Tech Opportunities
The Businesses Behind the AI Data Center Boom
Everyone's buying the chip stock. The quieter money is in power, cooling, and the concrete the GPUs sit on.
The AI boom looks like a software story and is actually a real-estate, energy and industrial story. A GPU is useless without a building, power, and a way to keep it from melting. That stack is where a lot of the durable money is.
Follow the constraint
The bottleneck isn't chips anymore — it's everything around them.
How it works
- GPUs
- NEED POWER
- NEED COOLING
- NEED BUILDINGS + LAND
- NEED GRID INTERCONNECT
Each arrow is a company, a supply chain, and a multi-year lead time. The constraint keeps moving down the stack to the most physical, least glamorous layer.
Where's the opportunity?
Electrical contracting, liquid-cooling installation and retrofit, and data-center commissioning are demand-swamped services right now.
Power generation, transformers, switchgear and cooling suppliers are 'picks and shovels' with long, visible order books.
Power engineering, HVAC/thermal, and data-center operations roles are in acute shortage and pay accordingly.
Boom-driven overbuild, power-availability limits, and interest rates can turn a shortage into a glut.
The unsexy layer wins
When everyone crowds the obvious trade (the chip), the edge moves to the constraint nobody wants to underwrite: multi-year power and cooling capacity.
Key takeaway
AI's physical bottleneck is power and cooling, not silicon. The businesses that supply and service the buildout capture durable demand while the model layer stays brutally competitive.