Is AI a short-term energy guzzler but a long-term saver?

As the debate on the uses, drawbacks and ultimate developmental direction of AI rages on, it’s clear that the world-changing technology is currently both a help and a hindrance to boosting the overall sustainability of our global energy sector. While offering an invaluable chance to enhance the efficiency and resilience of grids across the world, there’s no getting away from the fact that AI carries a hefty electricity demand of its own.

Powerful yet power hungry – AI’s energy bill is booming

The immediate downside of AI’s rapid usage growth and ever-widening set of industrial applications is that it is sucking up energy at unprecedented rates. In late January, the Financial Times released a report that demonstrated how global electricity demand from data centres could surpass 1,000 terawatt hours by 2026. This is double the level of 2022, when data centres accounted for 2% of total global energy consumption. Jumping to around 4% in just four years represents a frightening hike in demand and the emissions that go hand in hand with greater consumption.

While data centres are rightly considered to be the “backbone of our digital lives”, responsible for supporting everything from public WiFi networks to medical patient data, AI (along with cryptocurrency, the FT notes) is driving 

an explosion of the industry’s electricity consumption. Using ChatGPT, for example, can be almost 10 times as energy intensive as Google search. This is the lower end estimate on a sliding scale of peer-reviewed reports that suggest the actual difference could be 28, 68, or even 236 times higher.

Clearly, there is a question of timing to consider. AI is an emerging, breakthrough technology that is still very far from achieving its full potential and efficiency; denouncing its energy-guzzling qualities may be a touch unfair at this point, given its rapidly evolving state. However, at a crucial moment in human history where emissions need to be driven down quickly and technology is routinely viewed as our potential solution, it’s both ironic and somewhat demoralising to see AI hiking up global energy consumption so significantly.

Putting the industry on a path to decarbonisation – AI is essential for rapid efficiency gains

Fortunately, there is plenty of confidence that AI’s positive contributions to the energy industry will quickly surpass its running costs in terms of emissions and energy consumption. Experts within the industry and outside observers agree that AI is poised to revolutionise energy systems’ predictive maintenance, system efficiency and potentially even drive greater international energy security while driving down emissions.

The AI market is already worth an estimated $13 billion to the global energy sector, and serves over 50 use cases in power plants, national grids and energy production facilities around the world. This is just the tip of the iceberg. Predictive AI (the more industry-specific sibling of generative AI, which is all about generating new content and ideas) is set to assume a central role in the setup of next-generation facilities, as such systems can be fed mountains of data that allow for better decision making in real time while also planning ahead. System repairs or maintenance actions that could take  teams of human engineers weeks to identify, plan and execute will soon become the work of mere hours or minutes when enabled by AI systems integrated with drones, automated solutions and, of course, qualified humans to oversee it all. With improved oversight and response capabilities will come unprecedented efficiency gains, leading to the industry’s decarbonisation at scale and at pace.

Another crucial element of delivering this highly efficient and decarbonised new energy setup is the thorny problem of predicting supply and demand. As renewable energy production ramps up, along with the adoption of electric vehicles (EVs) and residential solar, both supply and demand levels are fluctuating much more quickly. Add the increase of extreme weather into the mix, and you have a much harder time balancing the grid. AI can help here too. Its ability to analyse vast datasets and usage patterns, while plotting in factors like weather, will eventually arm grid operators with the necessary insight to pre-emptively balance the scales. This will ensure, for example, that fossil fuel facilities are only in use when they are absolutely needed, along will a host of other efficiency boosters that reduce waste and the production of needless carbon emissions.

In it for the long term

AI is seemingly here to stay. The energy industry, which is under increasing political and societal pressure become greener while keeping pace with increasing demand, cannot afford to miss the opportunity to fully realise and understand the extent of its own production facilities, grid elements and the activities of its end users. 

The world is demanding that the industry decarbonise as quickly as possible; an outcome that seems impossible without the timely integration of human intuition and the predictive power of AI-enabled data crunching. The crucial question is how long it will take for AI to reach the tipping point where the efficiency gains it brings will outweigh its own carbon footprint.