79 % der befragten Energieversorger haben KI eingeführt – meist noch als Pilotprojekt.

Most energy AI is still a pilot

In Germany, 79% of energy suppliers have introduced AI, but 61% of those uses are still at pilot stage, according to PwC's energy AI study. That gap is the central finding of the Good AI Prism Sector Brief: Energy & Utilities. The value sits in operations, and most of the AI has not reached them yet.

The upside is large where AI is deployed well. The IEA's Energy and AI report finds 30–50% shorter outages with AI-based fault detection, up to 175 GW of additional transmission capacity without new lines, and incidents detected 500 times faster with satellites and sensors. Utilities use AI most for demand forecasting (62%), predictive maintenance and asset monitoring (53%) and customer service (48%), L.E.K. reports from a survey of more than 300 energy executives.

AI needs power, and power needs people

Energy meets AI from two sides. Data centres used 415 TWh in 2024, about 1.5% of world electricity, and the IEA projects about 945 TWh by 2030; their use grew 17% in 2025. In Ireland, data centres took 23% of metered electricity in 2025 (CSO). German data centres used about 20 TWh in 2024, with about 31 TWh expected by 2030 (Bitkom and Borderstep).

The workforce is the second bottleneck. 76 million people work in energy, but in advanced economies 2.4 workers near retirement for every new entrant under 25, and more than half of organisations report critical hiring bottlenecks (IEA, 2025). In Germany, 80% of municipal utilities name the skills shortage as a challenge, while 82% now see AI as an instrument of their digital transformation, up from 61% a year earlier (BDEW and EY).

The grid bottleneck has a price

Germany spends about €3bn a year on grid congestion management (Bundesnetzagentur, 2025). In the US, only 13% of grid connection requests filed between 2000 and 2019 reached operation (LBNL). Forecasting, congestion management and fault detection are where AI can show its value in the numbers a grid operator already tracks.

The rules: high-risk duties for safety components from Dec 2027

From 2 Dec 2027, AI used as a safety component in managing and operating the supply of electricity, gas, heat or water carries high-risk duties under Annex III of the AI Act, with dates as amended by Regulation (EU) 2026/1744. Systems that only assist or optimise are not automatically safety components. Data centres from 500 kW already report energy and sustainability indicators each year under EU energy-efficiency rules.

Germany adds cybersecurity and resilience duties: the NIS2 implementation act, the KRITIS umbrella act and the smart-meter rollout, where the Bundesnetzagentur opened proceedings against 77 operators in March 2026. In the UK, Ofgem set expectations on governance, explainability and human oversight; in Texas, Senate Bill 6 lets large loads of 75 MW or more be curtailed in emergencies.

What the market offers

The Good AI Prism places 24 entries in Energy & Utilities: 23 AI offerings and one unclassified record (Tibber). Seven manage flexibility, demand response and customer energy, among them Kraken, Tibber and Sympower. Six forecast renewables and trade energy, such as enercast and Volue. Six plan and operate grids, including envelio, GridOS and Gridscale X. Three optimise buildings and industrial energy use, such as BrainBox AI and EcoStruxure, and two monitor renewable assets. 14 of the 24 entries come from providers headquartered in Europe, five of them in Germany. The Energy & Utilities area shows every entry with its source and date.

What to do Monday

1. List every AI use and model, from forecasting to customer service, including tools staff bring themselves.
2. Mark the uses that act as safety components in supply and map all uses to AI, cyber and critical-infrastructure rules.
3. Bring in operations, IT/OT security, data protection and the works council before the pilot, not after.
4. Name accountable operators with override rights and document AI literacy measures.
5. Run one forecasting, maintenance or flexibility pilot with a baseline, and scale only on measured forecast error, congestion cost, outage minutes and customer outcomes.