Will Data Centers be the Biggest Challenge for AI in 2025?

As artificial intelligence innovation rapidly accelerates and companies adopt the technology across the board, demand is growing exponentially for the data that informs the large language models that drive AI. This means that even more energy-hungry data centers, packed with processors, must be built to fuel the expansion. ABI Research suggests that, by the end of this year, there will be 5,709 public data centers worldwide (5,186 colocation sites and 523 hyperscale sites).

In fact, Barclays Research suggests that AI demands will more than double data center electricity needs by 2030 (based on current grid capacity).The International Energy Agency (IEA), who work with governments and industries to shape a sustainable energy policy, have stated that “in 2023, overall capital investment by Google, Microsoft and Amazon, which are industry leaders in AI adoption and data center installation, was higher than that of the entire US oil and gas industry – totaling around 0.5% of US GDP.”

So, with this level of outlay, it’s important to understand why this investment in data centers is happening, and where it’s taking place.

What’s driving increased consumption?

Oracle’s Chairman, Larry Ellison, has stated that data center construction is being driven by the dual factors of rising GPU demand and increased power requirements – and this has resulted in a shift to liquid cooling: “This AI race is going to go on for a long time. It's not a matter of getting ahead, just simply getting ahead in AI, but you also have to keep your model current. And that's going to take larger and larger data centers.”

But AI data usage can vary

The amount of data used by AI is very much dependent on the model being used. So, smaller machine learning models are likely to only need hundreds or thousands of data points, while more complex deep learning models, like deep learning networks, will typically require large datasets, often ranging from thousands to millions of examples – especially for tasks like image and speech recognition.

The large datasets that help the model learn better also require more storage and processing power, with the quality and quantity of training data directly impacting the performance of the AI. Ergo, as AI becomes more powerful, more storage will be required. That said, each new generation of AI is also likely to be more power efficient than the last.

Where are these data centers being built?

AI data centers are now being built all over the world, with the biggest investments taking place in regions that offer the best conditions for data processing – so those with access to renewable energy, low latency, and proximity to major tech hubs.

Asia-Pacific currently has the highest concentration of data center locations, followed by Europe and North America – where Big Tech companies like Microsoft, Google, Amazon and Facebook, are expanding their data center infrastructures in states like Virginia, Texas, and Washington (northern Virginia’s “Data Center Alley” is known as the data center capital of the world).

Meanwhile, China is now investing heavily in AI and is building a number of data centers in Beijing, Shanghai, and Shenzhen as part of its national strategy to become a leader in AI technology. And in Europe, countries like Ireland, the Netherlands, and Germany are seeing an expansion in data center construction, with Dublin and Amsterdam particularly popular due to favorable regulations and tax incentives.

Objections and concerns

Then there is the thorny issue of where to actually station data center infrastructure. In the UK, there is often the issue of ‘nimbyism’, i.e. when the local population revolt against something they deem to be inappropriate in their locale. This manifested itself recently in England, where a proposed new data center situated on the outskirts of the village of Abbots Langley was rejected by local authorities after villagers opposed the plans.

This issue has also been raised in the US with Julie Bolthouse, director of land use at the Piedmont Environmental Council in Virginia, outlining the sheer scale of the data center operation in the state: “They’re so huge and imposing on a human scale. It’s basically a giant, concrete computer. And there’s the wires you can see over your head, and all throughout Data Center Alley now are substations and transmission lines and giant concrete boxes.” Meanwhile, in August of this year, the FT reported that water consumption by dozens of facilities in the state has risen by almost two-thirds since 2019 – a significant environmental concern.

AI will transform future energy systems

All of this and more will be discussed in detail at the IEA’s Global Conference on Energy and AI, which will take place on 4th and 5th December. Here representatives from governments, technology, the energy industry and civil society, will discuss how AI could transform energy systems in the future. One thing is certain though: with more advanced AI taking off, the demand for new data centers will continue to grow. Building these will require greater consideration and regulation in order for data demands to be realized while also taking into account environmental and land use concerns.