孫正義豪砸 750 億歐元落腳法國！SoftBank 蓋 5GW AI 資料中心，歐洲最大算力賭注開打

On the evening of May 30, Paris time, a press release from the French Ministry of Economy squeezed into a name that is rarely noticed in the tech circle - Dunkirk. This northern French port, close to the Belgian border, is famous for its massive evacuation of 300,000 soldiers during World War II. However, what it needs to undertake now is not the evacuation of soldiers, but rather the tens of billions of euros in SoftBank's pocket and a whole array of AI data centers that will soon rise.

The numbers announced by SoftBank are staggering, even for those who are used to big news: up to €75 billion, approximately $87 billion, will be invested to build a total of 5GW (gigawatt, or 1 billion watts) of AI computing capacity in France. This is the largest infrastructure investment commitment made by the Japanese SoftBank Group in Europe, and it officially brings the competition for "who can build enough computing power" from the United States, specifically Texas and Arizona, to the European continent. For Taiwanese readers, this news may seem distant at first, but in reality, almost every server, cooling system, cabinet, and power module in each data center has a Taiwanese manufacturer's name behind it.

Event Background

Over the past two years, the main battlefield of the global AI competition has quietly shifted from "whose model is smarter" to "who has enough electricity and computing power to train and serve." No matter how strong the model is, without sufficient GPU and data center support, it's like a sports car without a racetrack.

SoftBank plays a unique role in this game. On the one hand, it is an important investor and major customer of OpenAI. On the other hand, through its subsidiaries, such as Arm and SB Energy, it extends from chip architecture to energy and infrastructure. In other words, Masayoshi Son is not just betting on a specific AI application, but rather wants to be the one "selling shovels to the entire gold rush."

France is another key player. French President Emmanuel Macron has been actively positioning the country as the "European AI hub" in recent years. France has a card that others envy - nuclear power. About 70% of France's electricity comes from nuclear energy, which means low carbon emissions and relatively stable prices. This perfectly addresses the two most painful points for AI data centers: high energy consumption and the need to be environmentally friendly. French Economy Minister Roland Lescure stated in a statement that this investment proves France's ability to undertake large-scale digital infrastructure projects, relying on "competitive, sovereign, and low-carbon electricity." This statement is almost a footnote to the entire investment case.

Key Points

• Total investment of up to €75 billion (approximately $87 billion): Used to develop and operate up to 5GW of AI data center capacity in France, which SoftBank itself considers the group's largest AI infrastructure investment in Europe.
• Initial investment of €45 billion, 3.1GW: The first phase focuses on the Hauts-de-France region in northern France, expected to be completed and online by 2031.
• Three locations: Loon-Plage, Bosquel, and Bouchain in Dunkirk will be the sites of the first batch of data centers.
• Energy partnership: Collaborating with French electricity company EDF on the Bouchain project, and with Schneider Electric to build an industrial cluster in the port of Dunkirk, including local production of data center servers, power modules, and other equipment.
• Job creation commitment: SoftBank states that it will create thousands of high-skilled job openings, covering data center development, engineering, energy systems, robotics, operations, and advanced manufacturing.
• Masayoshi Son's statement: He said in a statement that AI is entering a new era, and "the country that builds the foundation for this transformation will shape the future of technology" - a message that is, to some extent, directed at all governments still watching from the sidelines.

Market Impact Analysis

For Taiwanese users: You may not directly walk into the data centers in Dunkirk, but you will feel their presence in your daily life. As European computing power supply increases, services like Gemini, ChatGPT, and Perplexity may see improvements in latency, stability, and availability in the European region, which could also impact the overall experience for Asian users. More importantly, the increase in computing power supply may reduce the unit cost of AI services over time - a practical benefit for Taiwanese users who frequently use AI for writing, research, and creating prompts.

For enterprises: For Taiwanese companies, especially those in the electronics and power supply chain, this is a visible order map. 5GW is equivalent to the full-load output of several nuclear power plants, requiring a massive amount of servers, switches, liquid cooling systems, power modules, and cabinets. While Schneider Electric will assemble power modules locally, the components and upstream server manufacturing still heavily rely on Taiwanese companies. For Taiwanese companies planning to expand into Europe, this is a signal: placing data and AI workloads in "low-carbon, sovereign" European nodes is becoming a compliance and ESG (Environmental, Social, and Governance) bonus, worth considering in advance.

For developers: For developers, the long-term effect of such large-scale construction is the "bargaining space for cloud computing power." As SoftBank, cloud service providers, and governments around the world rush to build data centers, GPU and computing resources are moving from extreme scarcity to relative abundance. The most direct benefit is the reduction in training and inference costs, and the increase in available high-end GPUs for rent. For engineers using Cursor to write programs and integrate various model APIs, this means that the structure of cloud bills may become more flexible in the future, and AI products that need data sovereignty in Europe will have more deployment options.

Future Development Trends

The true significance of this investment lies not in the €75 billion figure itself, but in the three trend lines it reveals.

First, the bottleneck of the AI competition has officially shifted from "chips" to "electricity." When data centers are measured in gigawatts, energy strategy becomes AI strategy. France's reliance on nuclear power to win this round will lead to more countries using "cheap and clean electricity" as a lure for investment, and the bond between energy and AI will only become tighter.

Second, sovereign AI is becoming a new political language. Europe has long been concerned about over-reliance on American tech giants, and this collaboration between Japanese funds, French land and electricity, and European local manufacturing is essentially building an option that is "not entirely subject to American cloud services." It can be expected that similar "localized computing power" projects will emerge in Germany, the UK, and even Asia.

Third, this is a marathon that will last for several years. The first phase is expected to be completed by 2031, meaning that there is a long period of engineering, power supply, and approval risks between today's commitments and actual production capacity. Whether the data centers will be completed on schedule and whether AI demand can support such a large supply are still unknown. For supply chain manufacturers, this is a long-term topic, not a story of overnight wealth.

TheAI Academy Summary and Commentary

Masayoshi Son is known for making big bets while others are still watching. This time, choosing to invest in Europe at a time when the need for computing power is most urgent and France is most eager to prove itself is a well-timed move. However, I would like to remind everyone that large-scale infrastructure projects are most likely to see a gap between "commitment" and "delivery," and the €75 billion is an upper limit, a vision. The actual landing speed and scale will depend on the execution over the next few years.

Commentary: This is not an AI investment, but a bet on an era where "electricity equals computing power"; those who understand will look for the shovel, not the gold.

For Taiwanese readers, the specific suggestion is: if you are a professional in the supply chain, put "European data center localization" on your observation list this year, especially paying attention to the demand for local production of power supplies, cooling systems, and servers in Europe; if you are a general user or content creator, do not worry about this big news, focus on mastering AI tools, accumulating useful prompts and workflows, which is the true way to enjoy the benefits of increased computing power.

Data Sources

• SoftBank Group official press release: https://group.softbank/en/news/press/20260531_0
• TechCrunch report: https://techcrunch.com/2026/05/30/softbank-says-it-will-invest-up-to-e75-billion-to-build-french-data-centers/
• CNBC report: https://www.cnbc.com/2026/05/31/softbank-to-build-up-ai-data-centers-in-france-with-major-investment.html

This article is based on publicly available information and official announcements. The investment amount is the planned upper limit and multi-stage commitment, and the actual landing scale, timeline, and benefits are still subject to engineering, power supply, and approval variables. This article is for industry analysis and information integration only and does not constitute any investment advice.