Masayoshi Son leads SoftBank to invest in French nuclear power. Can this cooperation with Schneider break through the dilemma of European computing power?
SoftBank to Invest in France to Build Computing Power Cluster, Join Hands with Schneider Electric in the AI Computing Power Battle
According to the latest information from Yujian Energy, on May 31st, SoftBank officially announced an investment of up to 75 billion euros in France to build a computing power cluster with an initial capacity of 3.1 gigawatts and a long - term target of 5 gigawatts. The project is located in the Hauts - de - France region, and SoftBank will join hands with Schneider Electric in Dunkirk to create an AI infrastructure and robot manufacturing industry center.
Before this, in April this year, the meeting between French President Emmanuel Macron and Masayoshi Son of SoftBank Group had laid the groundwork for this investment.
This is not the first time Masayoshi Son has made such a decision.
This 68 - year - old Japanese of Korean descent has achieved great success with Yahoo and Alibaba, but also suffered setbacks with WeWork and Uber. From 2021 to 2022, SoftBank's Vision Fund suffered a total loss of more than $50 billion, and Masayoshi Son publicly admitted that it was "the biggest mistake". However, after 2023, he placed his bets on AI. The listing of ARM, which he bought for $31 billion, brought a paper profit of more than $260 billion, and he also invested $30 billion in OpenAI, becoming the largest investor outside of Microsoft. Now, SoftBank's market value has reached 33 trillion yen, and Masayoshi Son's net worth has doubled to $67.1 billion in half a year.
This time, his gambling table has been moved to France.
Yujian Energy has always believed that Europe has been invisible in the global map of AI computing power for a long time. Data from the International Energy Agency shows that the average annual growth rate of global data center power consumption from 2023 to 2026 will reach 18%, but the construction of computing power infrastructure in Europe lags far behind that of the United States and China.
When OpenAI's flagship project in northeastern England was put on hold indefinitely, and when AWS and Microsoft are expanding their data centers by hundreds of billions of dollars every year, Masayoshi Son's decision means that European computing power is no longer just a follower but is passively involved in the computing power battle.
However, building a data center is not just about throwing money. The reason why Schneider Electric was selected by SoftBank is that this French electrical giant holds the truly scarce asset in the era of AI computing power - the full - stack data center infrastructure capabilities from the power grid to the chips.
Yujian Energy believes that behind this event lies the intersection of liquid cooling technology, power management, and system integration.
Liquid Cooling Technology: A Must - Have for Computing Power Projects
The heat density of AI chips is rising at an astonishing speed.
Schneider Electric's newly launched MCDU - 70 cooling distribution unit has a single - unit cooling capacity of up to 2.5 megawatts. Through Schneider Electric's EcoStruxure software to build a centralized control system, it has an expansion capacity of over 10 megawatts. Schneider's end - to - end liquid cooling product portfolio covers the full range from 105kW to 2.5MW, and all CDUs have been certified by Schneider's global expert network for full - life - cycle support.
Schneider has emphasized in public that the success of data centers increasingly depends on large - scale, reliable, and efficient infrastructure solutions. As AI computing power develops, power supply and cooling have become the key variables determining whether a project can be implemented.
Yujian Energy believes that SoftBank's choice of Schneider is not a simple equipment purchase but a consideration of its comprehensive strength in the industry.
Although Schneider is not very popular in the domestic public opinion field, as an established foreign company, Schneider has been deeply involved in the global data center industry for decades. Its business scope covers all the necessary links for AI computing power projects, such as low - voltage power distribution, medium - voltage power distribution, building automation, industrial automation, and digital infrastructure, and can be said to be very reliable.
Judging from the data, Schneider also performs well. It is understood that in the fiscal year 2025, Schneider Electric's global revenue exceeded 40 billion euros for the first time, reaching 40.2 billion euros, with an organic growth of 8.9%. The data center business performed the most prominently among all sub - industries, with double - digit growth in both sales and potential demand. More importantly, Schneider has obtained the status of a cooling distribution device supplier recognized by NVIDIA, and its liquid cooling technology has been integrated into NVIDIA's GB200 NVL72 server system, which is a pass for upstream locking in the entire computing power industry chain.
In the words of a senior person in the data center industry, the threshold of liquid cooling technology does not lie in making a single device, but in whether it can be delivered in batches, whether it can pass the verification, and whether it can ensure operation throughout the entire life cycle.
Schneider completed the integration of Motivair in the past year. Its end - to - end liquid cooling portfolio covers the full range from 105kW to 2.5MW. Its products meet NVIDIA's latest hardware certification standards and are supplied synchronously at advanced manufacturing centers on three continents.
By teaming up with Schneider, SoftBank has filled the supply - chain gaps before building the computing power project.
From Investment to Real - World Business: SoftBank Needs to Do the Math
Looking deeper, the cooperation between SoftBank and Schneider is essentially to solve a financial problem.
It should be noted that 75 billion euros is far from enough to build a 5 - gigawatt data center. According to the industry's common algorithm, it takes about $50 billion to build a 1 - gigawatt data center, and 75 billion euros can only cover about 1.5 gigawatts. This means that the seemingly high - profile announcement is more like a starting point, and more external financing will surely be introduced later.
SoftBank CFO Yoshiaki Goto had previously clearly explained their financing logic: Super AI infrastructure projects will adopt the "project financing" model. Partners will inject 10% to 20% of the total investment in each project as equity, and the remaining 80% to 90% of the funds will be raised from channels such as banks, funds, pensions, and insurance companies through loans and other forms.
SoftBank is also shifting its business focus from investing in AI - related enterprises through funds to areas closer to "real - world business" such as building AI infrastructure. This means that SoftBank no longer needs scattered invested enterprises but system integrators that can deliver large - scale facilities end - to - end.
Schneider's role on this computing power chessboard is clear. It not only provides liquid cooling equipment and power distribution systems but, more importantly, provides the full - process integration ability from the power grid to the chips.
Schneider has cooperated with NVIDIA to develop multiple sets of reference designs centered on power and liquid cooling, covering power management, cooling control, high - density cabinet systems, and even digital twin simulation tools. That is to say, Schneider can solve all the problems of power distribution, cooling, control, and automation for SoftBank in the construction in Dunkirk in one stop. Binding Schneider is more cost - effective and time - saving for SoftBank than individual procurement.
The Endgame of Computing Power Infrastructure: It's All About Integration
Looking at the entire industry, the cooperation between SoftBank and Schneider reveals a deeper industry trend: that is, the era of AI computing power infrastructure has entered a stage where multi - party capabilities need to intersect.
Currently, no single company can independently complete a gigawatt - level data center. This requires nuclear power suppliers to solve the power source, electrical giants to handle power supply and heat dissipation, chip manufacturers to lock in GPU supply, and capital providers to cover the huge upfront investment.
This is an upgrade of the industrial form. Computing power giants are no longer just simple capital buyers but packagers of infrastructure and organizers of system integration.
Schneider has obtained more than just SoftBank's order. The industrial center in Dunkirk will integrate AI infrastructure and robot manufacturing, which means that Schneider can build a complete business ecosystem here, from equipment manufacturing to engineering services and then to operation and maintenance support.
For an electrical equipment manufacturer, this is equivalent to upgrading the traditional "equipment - selling" model to the role of a "long - term operation service provider", and both customer stickiness and the business model have undergone qualitative changes.
By partnering with Schneider, SoftBank has established a long - term anchor in Dunkirk.
The total planned computing power of the entire project reaches 5 gigawatts. Once completed, its power consumption will be equivalent to the total output of 5 standard nuclear power plants, approaching the peak power demand of New York City. This scale means that there will be continuous infrastructure expansion needs in the future. By having Schneider as a long - term partner in this link, SoftBank has a stable support for the entire computing power supply chain.
Of course, investment also has another side. SoftBank's debt scale has risen in recent years due to continuous investment in AI. If the growth of AI demand in the future fails to meet expectations, the project will face the risk of pressure on investment returns.
Schneider's global manufacturing and delivery capabilities are certainly an advantage, but whether the project can be truly implemented ultimately depends on the continuity of France's future policies, the access speed of the power grid, and SoftBank's actual capital arrangement.
Partnering with Schneider may only be the first step of a long - distance journey.
This article is from the WeChat official account "Yujian Energy", author: Wang Mengjiao, published by 36Kr with authorization.