On Europe's Influence on the AI Value Chain

There's two things that inspired us to write this blog post: on one hand the recent introduction of deepseek and it's effect on the stock exchange, which was covered a lot on the media, as well as our recent effort to write a white paper on agentic AI and the role it might play in the future according to industry leaders. We felt the media coverage was in many aspects missing the point of who and what contributes to the AI value chain and in addition we find that it was portrayed as a battle between the US and China. Since we are actively working with many companies introducing AI into their operations and have connections to many European innovators in AI, we decided to blog about how we see the AI value chain and the influence that Europe has on it. 

AI Value Chain by midjourney

Let's start by looking at the value chain itself: often times when people in media talk about AI, they're merely talking about OpenAI's ChatGPT, Meta's Llama and how everything runs on NVIDIA chips. There's more to it though. NVIDIA for example is "only" designing the chips, they're made by Taiwanese company TSMC, which in turn relies on lithography printers from Dutch company ASML. And while many people are using apps like ChatGPT on a daily basis, the companies behind them make the majority of their revenue with B2B services, like including Copilot in a number of Microsoft products, adding license fees to their corporate customers, or services built on top of language models like automation (e.g. UiPath, which was started in Romania) or chatbots (e.g. Cognigy, which is headquartered in Germany). There's even European language models like France's Mistral. Granted, none of these get the media coverage or stock valuations like NVIDIA.

Selection of companies that we feel are important players in the AI Value Chain.

It is our understanding that language models, rather than being the main event, are rather important components in a larger puzzle. From automated handling of emails or incoming calls, to inserting data into ERM or CRM systems, all the way to intelligent agents handling entire business processes by themselves, this is where we believe most of the value is, and while those applications are clearly powered by language models, it is not crucial which model is chosen, as there are many which are sufficient. When looking at the share prices of companies working in that field (as an example we picked Microsoft, Apple, SAP and UiPath) over the last 2 years, we find that they appear correlated and are riding the growth expectations of AI. With SAP performing slightly better and UiPath being the most volatile. This is probably due to them being the newest, and "only" focusing on automation, where the others are all established in a multitude of fields. 

Share prices of Microsoft (Copilot), Apple, SAP and UiPath over the last two years.

Comparing companies from our infrastructure and supply chain layers reveals what many analysts have said for some time now. The utter difference in how the chip designer NVIDIA is expected to grow compared to how their chip maker TSMC is evaluated, shows that there is something wrong with investors expectations. Personally we believe that NVIDIA shares are in a bubble, which would explain the huge impact deepseeks benchmark results had on their share price. Add to that, that the PE ratios of NVDA and TSM are 70 (!) and 11 respectively at the time of this writing, and one can easily see how there is a bubble. (We wouldn't dare shorting it though, we're just observing).

Share prices of NVIDIA, TSMC and ASML over the last two years.

Lastly we take a look at some of the most important AI companies in China, namely Alibaba, Tencent and Baidu. Note that Deepseek is actually owned by 幻方. It is apparent that their shares don't follow the same growth expectations as their western counterparts, even though they're working on similar AI applications, which in our belief will be able to compete with western services in terms of their capabilities. The difference in share price expectations probably stems from them operating in a different market.

Share prices of Alibaba, Tencent and Baidu over the last two years.

In conclusion, the AI value chain is vast and there are many interesting aspects to it, beyond the few that get most of the media coverage. One can find European heavy weights in all the relevant layers, and of course they are deeply connected, e.g. German based Cognigy just recently got a new funding round involving US based investors.

There's many more mention worthy companies, but we wanted to keep the blog post concise. Just like there are many more layers in the value chain, e.g. consulting firms which would be the layer above the application/automation or e.g. the kind of AI research which is done in Universities.