Where Can The Guest Be Powered

On 8 October, four McKinsey partners spent 43 minutes walking a global audience through the firm’s Global Energy Perspective 2026. One slide did most of the work.

The strategic question is no longer ‘How much electricity will AI use?’ It is ‘Where can AI physically be powered?’

McKinsey webinar slide, 8 October 2026

Read it again — and swap “AI” for “guest”.

Every booking engine, chatbot, revenue system and AI concierge a hotel switches on draws on a data centre somewhere. Every guest who checks in draws on a grid somewhere. Every traveller who flies in burns fuel priced in a market still absorbing the largest supply shock in modern energy history. For travel, tourism and hospitality, energy has quietly left the engineering budget and taken a seat at the strategy table.

This Viewpoint reads McKinsey’s findings through one lens: what they mean for the businesses that move, feed and house travellers, and for the AI those businesses are racing to deploy. The numbers are McKinsey’s unless stated. The hospitality reading — and the frameworks — are Pertlink’s.

The transition has become an expansion

McKinsey has quietly retired a word. The energy transition, it now says, is an energy expansion. In 2025 global energy demand grew 2.7 per cent, with oil, gas, coal and renewables all growing at once. In the firm’s baseline, primary energy demand rises another 20 per cent by 2040. Renewables grow fastest. Nothing else shrinks quickly enough to matter.

The scenarios have widened too. A year ago the gap between McKinsey’s highest and lowest demand cases was 14 per cent; now it is about 28 per cent. “We’ve been humbled,” said Helsinki partner Anna Granskog. Diego Hernandez Diaz, who leads the Perspective, drew the strategic conclusion: there is a great deal of uncertainty, so take a view with conviction, place your big bets and work in an agile manner.

For hospitality the translation is simple. Do not plan for a world in which energy gets cheaper and cleaner on schedule. Plan for one in which it gets bigger, more contested and more local — and in which your AI and your guests draw on the same constrained grid.

Eight numbers every hospitality board should know

Most of McKinsey’s numbers concern oil fields and transmission lines. These eight land on a hotel P&L.

Sources: McKinsey Global Energy Perspective 2026 webinar and report; IATA via Travelweek (June 2026).

Collision one: AI’s power bill lands on the guest folio

Pertlink has spent much of 2026 arguing that hotels should price AI the way they price breakfast: per guest. Token Cost Per Guest (TCPG) is that measure. McKinsey’s Perspective adds the variable most TCPG models leave out. Tokens are electricity with a software wrapper.

Today’s prices feel trivially cheap. The report puts public API prices at roughly $0.20 to $5.00 per million input tokens and $1.20 to $30.00 per million output tokens. But those prices sit on a data-centre build-out growing about 25 per cent a year to 2030, in markets where power — not chips — is becoming the binding constraint. When the question becomes where AI can physically be powered, the answer has a price. Eventually it reaches the folio.

Three implications follow.

1. Where your AI runs now matters. About 45 per cent of data-centre capacity sits in North America; Diego put the US at roughly 40 per cent of global demand and China at about 25 per cent. For a resort in Bohol or Bali, the model answering a guest at 2 a.m. may well be running half a world away. In McKinsey’s Fragmented Energy Order scenario, AI demand splits by geopolitical bloc and sovereign AI grows, bringing latency, data residency and cost straight back into the guest experience.

2. Cheap tokens are not a law of nature. McKinsey’s cost index puts a fragmented world 12 per cent above today’s useful-energy costs by 2050, and a decarbonised one 9 per cent below. In the high-growth case, data-centre capacity reaches 551 GW by 2035. Either way, vendors will pass energy and capacity costs through — in token prices, usage tiers or “fair use” clauses. Build your TCPG budget with an energy uplift, not a straight line.

3. AI is also the cheapest kilowatt-hour you will find. McKinsey stresses that AI works on the supply side too: grid forecasting, predictive maintenance, faster project delivery. Inside a hotel the same logic holds — occupancy-linked HVAC, chiller and boiler maintenance, kitchen equipment scheduling, laundry timed to off-peak tariffs. The test for every deployment is simple. Does it save more kilowatt-hours than its tokens consume?

Context matters here. US hotels spent an estimated 3.3 per cent of total revenue on utilities in 2023 — $9.42 per occupied room — and electricity made up 58.9 per cent of that bill (CBRE). It looks small as a share of revenue. It is not small as a share of profit, and it is the line most exposed to everything in McKinsey’s Perspective.

Pertlink’s proposal: run two meters

Track TCPG alongside kilowatt-hours per occupied room night, and report them to the board together. One tells you what your intelligence costs. The other tells you what your hospitality costs. When both draw on the same grid, you cannot manage one without the other.

Collision two: the grid is the new pre-opening critical path

If one line from the webinar belongs on every development committee agenda, it is Diego’s.

If the grid is not built, the supply and demand will not get built.

Diego Hernandez Diaz, McKinsey

McKinsey calls the grid the transition’s “ultimate speed limit”. Investment must more than double, from $443 billion a year in 2025 to $970 billion by 2050, with transmission spending growing nearly four times as fast as distribution. Namit Sharma noted that most grid companies have raised capital spending about tenfold this decade. Equipment makers carry backlogs, renewables wait in connection queues and some data centres now wait five, seven or even ten years to plug in.

Hotels, resorts, airports and convention centres join the same queue. So do the all-electric kitchens, heat pumps and EV chargers that brand standards and guests increasingly expect. McKinsey’s baseline has about a quarter of the world’s passenger-car fleet electric by 2040. Your guests will arrive needing a charge.

Anyone who has opened a hotel knows the power conversation usually happens late: an MEP line item, settled once the design is frozen. That sequence no longer works. Power capacity is now a feasibility question.

• Ask the utility before you brief the architect. Secure connection capacity, and its timeline, at land acquisition.

• Order long-lead electrical equipment early. The report notes that transformers and cables can take two to four years to procure.

• Design for flexibility. On-site storage, load-shifting and demand response shrink the connection you need. In the US, virtual power plants already supply an estimated 30–60 GW of capacity (US Department of Energy, cited by McKinsey).

• Treat data centres as neighbours, not just suppliers. In some markets they are bidding for the same land, water and substation capacity as your next resort.

Collision three: energy cost becomes destination competitiveness

Namit opened the webinar by scoring the world against McKinsey’s “energy quadrilemma”: affordability, competitiveness, sustainability and security. A year on, it is worse on all four. For tourism, watch the competitiveness corner. EU gas costs about 6.5 times the US price, and EU electricity 120 per cent more. That is a structural handicap for every energy-intensive European product — spas, ski resorts, heated pools, convention centres.

The report’s system-cost view sharpens the point. In fast-growing markets such as China, India, Indonesia and Mexico, GDP is projected to outgrow electricity system costs, so affordability improves. In France, Germany, Italy, Japan, South Korea and the UK, system costs outgrow GDP. The political fight there is over who pays — and hotels, as large commercial customers, are rarely first in line for relief.

Then there is the journey. IATA expects the airline fuel bill to rise from $252 billion in 2025 to $350 billion in 2026, with industry profit roughly halving. McKinsey’s oil scenarios run from about 80 million barrels a day in 2050 to more than 130, and its baseline expects only about 12 per cent of aviation and maritime fuel to come from non-fossil sources. Pertlink’s reading: getting the guest to the destination is now as volatile a cost as keeping them comfortable once they arrive. Destinations that rely on long-haul arrivals should model fuel shocks as seriously as currency shocks.

The sun-belt advantage

Here is the good news for Asia-Pacific, and for every tropical destination. McKinsey’s cost curves show that solar, firmed with battery storage to 90 per cent reliability, already beats gas at low latitudes: about $94 per MWh at 10 degrees and $102 at 20 degrees, against a gas benchmark of about $125. By 50 degrees it costs $283. By 60 degrees, $576.

For island and coastal resorts across Southeast Asia, the Pacific and the Caribbean, that turns energy from a cost centre into an asset. Two more signals in the report point the same way. In Pakistan, rooftop solar grew from about 0.6 GW in 2022 to 6.3 GW by the end of fiscal 2025, with no climate policy behind it, because grid tariffs nearly tripled while panel prices fell about 70 per cent. And battery storage, said Anna Granskog, has become an asset class in its own right, built increasingly by infrastructure investors who earn their return trading the power. A hotel does not have to buy the battery. It can host one.

A caution: latitude is a proxy. Monsoon cloud cover, typhoon resilience and land constraints all move the number. That is exactly where AI earns its keep — forecasting load and solar yield, and sizing the storage.

Five futures, five hospitality playbooks

McKinsey assigns no probabilities to its scenarios. Neither should you. But each implies a different energy and AI environment for travel, and a different set of no-regret moves.

Cost index = McKinsey’s useful-energy cost index for 2050 (Continued Momentum 2025 = 100). Data-centre figures are GW of IT capacity. Implications are Pertlink’s.

What to do on Monday

Eight moves, in rough order of urgency.

1. Put kilowatts in every AI business case. Ask each AI vendor where inference runs, on what power and in which jurisdiction, and how energy and capacity costs flow into its price. Stress-test your TCPG against an energy uplift.

2. Run two meters. Report TCPG and kWh per occupied room night side by side, monthly, by property.

3. Use AI to cut energy before you use it to add features. HVAC, chillers, kitchens and laundry are where AI can pay its own power bill.

4. Make power a feasibility item. Grid-connection studies at land acquisition; long-lead electrical equipment ordered early; storage and load-shifting designed in.

5. Firm your own power where the sun allows. Solar plus storage for sun-belt resorts, with third-party-owned batteries where your capital is better spent elsewhere.

6. Watch the air bridge. Track fuel pass-through and route changes in your top source markets, and build short-haul and domestic demand where you can.

7. Plan against the boundaries, not the middle. McKinsey built Hydrocarbon Resilience and Fragmented Energy Order as stress tests. Use them that way — then pick the future you believe in, and commit.

8. Destinations: map power the way you map rooms. Tourism boards already plan room supply, airlift and attractions. Grid capacity, and the data centres competing for it, now belongs on the same map.

Signposts to watch

• How long the Strait of Hormuz disruption lasts, and what it does to jet fuel and food logistics

• Data-centre connection queues and power deals in your key markets

• The EU–US electricity price gap

• Battery and solar module prices, and any trade measures that move them

• AI chip efficiency gains, and how quickly they reach token prices

• Grid companies’ capital plans, and the tariff cases that follow

• McKinsey’s Global Materials Perspective 2026, due within weeks, on critical minerals

The last word

McKinsey closed with advice that fits a general manager as well as an oil major: take a view with conviction, place your big bets, stay agile. For travel, tourism and hospitality, the view is this. AI is not weightless. Neither is a guest night. Both run on power that is becoming more contested, more local and, in much of the world, more expensive.

The properties that win will know what their intelligence costs, where their power comes from, and how quickly they can get more of it.

The intelligence may be artificial. But the experience is human. From now on, both run on the same grid.

Sources and notes

1. McKinsey & Company, “Global Energy Perspective 2026” live webinar, 8 October 2026: Namit Sharma (Senior Partner, Amsterdam), Anna Granskog (Partner, Helsinki), Diego Hernandez Diaz (Partner, Geneva), Luciano Di Fiori (Partner, Houston). Quotations from the session transcript and slides.

2. McKinsey & Company, Global Energy Perspective 2026, September 2026. mckinsey.com

3. IATA industry outlook, reported in Travelweek, “Fuel bill for world’s airlines will soar from $252 billion to $350 billion, fare hikes inevitable”, 8 June 2026. travelweek.ca

4. Robert Mandelbaum (CBRE Hotels Research) and Joe Snider (IHG Hotels & Resorts), “Gaining Control of Utility Costs”, CBRE, April 2024 (US hotels, 2023 estimates). cbre.com

Token Cost Per Guest (TCPG), the two-meter proposal and all hospitality interpretations are Pertlink’s own and do not represent McKinsey’s views. Destination latitudes in the chart are approximate. Related Pertlink Viewpoints: the TCPG series; “Spend Wisely”; “The Front Desk Has No Passport Control. Yet.”; “Build the Truth Once”.

Made with the help of various SI [AI] tools, but always with a HITL