DDI Opinion: What Gordon Moore Would Say About “Enough Data Centers”

In the Danish debate about new data centers, one argument keeps coming back: we already have the data centers we need, so let's stop here.

I understand where that comes from. AI is moving fast, many people feel uneasy about where it is taking us, and data centers have become the most visible, physical symbol of that unease. You can't see an algorithm. You can see a large building next to a field. 

But the premise doesn't hold. And as an industry, we owe people a better conversation than the one we are having. 

We have been here before 

In the 1970s, computing meant IBM mainframes. Large, expensive machines in air conditioned rooms, owned by a few big companies, universities and public authorities. Few imagined that ordinary people would ever need computing power of their own. 

Then came the personal computer and the internet. Every company got its own server room in the basement, and computing spread into thousands of small, inefficient installations, often running half empty and cooled by whatever was at hand. 

When cloud computing arrived in the late 2000s, the debate sounded familiar. Is it safe to put our data in someone else's building? Do we lose control? Is it just hype? And will electricity use spiral out of control? 

What happened instead was one of the biggest efficiency gains in the history of computing. As workloads moved from scattered server rooms into purpose built data centers, the International Energy Agency's tracking of 2010 to 2020 shows internet traffic and data center workloads multiplying many times over, while data center electricity use stayed at around 1 percent of global demand. That was the decade before AI. 

What Moore's law actually taught us 

In 1965, Gordon Moore predicted that the number of components on a chip would keep doubling, and that this would lead to “such wonders as home computers”. At the time, that sounded absurd. He was right. For decades, the computing power on a chip roughly doubled every two years, and the energy needed per calculation fell dramatically. 

A widely cited study in Science found that between 2010 and 2018, the amount of computing done in the world's data centers grew by around 550 percent, while their energy use grew by around 6 percent. 

That is an extraordinary track record. But efficiency did not remove the need for new capacity. It made it possible to keep up with demand at all. At every turn, someone could have said we have all the computing we need. Each time, new uses appeared that nobody had predicted, and the answer was not to stop building but to build smarter. Each new generation of data centers is more efficient than the last, so freezing the building stock today is not the climate friendly choice. It freezes our digital infrastructure at yesterday's efficiency. 

Demand does not disappear when we say no. Hospital systems, bank transfers, research, public digital services and MitID all run in data centers, and rarely in just one. A single mobile payment can pass between a bank’s own servers and commercial cloud several times in seconds, so the line between “critical” capacity and everything else is blurrier than it looks. If that computing does not happen in Denmark, it happens elsewhere, often on grids with more fossil power and with less Danish influence over how our data is handled. 

AI does require us to think differently 

AI is not just more of the same. Training and running large AI models takes far more power per rack, often needs liquid cooling, and puts pressure on the grid at a scale we have not planned for before. The IEA expects global data center electricity use to more than double by 2030, with AI as the main driver. 

Efficiency will keep improving, through specialised chips, liquid cooling and models that do more with less, and our industry has every reason to keep pushing. But history also shows that when computing gets cheaper, we use more of it. So the past gives us reason for confidence, not complacency. AI asks us to plan grid capacity, location and energy supply together, to be clear about which needs we are building for, and to expect more of how new data centers contribute to the system they connect to. 

Part of the energy system, not just a user of it 

This is where Denmark has a real advantage. We have district heating networks, a large share of wind and solar power, and a long tradition of thinking in integrated systems. 

Excess heat can go to district heating and food production, as we already see in Odense and Albertslund. Data centers can be flexible consumers that adjust to the needs of the grid instead of adding to the strain at the worst moments. And large, long term electricity customers can help anchor investment in new renewables and grid infrastructure. 

That is the conversation I would like to have. Not whether we should have data centers, but what a good data center looks like, and what we as a society should expect of them. 

The fear we have become a symbol of 

Behind much of the resistance is a deeper fear, and it deserves to be taken seriously. People fear that AI will change or replace jobs. Pointing to the number of jobs a new data center creates locally does not answer that. It is a fear about whether there will be a place for them in what comes next. 

The tech sector must take its share of the responsibility for bringing people along. That means investing in upskilling and reskilling, working with vocational schools and universities on the education new jobs require, partnering with municipalities, and being present and transparent in the communities where data centers are built. It also means making sure the value data centers create is shared fairly, both with the communities that host them and with society at large. Nationally, tech companies, government, unions and the education sector need to sit at the same table and own the transition together. 

Denmark has done this before. Our transitions have worked when people could see a way across and when the gains were shared broadly. Nobody should be left standing on the platform while the train leaves. 

Building the next ones better

We don't have all the data centers we need. We have the ones yesterday needed. The task ahead is to build the next ones better: more efficient, integrated into the energy system, and with people on board, locally and nationally. I welcome that conversation, also with those who see it differently.

____



DDI Opinion is a monthly series where members and guests share their view on current topics, challenges and technologies in the data center industry. Have an idea for a piece? Send your proposal to camilla@datacenterindustrien.dk