Denmark's Gefion: The AI supercomputer that puts society first

Can it help us reimagine what "AI success" looks like?

Nov 20, 2024

Hi AIport readers,

I’m happy to bring you a local story with global implications.

It’s a look at Denmark’s new AI supercomputer, Gefion, and how it may serve as a more socially conscious model for investing in AI infrastructure.

From the way it’s been funded to the way it’s powered, Gefion challenges how we might measure “returns” on AI investments — is it short-term profits or long-term societal impact?

Thanks to Avi and Alex for having me.

Enjoy!

- Daniel Nest

In late October 2024, NVIDIA’s Jensen Huang and Denmark’s King Frederik X symbolically plugged in the country’s new AI supercomputer, Gefion.

It was quite a badass moment:

Watching the event unfold, you’d be forgiven for dismissing it as yet another story of investors throwing money at AI in the mad race to capitalize on the hype.

Yet if you look more closely at Gefion, you might walk away with a different narrative.

Could Gefion offer a different vision for socially responsible AI than the current product-driven one?

What makes Gefion different?

Granted, Gefion isn’t the first, largest, or fastest supercomputer in existence.

But I believe it is a symbol of a more conscientious approach to AI than we typically see.

Here’s why.

1. Societal impact vs. monetization

First, let’s take a quick look at the six pilot projects selected for Gefion’s trial run.1

See if you can spot a pattern:

“Unravelling CO2 reduction in Non-Metal Formate Dehydrogenase (FDH) using Machine-Learned Force Fields.” (Technical University of Denmark).

Reducing greenhouse gas emissions and supporting sustainable fuel development.

“Large-scale distributed simulation of quantum algorithms for quantifying molecular recognition processes” (University of Copenhagen).

The findings should help with future drug discovery.

“Multimodal Genomic Foundation Model” (University of Copenhagen)

Foundation for future developments in precision medicine and targeted treatments.

“Multi-Modal Document Understanding: Transforming Data Entry with Multi-Modal Precision” (Go Autonomous)

Optimizing the company’s models to be more energy efficient, among other things.

“Building an AI Care Companion with Large Video Pretraining.” (Teton and University of Copenhagen

Developing an AI-powered caregiving system to support medical staff.

“SAPIEN – Skilful Atmospheric Prediction with Intelligent Environmental Networks.” (Danish Meteorological Institute)

Improving weather prediction models and gauging environmental impact.

So, what do these projects have in common, aside from their titles sailing over most of our heads?

Hint: Almost all of them are either driven by public research institutions or dedicated to ESG (Environmental, Social, and Governance) initiatives — often both.

This puts Gefion in stark contrast to the purely profit-driven AI investments we’re used to seeing.

As the Danish Minister for Business Morten Bødskov puts it:

“[Gefion] represents an important opportunity for Danish companies and researchers to develop innovative solutions to future challenges.”

Gefion’s initial focus on societal projects isn’t just a happy accident, either.

ESG is built into the core of the entire initiative.

That’s because…

2. Public-private cooperation vs. venture capital

The ongoing wave of investments in AI startups comes primarily from venture capital and large corporations.

Gefion, on the other hand, is the result of intentional public-private cooperation.

The supercomputer is owned and operated by the newly formed Danish Center for AI Innovation (DCAI), which in turn was established through a collaboration of two key players:

The Novo Nordisk Foundation: an enterprise foundation that supports scientific, humanitarian, and social causes.
The Export and Investment Fund of Denmark (EIFO), a state-owned financial institution that supports various projects, but especially those that “facilitate the global green transition.”

These two organizations provide the bulk of Gefion’s funding.

DCAI is also quite explicit about the project’s focus on broader societal impacts.

While Gefion will eventually be open to anyone on a GPU-based fee model, DCAI clearly states that it’s especially looking to “catalyze certain application areas, including life sciences, healthcare, the green transition, and fault-tolerant quantum computing.”

This public-private backing makes Gefion less reliant on turning a profit and therefore less susceptible to the whims of the market.

3. Powered by renewable energy

Here’s the thing about AI models: They eat up a lot of energy.

One tidbit you might’ve heard is that a ChatGPT query consumes 10X the energy of a Google search.

That’s not to mention the costs to train these models in the first place.

There’s growing concern that AI labs and tech firms are putting a strain on the energy grid and undermining efforts to slow climate change.

Recently, major players have turned their attention to nuclear reactors to support the energy-intensive processes involved in training and running AI models.

But these projects have a timeline of 5+ years.

Gefion, right from the get-go, has carbon neutrality baked in.

The computer runs on a data center matched with 100% renewable energy. It’s one of 152 such data centers provided by Digital Reality, a US public company that invests heavily in making a complete switch to renewables.

According to DCAI, “the goal is to minimize [the environmental footprint] while maximizing the impact of the end results from these activities. DCAI and Digital Realty are committed to ensuring that the operations of Gefion are as sustainable as possible.”

Does Gefion put a crack in the “EU falling behind” narrative?

All of this got me thinking.

For the past two years, we’ve been increasingly seeing headlines like these:

The typical story goes like this: Europe got complacent, regulated itself to death, and lost out on AI talent and investments to business-friendly countries with less red tape like the US and China.

Here’s the thing, though: In most cases, the unspoken success criteria are profit and business competitiveness.

“...none of the world’s top 10 start-up ecosystems are in Europe. Given that AI offers such a huge potential to lift economic performance…it seems to us it’s pretty damn important if it can play such a crucial role in lifting the fortunes of Europe that it should be as widely accessible as possible”
- Nick Clegg, Meta’s Head of Global Affairs.

But what if such metrics for evaluating who’s “falling behind” are misleading, or at least incomplete?

What if we shifted our focus from simply “more AI” to “more of the right kind of AI”?

To me, one way to look at Gefion is to see it as a symbol of an alternative path for AI and its potential. A path that helps Denmark and the EU remain an attractive destination for AI talent while staying true to the region’s long-term societal and environmental goals.

Bjarke Ruse Sejersen, Founder & CEO of Go Autonomous, certainly sees the commercial value of a project like Gefion:

“Today, many are moving activities abroad...to raise the necessary capital, access infrastructure like Gefion, and attract the required talent. Gefion could help shift this trend. It could contribute to making Denmark a magnet for talent and leading companies in the field of AI."

But at the same time, with its focus on ESG projects, its public-private funding, and its reliance on green energy, Gefion also poses an implicit question:

What kind of a world do we want to build with AI?

I don’t know about you, but in today’s “AI gold rush” era, I find it refreshing to see a project that puts short-term financial gains on the back burner and sets its sights on our collective future.

I’m not alone in this.

In a panel discussion of the long-term vision for Gefion, Mikkel Flyverborn, Professor of Communication and Digital Transformations at Copenhagen Business School, is quite explicit:

“I think the people in the future will look back at us and ask, ‘What did these people use their new superpowers for? Did they solve their problem or did they create new problems?’ And right now a lot of people are engaging in technology development that actually doesn't support our societies, democracies, children's well-being, and so on. So I think we need to keep in mind that history will judge us.”

Can Gefion give history a reason to be proud of us?