[비즈한국] Back in April, I introduced the scene from the state banquet held at the Blue House in honor of French President Emmanuel Macron's visit to South Korea (Related article: The 33-Year-Old Startup Founder Sitting Next to Lee Jae-yong… The ‘Future’ France Envisions).
At that time, a 33-year-old startup founder from France was seated next to Samsung Electronics Chairman Lee Jae-yong. It was Arthur Mensch, CEO of Mistral AI, a French startup that had been founded only three years prior. At the time, the market valuations of the two companies differed by several dozen times. As I mentioned in my column, the fact that this defied the usual diplomatic protocol—where seats are typically assigned to business leaders of similar standing—confirmed that the meeting was no coincidence.
On the 8th (local time), Mistral closed its Series D round, marking the largest single round of funding in the history of European tech companies. The amount reached 3 billion euros (approximately 4.7 trillion won). The company broke its own record, set just a year ago, by increasing the scale by 1.5 times. Its corporate valuation has now surpassed 21 billion euros (approximately 32 trillion won). The lead investor for this record-breaking round was none other than Samsung Electronics. Samsung had previously invested in Mistral alongside companies like Nvidia in 2024, but this time, it took the lead in orchestrating the round.
This funding announcement came during President Yoon Suk-yeol's state visit to Paris (September 6–9). The deal between Samsung and Mistral was finalized between the two summit diplomacy events. Reports that Samsung was considering an investment of up to 1 billion euros in Mistral had already surfaced last July.
The Banner of Sovereign AI
Mistral is an AI startup founded in Paris in 2023. Established by engineers from Google DeepMind and Meta, the company has rapidly positioned itself as Europe's leading challenger to OpenAI by releasing a succession of open-source AI models. From its inception, the company adopted the thesis that "Europe can take the lead in the AI race" as its core business model. This is what is known as ‘Sovereign AI.’ It is a concept where nations and companies control not only their AI models but also the locations where data is processed, the infrastructure upon which models run, computing resources, and operational authority. At an AI summit in India last February, CEO Mensch stated, "Anyone operating AI workloads must hold the power button themselves. You should not depend on an external provider who can turn that button off."
In reality, Mistral's customers are responding to this concept. The focus on Mistral by the European Commission, French government agencies, as well as German companies like SAP and Helsing, is rooted in Europe's data regulatory environment—such as the GDPR—concerns over cross-border data transfers, and the desire to secure control over AI infrastructure. With this new capital, Mistral has announced plans to build 1 GW (gigawatt) of computing infrastructure within Europe by 2030.
Looking at the surface of this investment, the narrative is clear: it is read as proof of Europe's determination not to remain dependent on U.S. Big Tech AI. However, once you examine the list of investors, the story becomes more complex.
Sovereignty of Capital vs. Sovereignty of Technology
The round was led by Samsung Electronics, with the EU-backed Scaleup Europe Fund managed by EQT participating as a co-lead. Up to this point, the narrative reads as "Asian capital supporting European AI."
The complication arises thereafter. a16z (Andreessen Horowitz), Nvidia, Salesforce Ventures, and BlackRock are all included in the investor list. a16z is a prominent American venture capital firm that also holds a stake in OpenAI. BlackRock is the largest asset manager in the U.S., with assets under management exceeding 10 quadrillion won. Nvidia is a company with a dominant share in the data center GPU market, which is essential for training and running AI models.
Europe has bet 3 billion euros on ‘European AI’ to reduce its dependence on U.S. AI. Yet, the investor list for Mistral includes U.S. capital, and the core GPUs required to train and run the AI models are supplied by U.S. companies. If so, what exactly is the ‘sovereignty’ referred to in Sovereign AI?
Mistral explains that the nationality of capital and the sovereignty of AI are separate issues. They argue that the essence of sovereignty is not about whose money was received, but about where the data is processed, who holds the operational decision-making power, and which country’s laws are applied.
This logic is persuasive. Mistral is a French legal entity operating under the European regulatory framework. The fact that U.S. investors are shareholders does not turn Mistral into an American company. What truly matters for Sovereign AI is not the nationality of the investors, but the country where the data is processed and the laws are applied.
However, structural limitations persist, specifically regarding the relationship between capital and technology. For instance, Nvidia is an investor in Mistral while simultaneously being at the center of the supply chain for the core GPUs used by Mistral to train its large-scale AI models. If the U.S. government strengthens semiconductor export controls, Mistral's development speed could be directly affected. In fact, the U.S. government has continuously tightened export controls on high-performance AI chips since 2023. Whether ‘sovereignty’ can be fully maintained at this point remains an open question.
Imperfect Sovereignty, Yet Still Valid
Even though there is room for concern, this investment holds clear significance. Mistral is defining the market for Sovereign AI demand itself and pursuing a strategy to set the standards for that market. As long as European governments and companies are reluctant to entrust their data to U.S. AI, the demand for this market exists. The fact that it raised 3 billion euros is proof of that. That U.S. VCs participated in this deal is, paradoxically, a signal that they also take this market seriously.
Samsung's participation as a lead investor in a core European Sovereign AI company can be read as a signal that South Korean firms are no longer just observing AI technology from outside the U.S.-China technology hegemony, but are beginning to participate directly with capital and technology in the process of building European technological sovereignty.
Within South Korea, discussions surrounding Sovereign AI still seem to remain at a superficial level. The primary interest is in how many GPUs have been secured and how the performance of domestic AI models compares to those of the U.S. and China. However, the real questions of Sovereign AI do not end there. Issues such as whether necessary computing resources can be secured if U.S. semiconductor export controls are tightened, or which country's laws and government access rights apply when government and public data stored in Korea passes through the clouds and AI services of foreign companies, have not yet been sufficiently discussed in the public sphere.
Even if the answer presented by Mistral is not perfect, the questions they have raised are precise. What is AI sovereignty? Is it the source of capital, the control of technology, or where the data is processed? What answer is Korea preparing?
The author, Lee Jung-woo, has spent 17 years as a journalist covering various fields, including major industries such as automobiles, secondary batteries, and heavy industry, as well as defense, diplomacy, environment, education, and health and welfare. He has extensively covered industrial structural changes centered on mobility, energy transition, and sustainability from the front lines. He currently resides in Berlin, Germany, and serves as a partner at the startup accelerator '123 Factory'.