Outlook theme 3 of 4
Energy for intelligence
Every answer an AI gives starts as electricity.
The big picture
A single large AI data-centre campus can draw as much power as a small city. As AI usage grows, so does the electricity needed to run and cool the chips - and building new power supply takes far longer than building a data centre.
That makes energy the slowest-moving and hardest bottleneck in the entire AI stack. Companies that can deliver reliable power, connect it to the grid and store it are becoming as strategically important as the chipmakers.
Why now
In many regions, the wait to connect a new data centre to the grid now stretches into years. Technology companies are signing long-term power deals directly with generators and exploring every source - renewables, gas, storage and nuclear - to secure supply.
Where the opportunity sits
The layers I’m researching
Grid equipment
Transformers, switchgear, cables and power electronics - often sold out years ahead.
Generation
Renewables, gas and nuclear that can supply firm, around-the-clock power.
Storage
Batteries that smooth renewable supply and keep data centres running.
Efficiency & cooling
Liquid cooling and power management that do more computing per watt.
Signals I watch
How I’ll know it’s working
- ↗Data-centre grid connection queues and wait times
- ↗Long-term power purchase agreements signed by technology companies
- ↗Orders for transformers, switchgear and other grid equipment
- ↗Growth in battery storage paired with renewable generation
Risks
What could go wrong
- !Utilities and energy projects are heavily regulated, which can cap returns
- !Big efficiency gains in chips could reduce expected power demand
- !Energy projects face long lead times, cost overruns and policy changes
How I’m approaching it
I hold a dedicated energy sleeve built in 2025, focused on grid equipment and diversified generation and storage rather than any single technology, and rebalance it as the power picture evolves.
“Chips set the speed of AI. Electricity sets the limit.”
These are personal, forward-looking views, not predictions or recommendations. They may turn out to be wrong.