AI is consuming electricity ⚡️ and the world is reorganizing its data centers 🌍
- Ralph Schwehr

- Sep 23
- 3 min read
Back in Germany, I'm becoming increasingly aware that AI is completely reshaping the global economy. Artificial intelligence is the new heavy industry, data centers are the factories, and electricity is its raw material.
However, Germany is starting with ballast: Energy is simply too expensive and uncompetitive in this country. Even with wholesale customer rates ("firm power"), Germany is at a disadvantage. While hyperscalers in the UAE or the USA can secure 1 GW of continuous load for under USD 900 million per year , operators in Germany pay up to twice as much, even with subsidized tariffs.
“Anyone who doesn’t supply electricity competitively today loses compute and thus AI sovereignty.” Ralph Schwehr

Here is a comparison of electricity prices
region | Electricity price for industry (basis) | “Firm Power” wholesale tariff | Costs for 1 GW/year (8,760,000 MWh) | Sources |
🇦🇪 UAE | ~ 0.11 USD/kWh | ~ 0.05–0.09 USD/kWh | ≈ USD 438 million–788 million | |
🇺🇸 USA | ~ 0.07–0.12 USD/kWh | ~ 0.05–0.10 USD/kWh | ≈ USD 438 million–876 million | |
🇩🇪 Germany | ~0.16 €/kWh (~0.17-0.18 USD) | ~ 0.08–0.12 €/kWh | ≈ USD 680 million–1.02 billion |
UAE: ☀️🌴 (solar + nuclear power)
USA: ⚡🏭 (grid load & nuclear power)
Germany: 🏗️📉 (construction site, high costs)
Global energy strategies for AI
🇦🇪 UAE - Compute sovereignty with solar & nuclear power
The UAE is relying on predictable, low-carbon base load and affordable solar power to build an AI-capable digital infrastructure: Barakah now supplies up to 5.6 GW of nuclear power and covers about a quarter of the nation's electricity demand, while Dubai's solar park continues to grow (+1 GW in September alone). This reduces risks for hyperscalers, supported by capital (including G42/Microsoft, Khazna financing) and proactive industrial policy.
🇺🇸 USA - AI boom puts pressure on the internet
The US is experiencing a historic surge in demand: The PJM grid operator expects around 30 GW of additional peak load by 2030, almost exclusively from data centers. Tech companies are responding with "firm power" deals, prominently Microsoft's 20-year PPA for the restart of Three Mile Island (835 MW), as well as pilot projects such as Enhanced Geothermal. Strategically, think tanks are advising better load forecasting and faster grid connections.
🇩🇪 Germany - Efficiency requirement meets grid emergency
Europe/Germany is facing headwinds: The efficiency and waste heat quotas of the EnEfG (German Energy Efficiency Act) are sensible ways of tightening standards, but challenging to implement. At the same time, bottlenecks and bureaucracy are slowing down expansion: In Frankfurt, large-scale grid connections will only become realistic from the mid-2030s onward; operators are considering their own gas-fired power plants to even bring projects online.
The result: Locations with secure, competitive energy, such as the UAE or certain US regions, are attracting AI capacities, while Germany is in danger of falling behind despite high demand.

Key messages
UAE: Nuclear power (5.6 GW) + solar = predictable compute sovereignty.
Microsoft & G42: Investments worth billions ensure location attractiveness.
Khazna: $2.6 billion for data center in Masdar City.
USA/PJM: +32 GW peak load by 2030 AI as the main driver.
Microsoft: 20-year PPA reboots Three Mile Island.
USA/WRI: Recommendations: Accelerate load forecasting and grid connections.
Germany/EnEfG: PUE ≤ 1.2 & waste heat requirement ambitious, difficult to implement.
Germany/Network: Frankfurt: Large-scale connections only from 2030+.
Germany/Strategy: Operators plan their own fossil power plants.
Global: Locations with secure, clean energy attract AI capabilities.
📚 Sources
Reuters (16.04.2024): Microsoft invests $1.5 billion in G42 (UAE)
World Nuclear News (03/01/2024): Barakah: 4th reactor started - 5.6 GW nuclear power
PV Magazine (03.09.2025): Dubai's mega solar park grows: +1 GW
DataCenterDynamics (05.09.2025): Khazna: 2.6 billion USD financing for data centers
PJM (01/24/2025): Long-Term Load Forecast: +32 GW by 2030
Microsoft Blog (20.09.2024): Three Mile Island reboot via 20-year PPA
WRI (September 17, 2025): US load increase due to AI data centers – Analysis & Recommendations
Bird & Bird (04.06.2024): EnEfG: Efficiency & waste heat requirements
Clean Energy Wire (25.07.2025): Data center operators plan fossil power plants
Golem.de (18.08.2025): Frankfurt: Large-scale connections only from the mid-2030s
GlobalPetrolPrices (2025): UAE – Electricity Prices
Khaleej Times (2024): Reduced Electricity Rates for Industries
Utility Dive (2025): Adapting Utility Tariffs for Data Centers
Clean Energy Wire (2024): Germany Industrial Power Price Debate
CubeConcepts (2024): Europe's Industrial Electricity Prices
DODS Report (2024): Germany's Subsidy Plan for Industrial Power
🎯 Conclusion
The race for AI infrastructure is brutally clear: Whoever has electricity has compute. Whoever has compute has a future. The UAE and the USA are securing pole position with billions in investments and competitive energy supplies. Germany, on the other hand, is threatening to take itself out of the game in the grid and cost quagmire.
👉 Act now: Request the OAKAI deep dive and/or use OAKAI as a think tank . We provide information, timelines, and sources precisely tailored to your AI roadmap.



Comments