AI needs power: where can the guest be served?

Pertlink has done something most hotel tech vendors avoid: it has taken McKinsey's Global Energy Perspective 2026 and dropped it on top of hospitality operations. The argument, published on Hospitality Net, is simple and uncomfortable. Energy constraints will decide where AI can run and where guests can actually be served. Not a futurist thought experiment. A capacity planning problem.

The proposal that caught my eye: track AI token costs next to utility costs per occupied room. That is a metric hoteliers can put into the monthly P&L review tomorrow. Most hotels already watch energy per occupied room closely. Adding the inference cost of every chatbot, upsell engine and revenue model to the same line turns AI from a vague innovation budget into a line item with a number attached.

My take: this is the right framing. AI in hotels is not free and it is not magic. It is compute, and compute drinks power. The chains and PMS vendors that start measuring tokens per occupied room now will know exactly which AI features earn their keep. The rest will keep guessing. Pertlink has handed the sector a useful yardstick before the bill arrives.