Denmark needs a national long-term strategy for chip design

If Denmark wants to be a significant player in the European chip agenda, we must focus on our strengths and particularly pursue them, such as edge AI. Although the EU Chips Act will channel money to the member states, it will be far from enough to truly establish an ecosystem for developing and producing microchips in the race with the USA and Asia.

This was one of the conclusions when the Nordic region’s largest deep-tech fair and conference, Digital Tech Summit, in Øksnehallen in Copenhagen, which featured a panel debate between representatives from Meta, NVIDIA, the hearing aid manufacturer WSA, Aarhus University, and DTU Compute with Head of Department Jan Madsen and Moderator Jørgen Kragh Jakobsen, who is an analogue circuit designer and owner of the one-man company IC Works.

Since advanced microchips are a crucial prerequisite for digitalization, ensuring fast and efficient data processing, it is natural that the EU Chips Act is a significant topic when more than 4,500 people from universities and industry gather for a conference in one of the world’s most digitalized countries.

The EU’s vulnerability was revealed - we are too dependent

Ahead of the debate, Minister of Industry, Business and Financial Affairs in Denmark, Morten Bødskov gave a keynote speech, where he pointed out the rapid change and views on chip design in few years from everything is great to a global race in defining and developing new technologies.  

“Before the COVID-19 pandemic, I guess it's fair to say also that we did not worry that much about these, so to say, tiny plates of silicon. But it has changed since then. Everybody is focused on producing and developing chips now. Suddenly, it became very clear that most of our advanced microchips came from just one place, Taiwan. And that setup is simply too risky. We need more and we need more different suppliers of chips to make sure that we are able to address the critical needs that we have and the critical materials that we need. Otherwise, we cannot innovate, we cannot compete, and we will simply fall behind. We will lose jobs, we will lose investments, we will lose money, and we will lose possibilities to develop the good side of our society. Because the world today has not become more stable or more reliable. We did not return back, after the pandemic to the good old time, to the secure and stable supply chains that we saw before the COVID pandemic. Today, we are looking into geopolitical tensions, and they will define the times that we're living in now,” Morten Bødskov said. 

The risk of dependency called for a joint response from the European Union nearly three years ago with the European Chips act. China and the US had already invested fully in the microchip agenda. The new European Commission has made it a priority to follow up on the European chips act and they proposed a European chips plan. But how can Denmark contribute to that agenda? Well, the German government has for example set aside five billion euros in public funding for only one (chip factory, editor) factory.

“There's no doubt that we cannot follow that track when it's only about public money. We have to find ways where we actually can make a difference as we have done so many times before in our history,” the minister said.

Scepticism about the EU Chips Act - not ambitious enough

It aligned with the message in the debate after, when Michael Smedegaard, a Danish national who's had his entire career in the USA, now as Chief Audio HW Architect at Meta, mentioned that the EU have allocated 43 billion euros to the chips act, but it is far too little money. Which is one of the major points of criticism that he had heard about the EU Chips Act.

The goals are viewed as too modest compared with investments made in the USA and Asia. Insufficient funding as part of the same coin, allocated budget can be seen as too small to make a significant impact on EU's semiconductor industry. How EU spend the money is really important, and a discussion at national level in the EU is necessary to figure out how to strengths of the ecosystem in each specific country in the EU, he said.

Another element of criticism is that the EU Chips Act prioritizes research over production, neglecting the need for large-scale investments. This will delay the work.

The Chips Act doesn't address the brain drain of engineers leaving for the US.

The final point of criticism is that it's got a slow timeline where both the implementation and also the funding is seen as too slow to keep up with the rapid pace of the development in the semiconductor industry. The EU is up against the free market forces of private companies or mostly private, part public companies like TSMC.

Go for the white space – how about Edge AI?

“I think we need to take a hard look at our core competencies across the spectrum of the entire tech sector and identify areas that contain more white space than the red ocean. When we talk about AI and chips, everybody is thinking about NVIDIA at the moment. And that's a runaway success story. Can you repeat that? Probably not if you try. But if you invest in all the areas where Denmark has potential, it's a good likelihood that a new NVIDIA, if you measure it as a success, will emerge from that ecosystem. So, I think a public debate about how we do this, how we allocate the funds, what our core competencies are, is needed,” Michael Smedegaard said and pointed out one area, where Denmark could play a great role: Edge AI. 

Edge AI/computing involves running smaller artificial intelligence models directly on devices rather than relying on cloud-based solutions. While large language models (LLMs) like Meta’s Llama or OpenAI’s ChatGPT have 70-80 billion parameters, there are also smaller models with fewer than 10 billion parameters. These smaller models can operate locally on devices using application-specific semiconductors/chips, which is a niche but growing area. This approach reduces the need for constant cloud connectivity and can enhance performance and privacy.

“And let's also not forget that most of the successful semiconductor companies are fabless. You don't need a fab to become a semiconductor company. Apple doesn't have a fab. NVIDIA doesn't have a fab. Broadcom, AMD, etc. So, the notion of what you need and what you need to invest and where you get the maximum output from your investments, that's a very detailed debate that I think that we must have, Michael Smedegaard said.