According to the German business newspaper "Handelsblatt" on June 9, Nvidia CEO Huang Renxun started his trip to the European Three Kingdoms this week. His first speech at London Science and Technology Week ended. He immediately rushed to Paris to host the GTC Technology Summit and will meet with German Chancellor Merz in Berlin.
This trip across Britain, France and Germany is interpreted by the industry as Nvidia's key move in the global AI computing power battle - when European countries spent tens of billions of euros to build local AI super factory, Huang Renxun was trying to weave a new ecological network with H200 chips and the "sovereign AI" strategy.
Europe's computing power dilemma and Nvidia's ambition
The current AI computing power gap in Europe is shocking: According to Statista data, 68% of the European AI computing power demand in 2024 rely on imported GPUs, while STMicroelectronics, the largest local chip manufacturer, has almost no presence in the field of advanced processes.
This structural imbalance forces the German government to separate 5 billion euros in its 2025 budget to support the construction of local AI infrastructure. But what is embarrassing is that even if Europe spends 8.5 billion euros to build the "European largest AI park" in the suburbs of Paris, its core computing power will still be supported by Nvidia H200 chips - the reasoning speed of this chip is nearly 1 times higher than the previous generation, and the video memory bandwidth is as high as 4.8TB/s, almost monopolizing the global high-end AI computing power market.
Double variations of technical hegemony and geopolitical game
The signals released by Huang Renxun this trip are intriguing: at the VivaTech Summit in France, he announced with the Paris Institute of Integrated Technology to develop the "European version of the model"; and in Berlin, his talks with the German Chancellor focused on the "path of autonomous AI computing power."
This seemingly contradictory statement is actually in line with Nvidia's "sovereign AI" strategy - this strategy has been implemented in Singapore, Japan and other countries. By helping countries establish basic models for local training, it not only avoids geopolitical risks, but also deeply binds the customer ecosystem. What is more noteworthy is that Nvidia is pushing European companies to adopt its Blackwell architecture chips, which accounted for 70% of data center revenue in Q1 2025, forming a de facto monopoly on technical standards.
Butterfly effect of industrial chain reconstruction
Europe's computing power layout is triggering a chain reaction: British AI chip company Graphcore was forced to reduce its business in China and lay off employees because it could not obtain Nvidia-level government support; while French start-ups such as Mistral AI chose to cooperate with Nvidia in exchange for computing power resources. Behind this "dependence innovation" is the deep crisis in the European semiconductor industry - its chip production is expected to drop from 8.1% to 5.9% globally, and there is a lack of independent processor and memory chip companies. What’s even more serious is that even if the AI super factory is built in Europe, its software ecosystem will still be subject to the Nvidia CUDA system. This ecological barrier is exactly what Huang Renxun repeatedly emphasized at the London Technology Week.
Conclusion
When Huang Renxun discussed "European AI sovereignty" with Merc at the Prime Minister's Office in Berlin, the Brandenburg Gate outside the window witnessed a game of computing power. Can Europe take the path of independence under the shadow of Nvidia's technology? When the H200 chip roars in the Paris park, is Europe's AI dream put on wings or is it covered in shackles? Welcome to leave your opinions in the comment section.
(Content source: IT Home, German Business Daily "Handelsblatt"; data reference: Statista, Global Network, First Financial Daily)
