Why Jensen Huang Can't Help European AI

At the recent GTC Berlin conference, Nvidia founder Jensen Huang issued this call to Germany and Europe. Huang's European trip took him to several countries, including France, the United Kingdom, and Germany.

Huang and Nvidia can offer only limited help for the technological prosperity of Europe's homegrown artificial intelligence. The core reason is not that Nvidia's technology is insufficiently advanced, but precisely that Nvidia is too powerful.

The full-stack AI factory Nvidia delivers to Europe is an integrated platform, not a single high-performance chip. This system covers GPU hardware, CUDA-X libraries, and models. These high-performance chips, CUDA-X, and the model layer are all things European companies are incapable of participating in building. Even if European companies accept it, all they can do is have companies such as Germany's Siemens or France's L'Oréal build their own application layers on top of it.

Nvidia also cannot spur European companies to develop similar competing products.

Nvidia's capabilities are so outstanding that almost no European company competes with it head-on. This is very different from the situation in the United States and China.

In the United States, AMD and Intel develop accelerator cards, and internet giants Amazon and Google also develop their own dedicated chips; China, subject to restrictions on imports of high-end chips, is growing a domestic computing ecosystem made up of companies such as Huawei, Cambricon, and MetaX.

Market competition spurs technological progress and economic development. Nvidia's hefty profits are obvious and should attract more European companies to compete on the same stage. But European companies seem to have given up their ambitions and are unwilling to step forward and fight.

The same is true of chips, and similarly of large models.

As far as I know, some large German companies operating in China connect at the underlying level either to models from America's Anthropic and OpenAI or to Chinese Alibaba models. Europe's representative model company is France's Mistral AI, but it currently shows no sign of gaining global recognition.

Europe's workplace environment is also unconducive to technological iteration. A Chinese researcher who visited Mistral AI told me that while Chinese and American researchers burn the midnight oil pushing large models to the frontier, Mistral AI's employee vacation system seems overly lavish.

Europe's reserve of AI talent is not as abundant as that of the United States and China. Nokia once stood at its zenith in the mobile phone sector, and Siemens also once had a mobile phone business, but both missed the mobile internet wave, and the corresponding software and hardware talent lost the soil to grow in Europe.

New technologies and innovative companies do not arise out of thin air; creators often come from the previous generation of technology and industry. The Transformer architecture that laid the foundation of artificial intelligence was born at the hands of Google employees, and Google was the king of the mobile internet era. The career résumé of Anthropic founder Dario Amodei even includes Baidu, Google, and OpenAI.

Financing for European startups is also less convenient than in China and the United States.

Huang said he will invest an additional $2 billion in Europe in the near term, after having already invested $8 billion in Europe. This may make Nvidia the largest investor in European artificial intelligence and ease the financing difficulties of local startups.

This money is spread across multiple countries including France, Germany, and the United Kingdom, involving more than ten or even dozens of companies, and the average amount each company receives will not be large. The companies Nvidia invests in are mostly in application scenarios for Nvidia products such as healthcare, business software, and robotics.

The key to artificial intelligence lies in chips, models, and data. U.S. companies in these fields routinely raise billions or tens of billions of dollars in financing—for example, OpenAI is now seeking $30 billion in financing; Chinese company MetaX raised more than 10 billion yuan before its listing, and DeepSeek is reported to be raising 100 billion yuan.

Even if AI seedlings are born in Europe, they will not easily find sufficient funding locally in Europe comparable to that in the United States and China.

In 2014, when DeepMind was acquired by Google, Europe's AI financing predicament was already obvious. The company was founded in the United Kingdom, and founder Demis Hassabis tried to keep it independent, but it was difficult to obtain venture capital support in the UK, Europe's financial center, and only afterward was it taken in by Google. Demis Hassabis also joined Google and in 2024 won the Nobel Prize in Chemistry for "protein structure prediction."

The United States and China are the two major centers of this wave of artificial intelligence, while Europe is in a relatively marginal position, caused by multiple factors including historical industrial ruptures, insufficient talent accumulation, and a lack of funding. Huang hopes Nvidia's AI factories will be accepted by European companies, and he has also brought a large sum of money. But these have only a limited effect on generating a homegrown AI ecosystem in Europe.

Europe's pace in catching up with the United States or China in artificial intelligence seems an almost impossible task. The best approach for European companies remains to embrace American and Chinese technology and attempt breakthroughs at the application layer. Some European companies have already been actively opening robotics research centers in China, conducting R&D together with Chinese companies to connect the links of this emerging industry.

Europe should no longer over-regulate artificial intelligence, especially when the United States and China make technological progress, rather than regulation, a priority. Necessary regulation is for safety, but excessive regulation may be needless worry and shut innovation out.

As Huang said, if an industry has not even developed, then there is nothing to regulate.