Feiyu Xu: Europe needs its own AI ecosystem

Ms. Xu, several important industries in Germany are in crisis. Can AI close a growth gap here?

Yes, AI is a huge opportunity for Germany. It contributes to developing intelligent or new products and helps optimize processes to increase efficiency and cut costs.

Is there even any room left for Germany and Europe in the AI race between the US and China?

We in Germany are too modest. We haven't been aggressive enough in the last few years. If you look at basic AI research – deep learning, machine translation, etc. – German scientists rank among the best in the world. But to drive commercialization, AI adoption by the industry, we have not invested enough. That's why we should now invest heavily in infrastructure, research and innovation. And then, I believe, we can quickly return to being among the best in the world. You can't reap without investing.

Where exactly are China and the USA ahead?

The AI ecosystem consists of chips and the hardware infrastructure, the cloud infrastructure, the infrastructure for training AI models, and the training and deployment infrastructure for AI models. It is rounded off by the applications that make AI technologies usable in practice. All of this is the AI technology stack. In terms of hyperscalers – the large cloud infrastructure used to enable scaling in Big Data and cloud computing – the United States is leading with companies such as Google, AWS, Microsoft and Meta. But China is also at the forefront with Alibaba, Tencent and Baidu. Around 60 percent of large language model companies are based in the USA and over 30 percent in China.

And where does Europe stand?

Europe only has a very limited number of companies developing large language models. We also don't have a hyperscaler in Europe. Many German and European companies develop AI applications based on US and Chinese AI stacks. The EU, national governments, and businesses investing in start-ups must work together to fill the gaps in the AI stack. Many are aware of this, but there is a lack of investment. Talking about it or complaining that we are perhaps not as far along as the US or China will not cut it.

Access to data is essential for AI development. Do we need to reconsider our regulatory framework?

With the GDPR or the Data Act, Europe is playing a leading role in the careful handling of data. The Data Act also aims to encourage companies to share data in order to build applications. China, but also other countries, take Europe as an example. But regulation alone is not enough. Additionally, companies need to be offered services and technologies to help them find out, for example, which data they can use, which not, and which technology is accessible under regulation.

What could such services look like in practice?

The EU could offer an app. Developers could then run the data through this app, and it would tell them: This data contains personal data and cannot be used, or: This use case is safe. I think offering something like this is very important, especially for small and medium-sized enterprises.

To what extent can cooperation with China in the development of AI work – against the backdrop of the escalating tech conflict with the US? Do we have to choose which tech stack we want to be in so that we are not potentially sanctioned for cooperating with China?

Cooperation with China is a very abstract issue. Many companies are global companies and offer products or services to the Chinese market. Conversely, many Chinese companies operate globally. If you don't want to miss out on the Chinese market, you must take a close look. Where are you in the AI ecosystem? At what level do you need partnerships? If a company is operating in China, for example, it needs Chinese language models; for the US, it needs Western language models. It's about how you can operate as a global company in order to be successful. On the other hand, a closer look at AI research partnerships reveals that US universities collaborate the most with Chinese universities.

But doesn't this contradict current US policy, which prevents the export of certain AI chips or machines to China?

I think we German or European countries may need to learn to understand the Americans a bit better. What they say and what they do can be different. Apart from that, I think we're all curious to see what next year has in store for us.

Is European AI even possible without becoming entangled with the US, and without China? This question was one of the sticking points that led to the split in your previous company, Nyonic.

It is vital for Europe's sovereignty that we have our own AI technology stack. But globally operating companies still depend on international cooperation in order to be successful.

Which AI issue do you consider so relevant that you would set up a company for?

I am currently supporting start-up companies that support the development of generative AI for industrial applications. That was also Nyonic's original idea – to make generative AI successful for the European industry. The second area is the combination of AI with synthetic biology. We know how hard life is for patients who have kidney failure, liver failure or heart problems. If we can successfully combine AI with synthetic biology, we can improve the quality of life of many people and save many.

Let's return to regulations. Different regional regulations can cause tensions. Do we need global interoperability?

Absolutely. Humanity is in a very critical phase. However, we must always keep the most important goals for humanity in mind: Eradicating famines, creating better medical conditions for everyone and preserving our climate. This requires global cooperation. We have also seen this in the fight against the pandemic. That's why I also believe in cooperation on AI, because AI needs data. Let's look at rare diseases: There might only be a few cases in some countries, but many more in others. Without collaboration, efforts to fight them would not be successful.

Furthermore, we need an organization like the UN for AI regulation and data regulation to enable global collaboration on technology applications.

Do such initiatives already exist?

There are initiatives. Last year, the US, the EU, China and Australia, along with 28 other countries, held the first AI Safety Summit in the UK in November. The US and Chinese governments have also met twice to coordinate their AI regulations.

So, do we have to think bigger?

Absolutely!

Dr. Feiyu Xu studied at Tongji University in Shanghai and Saarland University and habilitated in Big Text Data Analytics. She has published more than 100 scientific papers and co-founded Nyonic, which she left in March 2024, and was Senior Vice President, Global Head of Artificial Intelligence at SAP from May 2020 to June 2023, where she led the company's AI strategy. Xu is Chairwoman of AsiaBerlin Forum e.V. and Non-Executive Director of Airbus Group, member of the Supervisory Board of ZF Group, and member of the Board of Directors of ChainIQ Group and Zühlke Group.