Energy supply shouldn’t slow the growth of AI datacenters

Energy shortages are unlikely to hinder the expansion of AI datacenters, recent research by Ark Invest suggests. Although growth in electricity production has averaged ~2.7% at a global annual rate over the last five years, ARK’s research estimates that incremental demand from AI datacenters will be 0.7 percentage points, bringing growth in global electricity demand to 3.4% at a compound annual rate by 2030, as shown below.

According to ARK’s research, electricity represents only ~9% of the total costs of AI datacenters, leaving ample room for companies to invest in fast, off-grid power solutions without damaging datacenter economics, especially given the high returns on investment we expect from AI advances.

There is consensus among some experts that the impact of AI on energy systems will depend on both demand and efficiency. There is an urgent need to figure out how to manage datacenters in a way that maximises the use of renewable energy, said Chris Noble, co-founder and CEO of Cirrus Nexus, a cloud computing management company that uses datacenters owned by Google, Microsoft and Amazon.

Today, most datacenters seek to operate in a ‘steady state’, so that their energy consumption is reasonably stable. This leaves them at the mercy of the grid to which they are connected and whatever the daily mix of natural gas, nuclear power and renewable energy generation. The Financial Times reported in May that AI datacenters in the US already demand 15 gigawatts (GW) of energy annually – or almost the capacity of all US solar farms.

Technology giants are already the biggest buyers of renewable energy. Microsoft, for example, is trying to add mostly renewable energy to the grid and is investing heavily in carbon sequestration, which the U.S. Geological Survey defines as the process of capturing and storing atmospheric carbon dioxide. And the tech giant is doing more: ‘Our commitment to having our business powered 100 per cent of the time by 100 per cent zero-carbon energy by 2030 is also guiding our work on utility-scale battery storage, grid transformation and environmental justice,’ said Michelle Lancaster, senior director of global strategy at Microsoft.

It’s interesting to see technology companies looking for ways to reduce their dependence on more polluting energy sources. Changing datacenter operations is a creative way of harnessing excess renewable energy, helping to reduce the carbon footprint. This attitude shows a commitment to sustainability and can positively influence other industries to consider more responsible ways of operating.

Efficiency improvements in the chip’s hardware can also have a big impact on reducing energy use. This year, NVIDIA launched a new line of GPUs (graphics processing units) with 25 times lower power consumption than its previous models.

As far as possible, large companies say they want to remain carbon neutral, even when they increase their energy use. They are not succeeding at the moment. But they all know that, despite rapid advances in efficiency in chip models and AI, the energy use associated with AI will increase. In January, the International Energy Agency (IEA) predicted that global demand for electricity in datacenters will more than double from 2022 to 2026, with AI playing an important role in this increase.

According to the latest calculations, however, this increase will still represent a very small share of overall electricity use in rapidly decarbonising energy systems. Even with the projected growth in AI energy demand by 2030, datacenter use for AI in the US will represent around 0.4% of US emissions in 2022.

‘While the headlines may paint a grim picture of AI’s carbon footprint, the reality is that its direct impact is more of a blip than a seismic shift in our energy landscape,’ notes Dr Sarah Chen, a leading researcher in sustainable technology.

If anything, regulators are already taking action. Singapore announced a sustainability standard for datacenters in tropical countries last year, and the European Commission has moved towards regulating the sustainability of datacenters across the EU. In the US, Massachusetts Senator Ed Markey introduced a bill to study the environmental impacts of AI in February, and the House Energy and Commerce Committee held a hearing on AI energy use in June.

Changes are underway

Cushman & Wakefield’s fifth annual global datacenter market comparison ranking concludes that as AI has been shown to substantially increase demand for datacenters around the world, both site selection strategies and datacenter projects are being altered.

‘With interest in cloud expansion and the deployment of artificial intelligence and machine learning, datacenter developers have rushed to take large sites with a clear path to power,’ says Ali Greenwood, Managing Director at Cushman & Wakefield. ‘2023 ended as a record year in terms of market growth, tightening vacancy rates and rapid absorption in datacenter space. The stage is set for a 2024 full of emerging markets and new solutions to the energy puzzle,’ he adds.

The largest markets are positioned for continued growth, at least until there is pressure on local grid capacity or political imperatives change.

A report recently published by Dell’Oro Group predicts that spending on AI infrastructure will exceed one trillion dollars in the next five years, based on global datacenter investment growing at a compound annual growth rate (CAGR) of 24% until 2028. ‘AI has the potential to generate more than a trillion dollars in AI-related infrastructure spending in cloud and enterprise datacenters over the next five years,’ said Baron Fung, Senior Research Director at Dell’Oro Group.

The positive side of AI

Although AI could increase global energy consumption in the short term, its long-term potential as a decarbonisation tool is immense. We are witnessing a technological revolution that could finally decouple economic growth from emissions.

Take the UK, for example, which has managed to increase its GDP per capita by almost 50 per cent since 1990 while halving domestic emissions. AI could be the key to accelerating this trend on a global level.

However, to unlock the full potential of AI as a climate solution, stronger climate policies are needed, including:

• Putting a price on carbon to encourage AI solutions that minimise emissions;
• Double support for clean energy;
• Create regulatory structures that incentivise the development of AI for sustainability.

‘If we play our cards right, AI could be our trump card in the fight against climate change,’ says Emma Thompson, policy advisor at the Global Climate Initiative.

The future is in our hands. By harnessing the power of AI responsibly and strategically, we have the opportunity to create a more sustainable world. It’s time to embrace AI not just as a technological marvel, but as a powerful ally in our most critical global challenge.