Artificial intelligence is reshaping global electricity demand at unprecedented speed. Between transformer shortages, high-voltage cables, and nuclear output, ten listed companies embody the invisible energy infrastructure that makes AI possible. A data-driven tour of a theme now central to markets.
🔑 Key takeaways
- A large AI data center consumes as much electricity as 100,000 households, equivalent to 100-300 MW.
- The global data center fleet grew from 8,000 to 10,000 units between 2023 and 2025 (+25%).
- Microsoft, Alphabet, Amazon, and Meta will invest nearly $700 billion in AI infrastructure in 2026.
- Transformer and turbine lead times now reach 2 to 3 years.
- The IEA forecasts a doubling, BloombergNEF a tripling of data center power demand by 2030.
Unprecedented power demand
The AI boom imposes a new physical constraint on global power grids. A large data center built for model training or inference can draw between 100 and 300 megawatts, the equivalent of a medium-sized city. According to the International Energy Agency (IEA), such a site consumes as much electricity as 100,000 households.
The global fleet jumped from 8,000 to 10,000 units between 2023 and 2025, a 25% increase. This growth lifts data centers to 1-2% of worldwide electricity consumption. In the United States, their share already reaches 2%, on par with the aviation sector. In Ireland, operational sites account for over 17% of national demand.
Looking ahead, the IEA expects a doubling by 2030 while BloombergNEF projects a tripling — two estimates that frame a floor and a ceiling for equipment makers and power producers.
Supply chain bottlenecks
The electrical supply chain is under severe strain. Lead times for transformers, turbines, and circuit breakers now reach 2 to 3 years, locking in record order books for several players and limiting how fast new capacity can come online.
Battery energy storage (BESS) is growing more than 25% per year through 2030 according to BloombergNEF, providing room to smooth loads and stabilize grids. Meanwhile, hyperscalers confirmed combined AI infrastructure investments of nearly $700 billion for 2026, an amount that catalyzes the entire electrical value chain.
« Electricity has become the limiting factor for AI growth. The supply chain can no longer keep pace with orders. »
Sector analyst, BloombergNEF (cited by Socomec, 2026)
| Company | Ticker | Core segment | Data center exposure |
|---|---|---|---|
| Schneider Electric | SU | Energy management, UPS, cooling | >24% of revenue targeted |
| Legrand | LR | PDU, UPS, internal distribution | >20% of sales |
| Nexans | NEX | High-voltage and submarine cables | Multi-year backlog |
| Siemens Energy | ENR | Transformers, HVDC, gas turbines | Grid +41% YoY |
| GE Vernova | GEV | Gas turbines, US grid | US leader |
| Eaton | ETN | Switchgear, UPS, busway | Record backlog |
| Vertiv | VRT | Critical power, liquid cooling | Nvidia partner |
| Constellation Energy | CEG | US nuclear | Long-term PPAs |
| Vistra | VST | Nuclear + gas | More cyclical |
| HD Hyundai Electric | 267260.KS | EHV transformers | $5.185B 2026 target |
Ten stocks to ride the AI power wave
Analysis of the ten listed companies identified by sources reveals several exposure profiles, from equipment manufacturing to low-carbon power generation.
Schneider Electric (SU) is the global leader in energy management. It delivers the full electrical architecture of a data center — medium and low-voltage distribution, UPS (uninterruptible power supply) systems, and liquid cooling since the Motivair acquisition. The data center segment should exceed 24% of revenue in coming years.
Legrand (LR) equips internal infrastructure with power distribution units (PDUs), UPS systems, and power management tools. Over 20% of sales are tied to data centers. The valuation, viewed as stretched by several observers, already prices in part of the quality premium.
Nexans (NEX), the world’s second-largest cable maker, produces high-voltage cables and submarine interconnections linking power plants, wind farms, and data centers to the grid. This link, one of the tightest in the chain, offers multi-year visibility.
Siemens Energy (ENR) leverages its Grid Technologies division, which builds transformers, circuit breakers, and HVDC (high-voltage direct current) systems. Grid Technologies orders jumped more than 41% year-over-year.
GE Vernova (GEV), spun off from General Electric, is the backbone of US AI power. Its aeroderivative turbines, deployable in weeks, answer demand for fast, dispatchable capacity where grid connection lags behind.
Eaton (ETN) offers switchgear, UPS systems, and busway solutions. The Electrical Americas division shows strong backlog growth, largely AI-driven. The Boyd Thermal acquisition strengthens its cooling position.
Vertiv (VRT) supplies critical power and cooling infrastructure. The company collaborates with Nvidia on 800V DC architectures — a strong signal of its position at the heart of the AI roadmap.
Constellation Energy (CEG), the largest US nuclear operator, signs long-term power purchase agreements (PPAs) with major tech players, turning the operator into a strategic supplier of AI infrastructure.
Vistra (VST), an independent power producer, combines nuclear and gas, capturing both baseload and peak demand. More cyclical than Constellation, Vistra benefits from rising wholesale electricity prices.
HD Hyundai Electric (267260.KS), a global leader in extra-high-voltage transformers (765 kV), raised its 2026 order target to around $5.185 billion and is expanding its Alabama plant.
Beyond equipment makers: storage, water, and semiconductors
Battery energy storage systems (BESS) have shifted from a backup role to an active element of energy strategy. In mature markets like the UK, BESS cuts peak charges by 10 to 15% and stabilizes the grid with millisecond-level frequency regulation, turning data centers into grid-stabilizing actors.
The AI boom also pressures water and raw materials. A large data center can consume up to 19,000 cubic meters of water per day, equivalent to a city of 10,000 to 50,000 inhabitants. Germanium and gallium prices, essential for semiconductors, rose 25 to 30% in 2023-2024 after Chinese export restrictions. RAM prices jumped 172% between early and late 2025, and Nvidia announced no new GPU launch in 2026 — a first in five years.
On the regulatory side, the Netherlands restricted data center construction in 2022 above certain thresholds, with Microsoft circumventing the limit via three separate towers, raising concerns in the Dutch parliament over grid capacity. In the Philippines, where 7% of the population still lacks electricity, AI investment inflows trigger grid instability and higher bills.
Conclusion
Electricity has emerged as the limiting factor for AI growth, ahead of GPUs and memory. Companies that produce, transmit, and distribute this power — from transformer makers to nuclear operators — form the invisible infrastructure of artificial intelligence and offer investment opportunities across the entire value chain.
For investors, two scenarios emerge. An optimistic scenario, aligned with BloombergNEF, sees demand tripling by 2030 and supporting prolonged re-rating of both equipment makers and nuclear producers. A cautious scenario, aligned with the IEA, forecasts a doubling, sufficient to keep order books at record levels but without a major upside break. In both cases, the gap will widen between players able to deliver and those choked by lead times.
Sources
- Cryptoast – AI Energy: the 10 best stocks to invest
- EasyBourse – Mapping a technological revolution
- Eyes-on-Europe – Data centers and AI: who pays the ecological bill
- Socomec – How energy storage boosts data center ROI
This article is for informational and educational purposes only. It does not constitute investment advice. Do your own research (DYOR) before making any decision.
