I
30:40
Unidentified speaker (speaker_0): mean the inference market behaves very much like a commodity market, which is pretty high margin today. In that if you have supply, if you have the right quality of service for certain kind kinds of models, you can sell them on demand because you have a lot of demand coming directly for from customers that are non committed. And then for customers that are sufficiently mature,
31:00
Unidentified speaker (speaker_0): they need to have guaranteed quality of services. They need to have service level agreements, and there they need to have reserved physical capacity. And that's something that we can provide because we have the right forecasting models and we have the right supply. So you have these two things you have one which is on demand. which very much behave as a commodity market where pricing matters a lot. And then you
31:20
Unidentified speaker (speaker_0): have uh at a premium side of the market where the customers actually want to have access to want guarantees on the latency, want guarantees on the speed of generation for a certain number of requests. And the more they can be precise about the shape of their demand, the better or our pricing can be. The better we
31:40
Unidentified speaker (speaker_0): can shape our own infrastructure to make sure that they uh have the right criteria of service for the right price.
31:45
Unidentified speaker (speaker_1): We have so many different Neo Lab providers. There have been 102 different Neo Lab providers created. Every week there's another deep mind spin out and another. And so many. What will separate Neolabs Neo labs that thrive and those that don't?
32:00
Unidentified speaker (speaker_0): Well, the first thing, I mean, we started as a Neolab ourselves. It wasn't called that way in 2020 twenty three, but uh effectively when you start a company you have a bunch of silver bullets to hire the best, and we used all of our silver bullets to hire researchers. And so effectively we started as a research company. I think what has allowed us to scale and to now operate in
32:20
Unidentified speaker (speaker_0): a much wider way, still with this number one pillar, which is to deliver models, and that's why Stra Large 4 has been very successful for us. And I think it's very exciting as well in the new releases that are coming, because we're making pretty good progress. What has been successful is that we focused early on on a market that would scale. So we went for the enterprise market, which comes with a certain
32:40
Unidentified speaker (speaker_0): number of complexity, because of course uh enterprises' customers are demanding as they should be, and we have adapted our product to meet those demands. But the fact that we were able to both operate at the research level, at the highest level, and able to constantly recruit and invest further in making stronger and stronger models. And the fact
32:59
Unidentified speaker (speaker_0): that we built a customer centric motion where we understood the problems of our customers have allowed us to scale. And that's something that every Neo lab needs to think about when they start building their business.
33:10
Unidentified speaker (speaker_1): But how many people are there actually in Europe who truly have the ability to create these companies and have have been it and seen it like you have? I don't know, maybe a Jan McCun has done. Is it 20? Is it a hundred?
33:23
Unidentified speaker (UNKNOWN): I
33:23
Unidentified speaker (speaker_0): mean, I'm not a VC investor, so I would you probably know it better than I do. But there's a a lot of talent in Europe generally. You have a lot of entrepreneurial talent, you have a lot of researcher Search talent, your and you have a pretty good combination of those two. So that's why you see a lot of new neo labs uh in Europe. The hard thing is some. it's not only about being talented, it's not only about being able to sell uh and having
33:43
Unidentified speaker (speaker_0): something that you can successfully create a business with. It's also today about being able to access compute. I think a big bottleneck today for many neolabs that I know is that accessing GPUs to actually do interesting research in certain verticals is hard because there's an enormous supply crunch. And so that's why in the
34:03
Unidentified speaker (speaker_0): supply we have, we are reserving some capacity for certain neo labs because we believe that they can grow and that we can grow with them. But effectively, if you're a founder today and you want to do something which is very research-centric, the biggest problem you have is to get access to it of compute with a decent price.
34:21
Unidentified speaker (speaker_1): Researchers there. How crazy is the researcher hiring market today?
34:26
Unidentified speaker (UNKNOWN): Well,
34:26
Unidentified speaker (speaker_0): it's pretty competitive, I should say. Uh again, if you have the right mission, if you have the right compensation structure, if you're doing something that matters for the world, people will join us. And I would say uh we're very grateful for the dedication of all of our employees to making uh Mistral successful. The quality of the Mistral Stra Large Large4 4 is
34:46
Unidentified speaker (speaker_0): has been a success, uh, has a lot of traction today, and is leading the pack in the open weight category and coding and in cyber defense. It's only due to the dedication of this team in the last three, four months, the resilience as well, because when you run a model, we've we've posted our tech report uh recently, oftentimes it doesn't work the first time. And so
35:06
Unidentified speaker (speaker_0): you need to be able to course correct in a way where okay, you're going to spend two more months to actually make it work, and it's going to be painful. Uh, we're expecting to release, but we'll need to release a little later. It's going to be painful for everyone, but everyone needs to be all hands on tech and do the right thing. And I'm very grateful for the resilience of the team, uh,
35:26
Unidentified speaker (speaker_0): given the the scarcity of resources that we have.
35:29
Unidentified speaker (speaker_1): I just had Satya from Microsoft on the show and he talked about Mustafa's statement before that we'll see a hundred billion dollar training runs in the next few years. Do you think that's a little bit hyperbolic? Or is that actually well within reach?
35:41
Unidentified speaker (speaker_0): You do have a data wall before, I should say, in that uh at some point if you want to make your model very strong at certain domains like chip design, for instance, or if you want to make it very strong at operating industrial engineering workflows like we do, you do need to have the right environments to train them. And this part takes some skill and taste, is not something that you can
36:01
Unidentified speaker (speaker_0): scale as easily as just stacking compute. So I would say the the bottleneck that we're all running into today is effectively to be able to pro to source enough quality data. It's compute, we need more compute, we're deploying more compute. Getting to a hundred billion of training runs I think is not for tomorrow clearly.
36:20
Unidentified speaker (speaker_1): If everyone's so short on compute, why is Nabius Nebius forty 40% percent down? Why is CoreWeb 40% down? Why is the AI trade cracking in public markets if it's so clear that everyone is so compute starved?
36:34
Unidentified speaker (speaker_0): It's a little unclear to me how markets are reacting today. What we see is that we're fully able to finance our own compute, given the quality of the collateral that we have in terms of revenue commitment, e-corpe e-corpay commitments. We as a company operate as a vertically. integrated company. So we build compute, but for
36:54
Unidentified speaker (speaker_0): higher level purposes, for serving the enterprises, for building business applications. So we are anchored on the value creation of our customers, which means that at the financing level, we haven't had to go and leverage enormously the company on debt financing. And so that gives us a lot of confidence in what's going to happen. But effectively,
37:14
Unidentified speaker (speaker_0): as in any industrial build-out,
37:17
Unidentified speaker (speaker_0): it's tempting to get a lot of leverage, but at some point it's leverage break. And so that's why we're very happy to be vertically integrated and to operate from business application creation all the way to the infrastructure part.
37:30
Unidentified speaker (speaker_1): What did you not do with the benefit of focus? That with hindsight, you go, I wonder if we should have done that.
37:36
Unidentified speaker (speaker_0): We did not go deep into the consumer. market, but I think we're very grateful that we did not do it. We did not go into cheap design. I think that's uh
37:46
Unidentified speaker (speaker_1): pause. What like why why do you feel grateful you didn't go into the consumer market?
37:51
Unidentified speaker (UNKNOWN): I
37:51
Unidentified speaker (speaker_0): think it's a very packed market. And it's very hard to operate at the highest level with enterprises and also to run what is inherently going to be an ads business. So that part, that decision to focus on the enterprise segment. also that decision to focus on the four verticals that we have decided to focus on manufacturing, financial services, defense, public
38:10
Unidentified speaker (speaker_0): sector, has been very critical to focus the company. Because the problem with artificial intelligence is that it can do everything. And you need a source of focus. And so the source of focus we've picked comes from the market. So we want to have the models that serve the best agentic engineering workflow. workflows, that serve the best uh financial decision making, that serves
38:30
Unidentified speaker (speaker_0): the best defense purposes and everything related to state sovereignty as well.
38:34
Unidentified speaker (speaker_1): With RSI, do you need a source of focus?
38:38
Unidentified speaker (speaker_0): I mean, using AI to improve AI models do require to focus as well, because you have limited amount of resources to spend on problems. So at the end, the bottleneck here is indeed compute. The bottleneck is also that everything that you do that which is recursively self-improving or just for self-improving because it means the same thing,
38:58
Unidentified speaker (speaker_0): requires to have well defined problems. If you have a well-defined problem today with artificial intelligence, you can later climb it. In many cases in enterprises, the problems are not well defined. And so the true friction to economic growth is not the fact that we cannot solve hard problems that are well well-defined. defined, it's that we it's hard to solve ambiguous problems where you
39:18
Unidentified speaker (speaker_0): have ambiguous objectives, where you have humans in the loop. And that part you cannot really solve with self-improvement. You can only solve with models being deployed and interacting with humans. And that's the part that we are focusing on, I would say. But we we are investing a lot on making self-improving Improving loops on well-defined problems as well. But that is the part that is
39:37
Unidentified speaker (speaker_0): constraining the explosion. of AI usage, I would say, in the enterprise. The poorly defined problems on which you want to deploy AI, but you have no way to lay the climb because the definition of the problem is not proper. And so you need to spend some time. That's why you have FDEs, that's why we have FDEs to try and better define the problem.
39:55
Unidentified speaker (speaker_1): Arthur, I've got to be blunt with you. I joked at the beginning that you're like Europe's last hope. There is a, and I don't mean that like super dramatically. but there is a very big chasm between you and anyone else in terms of Neo Labs that will aspire to be the next Mistral. Do you and the team feel the pressure?
40:11
Unidentified speaker (speaker_0): We do feel some pressure, but as anyone in the field, this is uh fast growing, fast moving. You need to, as an entrepreneur in AI, you need to deliver on the model side because that's the one thing that shows the progress you're making. Then enterprises need to see the value, so you need to have people that are understanding the true
40:30
Unidentified speaker (speaker_0): business problem. So it's a fairly different problem to solve. So we do feel some pressure, but I think it's a very healthy pressure. Uh, it's a very important moment in time and in history. We need to be meeting the expectations of the general opinion. At the end, we need to show that AI is useful, the kind of deployment we're making, the kind of noise this is making in the market.
40:50
Unidentified speaker (speaker_0): This technology needs to appear as a positive technology. We need to show the positivity of that technology. And the only way in order to do so is to promote sovereign AI as a category, is to make sure that uh that creates economic growth and opportunities for everyone, to make sure that the gains of AI is well distributed. And so, of course it creates some
41:10
Unidentified speaker (speaker_0): pressure, uh, because it's not a given. If we don't succeed, I don't know what what's going to happen in the world. I think so. We do have a role to play. I think the mission we play is important and it's pressure, but it's also quite exciting.
41:22
Unidentified speaker (speaker_1): Will AI inflat GDP growth by more than two percent a year?
41:26
Unidentified speaker (speaker_0): Well, that's the entire question. What kind of uh bottleneck do you go and solve to actually inflect that GDP? That's the reason why we were very bullish on industrial engineering. At the end, if you want GDP growth, you need more production. You don't need more intelligence. You need more production. In order to have more production, you need better
41:46
Unidentified speaker (speaker_0): factories, you need objects that are more efficient, you need better planes that are flying for longer with the same amount of oil. You need better energy production that are more efficient, less carbon-intensive. Intensive, You need better SML ISML machines. You need better services also in financial services. In industry and engineering, in particular, using AI allows
42:06
Unidentified speaker (speaker_0): you to do better design of objects. So if you set up the agent tick loop to design safer cars, for instance, you can have your car design team go much faster and iterate faster on that product. That means that you can sell more, that means that you can sell more cheaply, that means that you're building safer products. In the aerospace industry, if you design a
42:26
Unidentified speaker (speaker_0): better plane, it's going to be out there for 40 years. If it runs faster with less fuel, then it's going to create GDP growth. So at the end, what truly matters is to enable AI to have a physical effect. This effect we find mostly in manufacturing and financial services.
42:45
Unidentified speaker (speaker_1): I'm aware of time. I'm gonna do a quick fire round with you. Okay.
42:48
Unidentified speaker (UNKNOWN): Please
42:48
Unidentified speaker (speaker_0): do.
42:49
Unidentified speaker (speaker_1): What do you believe? today that most around you would disagree with?
42:53
Unidentified speaker (speaker_0): I think we strongly believe that this is a fragmenting market and that sovereign AI is going to be the most important category of AI deployment in the enterprise. In that when enterprises are going to deploy the technology, business continuity, cost, and control over their Or data and in IP is going to be the one thing that matters. So I think that's a very strong conviction that we have.
43:13
Unidentified speaker (speaker_0): that is shared by a lot of enterprises and shared by a significant part of the market, but of course not shared with some of our competitors.
43:21
Unidentified speaker (speaker_1): What is a extremely wild prediction today that will be very common in five years' time about the way that we live, the way that we work?
43:29
Unidentified speaker (speaker_0): I think the way we work as engineers has already changed for software engineers, is going to completely change for any kind of engineers. So I'm super excited. excited by that approach of designing better products in a way which is fairly autonomous, uh, thanks to artificial intelligence. So it does not concern everybody, but
43:49
Unidentified speaker (speaker_0): everyone with a technical background, and not only the software engineers that were the first affected, and the mathematicians, everyone with a technical background will be operating AI agents in a way that will give them enormous leverage. So this in itself is going to be a major shift in the way the labour market is actually working.
44:08
Unidentified speaker (speaker_1): What do you do about going public? Because one day you will want to, okay. But you can't go public in Europe. You
44:16
Unidentified speaker (UNKNOWN): can
44:16
Unidentified speaker (speaker_0): absolutely go public in Europe.
44:18
Unidentified speaker (speaker_0): Yeah, you need to.
44:19
Unidentified speaker (UNKNOWN): Are you going to
44:19
Unidentified speaker (speaker_1): go public in France? No.
44:22
Unidentified speaker (speaker_0): What used to be true ten 10 years ago, I think, is no longer true. There's a huge appetite from investors from a across the world to invest in Europe, both on the private equity side but also on the public equity side. And so the bottlenecks that used to be there are no longer there. So we at Mistral are of course getting prepared on the you know, you have a lot of operational constraints when you when you go and become
44:42
Unidentified speaker (speaker_0): public. We're also pushing and and delivering on the the commercial milestone that we uh we need to deliver on. But uh I don't think this is any more anymore a bottleneck. And there are widely successful companies, uh some of them are stakeholders that are firstly listed in Europe. So there is no reason why not to go and
45:01
Unidentified speaker (speaker_0): and an IPO in Europe at some point.
45:04
Unidentified speaker (speaker_1): Final one for you dude. What are you most excited about when you look forward to the next five, seven, ten years that not many people think about?
45:12
Unidentified speaker (speaker_0): I'm quite excited by how AI is going to enable foundational research in domains where we lack understanding of the world. So, in physics, for instance, the fact that we can both run AI simulations at a scale that was nowhere reached, and that's the team we bought called Amy is specialist in that, is going to
45:32
Unidentified speaker (speaker_0): enable new discoveries. And so the fact that we can build simulators and that we can have agents operating those simulators. to understand better the world is going to be an enormous enabler of the understanding of the world. So I come from a math and physics background. And so I'm quite excited by the opportunity ahead of us in the next generation of researchers.
45:52
Unidentified speaker (speaker_1): Do you know what, Arthur? I blame you. I blame you when you sat on the bench and you didn't speak at all like you speak now. Like the author then and the author today are just like paradoxical. paradoxically different with respect. It's amazing to see the leader that you are and you've become. I I I honestly dude, I have so much respect, admiration, gratitude for you and all you've done
46:11
Unidentified speaker (speaker_1): for Europe. But wow, it's amazing.
46:16
Unidentified speaker (UNKNOWN): Thank you
46:16
Unidentified speaker (speaker_0): for the praise. There's still a lot of things we need to deliver on, but uh we uh quite happy with the progress we've done. Uh I think the mission is important. Uh and so if anyone watching us uh wants to join us in that journey, uh please uh please uh contact
46:30
Unidentified speaker (UNKNOWN): me.
46:32
Unidentified speaker (speaker_1): But before we leave you today, base 44 is one of those products I used, and I just thought straight away: man, I wish I'd had this when I started 20 VC with $50 and no contacts in Vansure. I wanted to see what it could do. So I built a guest research hub for the show. Upcoming guests, company notes, previous interviews, sponsored. sponsor contests, follow
46:52
Unidentified speaker (speaker_1): ups, all in one amazing location. I described it in plain English. A few minutes later I had the product open in front of me. Database, login, backend, and hosting already handled and just all in one. It made it so easy. And that's what I loved. I didn't write a long spec, I didn't stitch five different tools together, I built the thing and I started using it. For any founder listening,
47:12
Unidentified speaker (speaker_1): that's the point. The faster you turn an idea into something real, the faster you launch. Apps, dashboards, internal systems, AI agents, whatever you need to build, base44 gets you there. Go to base44.com forward slash 20vc. That's base44.com slash 20vc. While base44 helps you build faster,
47:32
Unidentified speaker (speaker_1): Laradin helps you measure AI impact. Every company now runs two workforces humans and AI. agents. Larridin is the first unified dashboard across both workforces that measures their work and optimizes their impact. Larridin answers four questions who's using AI, which tools, what is it costing,
47:52
Unidentified speaker (speaker_1): and was it actually being produced for every team across every vendor? AI ROI stops being a guess and becomes a number that you can consistently and confidently measure. Gainsight, Vertiv, SurveyMonkey, and many other enterprise organizations run their AI programs on Larridin. Thousands of AI tools and
48:12
Unidentified speaker (speaker_1): millions of dollars of AI spend are measured. You bet on AI, Larridin proves that it's working. Visit Lauridin.com to learn more. That's L-A-R-R-I-D-I-N.com. While Lauradin Larden measures AI impact, shop helps you buy smarter. My agent will summarize a board deck, researcher founder, and draft
48:32
Unidentified speaker (speaker_1): my notes. Ask it to buy something though, and it just falls apart. Wrong size, dead links, stuck at a login, and I end up doing it myself. That's why the shop skill from Shopify is supporting today's show. Head over to shop.md, copy the prompt, paste it into whatever agent you already use, and that's it. Your agent can now search millions of stores
48:52
Unidentified speaker (speaker_1): on Shopify, over a billion products. Sign in with Shop and it knows your preferences and order history. So the pics actually look like you, and nothing is bought until you approve it in Shop. I've installed the Shop skill with my agent, and I love what it found for me. So I needed protein powder in a specific flavor, under a specific budget
49:12
Unidentified speaker (speaker_1): I didn't want to spend too much, and I wanted it before the weekend. So I just typed that into my agent, exactly how I'd say it. Before it would have been stuck at checkout. This time though, it came back with exactly what I wanted. My details were already filled in. I hit approve and that was it. Turn your agent into a personal shopper with the shop skill. Visit shop.md
49:31
Unidentified speaker (speaker_1): slash 20 minute VC. VC That's That's shop shop.md dot MD slash slash 20minute twenty minute VC. VC.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.

More from The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch

The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch

20VC: Jev Creator on Why Data Is More Important Than Compute | Will US Open-Source Models Catch Chinese | The Cost of Intelligence and Tokens: What No One Sees But Needs to Know with Diogo Almeida

Episode discussion: Jev Creator argues AI reliability and data quality trump compute, predicting a shift from front-end apps to background automation.

Speakers’ claims: Jev Creator (Diogo Almeida) argues that the primary bottleneck for AI automation is not compute or speed, but the reliability of intelligence, defined as the consistency of decision-making rather than system uptime. He posits a 'bitterer lesson' where task definition and data quality are more critical than algorithms or compute, which he claims have plateaued in effectiveness since GPT-4.5. Creator criticizes current market hype, benchmark-maxing, and the 'AI coworker' model, predicting a 'SaaS Palooza' where incumbents win by integrating AI into workflows rather than replacing them. He also discusses TypeSafe AI's fundraising strategy, hiring philosophy, and the geopolitical landscape of open-source models.

AI ReliabilityData QualityVenture Capital
1h 14m2 assets22 claims to verify
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch

20VC: Taking on NVIDIA: Fractile CEO on the Future of Chips | What Everyone Gets Wrong About Capex Spending | The Winners and Losers in Neo Clouds with Walter Goodwin

Episode discussion: Walter Goodwin on Fractile's 3D-stacked memory chips, AI infrastructure bottlenecks, and NVIDIA's future valuation

Speakers’ claims: Walter Goodwin, CEO of Fractile, argues that frontier AI intelligence is not yet commoditized, distinguishing it from general cloud compute. He positions Fractile's 3D-stacked memory architecture as a solution to the memory bandwidth bottleneck that limits LLM inference speeds. Goodwin contends that land and power, rather than silicon, are now the primary constraints for AI infrastructure, favoring energy-efficient designs. He predicts NVIDIA will reach a $10 trillion valuation within five years, driven by the finite nature of data center power and the need for chip replacement cycles.

SemiconductorsAI InfrastructureVenture Capital
1h 16m3 assets25 claims to verify
All insights