The AI investment boom is creating a second group of winners beyond chipmakers: companies supplying transformers, switchgear, backup power and advanced cooling systems.
McKinsey forecasts nearly $7 trillion of global data-centre investment by 2030. That construction wave is colliding with long waiting times for grid connections and essential electrical equipment.
Why power is the real bottleneck
An AI server is useful only when a facility can power and cool it continuously. Developers may want new capacity delivered within six months, but grid connections can require two years in emerging markets and much longer in some developed economies.
Demand is therefore rising for high-voltage transformers, uninterruptible power supplies and systems that remove heat from dense GPU racks. Liquid cooling is gaining attention because traditional air cooling becomes less efficient as computing density increases.
Asian manufacturers are benefiting from the trend, but rapid competition and component shortages can pressure margins. Investors should distinguish companies with firm orders and manufacturing capacity from those benefiting mainly from AI-themed expectations.
What could slow the boom
Electricity prices, community opposition, financing costs and shortages of skilled engineers could delay projects. Some proposed data centres may also be speculative, making order backlogs less reliable than they first appear.
The durable opportunity lies in equipment required across multiple chip generations. GPUs may change quickly, but every data centre still needs reliable power distribution, cooling and physical infrastructure.
