Artificial Intelligence (AI) is driving a new digital revolution and generating significant opportunities. It is revolutionizing energy systems by helping us improve consumption e efficiency, reduce emissions, and facilitate the integration of renewable energy sources. Although AI is raising concerns as regards today’s infrastructure and energy system or on its climate impact, we believe that they can be overcome because of the contributions AI is already making to our energy transition.

I believe that businesses and European decision makers can together ensure that AI does not become a barrier or stumbling block to the energy transition but a building block to enable new heights of green growth.

How AI challenges today’s energy systems

Meeting AI’s electricity demand

Data centers today are a small fraction of our energy needs: only 20% of the world’s energy consumption is electrified and data centers accounted for less than 2% of that electricity demand[1]. In turn, AI is only a fraction of data center demand. However, each of those is increasing considerably and AI faster than anything else.

In 2022, data centers used 460TWh of electricity; by 2026, they will need anywhere from 620TWh to nearly twice that much. In the words of the IEA, that’s “roughly equivalent to adding at least one Sweden or at most one Germany”.

The IEA’s estimate is among the most conservative, with Barclays forecasting double the annual growth[2].

Schneider Electric’s Sustainability Research Institute has worked on different scenarios on AI’s energy consumption. In a world of pure abundance without boundaries, we would expect AI energy consumption to increase thirteen-fold over a decade[3], even accounting for data centers being abandoned as new and more efficient centers came online.

Keeping pace with infrastructural needs

After decades of stable electricity demand, grids are not sufficiently equipped to handle the structural increases needed for the transition. Data centers are straining against a pre-existing bottleneck and strengthening the case for urgent investment in expanded and smarter distribution grids.

Ireland is emblematic of this. Today it hosts over 80 data centers with over 50 more on the way. In 2022, they used 17% of Ireland’s electricity (5.3TWh), more than every urban home in the country. At current pace of growth, they will need 32% of all of Ireland’s electricity[4], with all the corresponding challenges to the infrastructure.

Overcoming these challenges

There is a pressing need to ensure that AI technologies are provisioned sustainably. Schneider Electric is committed to Responsible AI, emphasizing the importance of measuring AI’s environmental footprint and collaborating with industry leaders to create energy-efficient data centers.

Data center efficiency

Energy centers have made dramatic strides in their energy efficiency from everything to the algorithms to the chips to the cooling and are set to make many more. In a data center, 40% of energy is consumed by the computing itself, being converted into waste heat. The computer effectively becomes an electric boiler and so another 40% must be used in cooling. The remaining 20% is everything else, including surrounding IT[5].

Early data center designs were inefficient and fan-cooled but as scale increased, so did density and efficiency. This gave way to the hyper-efficient systems of today, for which Schneider Electric provides end-to-end AI-ready electrification and cooling solutions that align the rise of AI with the energy efficiency needs for sustainability.

We’re hopefully at the beginning of a new wave of efficiency gains.

The European Union’s Energy Efficiency Directive now requires data center operators to publish basic data and to share more detailed data to the European Commission so that it can aggregate a snapshot of energy consumption across Europe.

Frugal, flexible, responsible use

There are different AI tasks (training; inferencing), different types of AI (generative; analytical), for different purposes (industrial; consumer), and on different timescales (immediate response; overnight processing). The greatest bottleneck will be on generative AI inferencing. This points us to the highest gain solutions: algorithmic efficiency in generative AI is expected to improve 4 times per year and mixed precision techniques could even make its energy consumption plateau; prioritizing high impact use cases can drive consumption down further (from an overall data center consumption of 1370TWh to half as much, or 785TWh); flexible or staggered use can allow that consumption to flexibly match renewable production.

Additionally, to be taken up widely, AI needs to be trusted. This can be achieved through governmental incentives, such as the EU AI Act and the upcoming Apply AI Strategy, but also through the development of corporate Responsible AI strategies and processes.

Cybersecurity best practice 

Widespread adoption hinges on building trust in AI solutions. While AI introduces new cybersecurity challenges, it can also enhance defensive capabilities through advanced detection and protection tools. Ongoing conversations around acceptable risks and the establishment of a robust cybersecurity culture are crucial for responsibly integrating AI into energy systems. 

Transforming Energy Systems with AI

The reason that AI can be a net positive for our transition, is because of its incredible ability to handle massive volumes of data, simplify complexity, and act autonomously to optimize energy systems, especially digitalized and electrified ones.

Optimizing for efficiency

AI has the capability to transform energy-intensive operations in factories, buildings, and district heating networks. District energy networks play a crucial role in achieving zero-carbon heating and cooling, with a goal of 350 million connections in urban areas worldwide by 2030, according to the IEA[6].

The growth of these systems is expected to meet approximately 20% of global space heating requirements.

What can AI do to help optimize this crucial energy system? For instance, a district heating utility Croatia, which serves over 8,000 residents, employs Schneider Electric’s District Energy system with an AI Load Forecasting module. This technology, which complements our district energy software, predicts heating demands, optimizes energy consumption, reduces both operational and maintenance expenses, and lowers emissions. This AI advancement is helping our customer forecast the amount of energy needed to heat the district, in turn allowing the district to optimize the operation of chillers, boilers, and energy storage systems.

Optimizing for flexibility and prosumers

Schneider Electric also makes software designed to bring the power of AI to flexibility, enabling prosumers (consumers producing their own energy) and grid demand-response.

We combine our EcoStruxure Building Operation software with our EcoStruxure Microgrid Advisor so that energy is produced, consumed or stored exactly where it needs to be. Large buildings are even able to go beyond net-zero by distributing excess energy to hundreds of homes nearby.

This is not just theoretical; it’s already happening. For instance, Schneider Electric partnered with Citycon on the Lippulaiva project in Finland, enhancing sustainability in the country’s second-largest urban shopping center. Utilizing the AI-enabled systems mentioned above, Lippulaiva became an energy prosumer, reducing annual CO2 emissions by 335 tons and achieving a 3-year investment payback. Today, it stands as Europe’s first energy-self-sufficient, sustainable, and carbon-neutral shopping center!

This will become more and more valuable as we integrate more variable renewables onto the grid, increasing the need for intelligent demand-response. In markets with high volatility or negative prices, we have already seen huge benefits in the application of solutions as EcoStruxure Microgrid Advisor.

Conclusion

By the end of the decade, we can expect AI to be integral to the energy sector, driving unprecedented efficiency and innovation.

This will be also about collaboration: no one can do it alone. At Schneider Electric, we partner with the world’s leading chip manufacturers to design reference architectures that support high compute workloads with optimized uptime and efficient cooling.

New use cases will emerge, and AI will play a crucial role in achieving sustainability goals. The energy transition is not just possible; it’s essential for a sustainable future.

While AI presents challenges, its transformative potential in the energy sector is immense – and the technology emerges as our best ally in the journey towards net zero. By developing and applying AI responsibly, we can revolutionize energy systems and create a more sustainable world.

[1] https://www.iea.org/reports/electricity-2024

[2] https://www.ib.barclays/our-insights/ai-revolution-meeting-massive-infrastructure-demand.html?cid=pressrelease_site_PoweringAI2_

[3] https://www.se.com/ww/en/insights/sustainability/sustainability-research-institute/artificial-intelligence-electricity-system-dynamics-approach/

[4] https://iea.blob.core.windows.net/assets/18f3ed24-4b26-4c83-a3d2-8a1be51c8cc8/Electricity2024-Analysisandforecastto2026.pdf

[5] Ibid.

[6] https://www.iea.org/reports/350-million-building-units-connected-to-district-energy-networks-by-2030-provide-about-20-of-space-heating-needs