By 2040, Travel Won't Be an Industry — It Will Be Infrastructure
My thesis is simple and, I suspect, unfashionable: by 2040 travel will stop behaving like a discretionary consumer category and start behaving like infrastructure — booked by agents, priced like utilities, and measured by throughput rather than wanderlust. The romantic vocabulary we still use — "trip of a lifetime," "bucket list," "getaway" — will sound as quaint in 2040 as "long-distance call" sounds today. The reason isn't that humans will stop wanting to move. It's that the layer between human desire and physical movement is being rebuilt, and once it's rebuilt, travel becomes something we provision, not something we plan.
Why this matters now
Travel is already too big to be treated as a lifestyle vertical. The World Travel & Tourism Council projected travel and tourism would contribute roughly $11 trillion to global GDP in 2024. That's not a sector — that's a tectonic plate. The UN World Tourism Organization has long forecast that international tourist arrivals will reach 1.8 billion by 2030. If that arrival curve continues bending upward into the late 2030s, by 2040 the world will be processing human movement at a scale that simply cannot be coordinated by the current patchwork of search engines, OTAs, loyalty programs, and gate agents reading off screens.
The pandemic should have taught us how fragile that patchwork really is. Travel volumes fell about 75% in 2020 according to McKinsey's analysis. A 75% drawdown in any other trillion-dollar sector would have triggered a permanent regulatory rewrite. In travel, we papered over the cracks, restored the old workflows, and went back to telling ourselves that booking a flight on a clunky website was a normal way to coordinate a hundred-billion-dollar logistical chain. It isn't. And the pressure to fix it — from climate accounting, from labor shortages, from agentic AI, from sheer volume — is what will force travel to mature out of "industry" and into "infrastructure" by 2040.
The agent layer is forming faster than the incumbents admit
Here is the data point I cannot stop thinking about. An Expedia survey reported that around 8% of travel discovery and booking is already happening through AI. Eight percent sounds modest until you remember that this is a behavior that essentially did not exist three years ago. Adoption curves of that shape — zero to high single digits inside a single product cycle — are exactly what e-commerce looked like in 1999 and what mobile booking looked like in 2010. Both became the dominant channel within roughly a decade. There is no compelling reason to expect agentic travel discovery to behave differently, and several reasons to expect it to move faster, because the underlying model improvements are exogenous to travel itself.
Project that curve to 2040 and the implications cascade. If a majority of inspiration, comparison, and booking is mediated by an AI agent acting on a traveler's behalf, then the entire surface area of "marketing to travelers" collapses into "being legible to agents." Hotels won't optimize for human eyeballs scanning a results page; they'll optimize for structured availability feeds, verifiable amenity claims, and machine-readable cancellation policies. The supplier that lies to the agent — about a renovation, about a beachfront, about Wi-Fi speed — gets downgraded the way spam sites got downgraded in search. Trust becomes a wholesale relationship between a property and an agent fleet, not a retail relationship between a property and a guest.
This is what I mean by infrastructure. Roads do not advertise. Power grids do not run loyalty promotions. As travel's coordination layer becomes machine-mediated, the supplier-side experience starts to resemble the way utilities interact with grid operators: through standards, audits, and SLAs.
The supply graph is already too large for humans to navigate
Look at the supply side as it exists today. Booking.com alone lists roughly 3.4 million properties on its platform, a figure composed of about 475,000 hotels, motels, and resorts, alongside roughly 2.9 million non-hotel properties. That ratio — non-hotel inventory outnumbering traditional hotel inventory by more than six to one on a single major platform — tells you how absurd it has become to ask a human to "compare options."
No one compares 3.4 million options. They glance at the first ten, sort by price, accept whatever the algorithm surfaced, and convince themselves they "researched" it. That charade has been the open secret of online travel for fifteen years. By 2040 it will be over. An agent can actually evaluate the long tail. It can rank a 12-room guesthouse in a town the traveler has never heard of against a chain hotel three blocks away on the dimensions that specific traveler cares about. The economic consequence is that the long tail of supply — independent operators, regional brands, alternative accommodations — finally becomes addressable. The marketing dollar advantage that global chains have spent decades accumulating becomes much less defensible when the buyer is a machine that does not know what a billboard is.
That is a redistribution event. It will not be small.
Pricing will look like electricity, not like airline seats
The second piece of the infrastructure thesis is pricing. Airlines and hotels already practice dynamic pricing, but the dynamism is crude — based on cabin, day-of-week, lead time, and a few demand signals. By 2040 the agent layer will demand and reward pricing that is continuous, transparent, and bundled across legs of a journey.
Imagine your agent assembling a six-day trip across three cities. It does not want twelve separate transactions with twelve separate cancellation policies and twelve separate loyalty currencies. It wants a single quote for the journey, with substitutability priced in: if a flight is delayed, the hotel night and the restaurant deposit reprice automatically. That is what infrastructure pricing looks like — closer to how a data center buys compute or how a factory buys electricity than how a person buys a vacation. Suppliers who refuse to participate in that fabric will be routed around, the way agents already route around websites that block them.
Loyalty programs, in this world, are in serious trouble. Their entire economic logic depends on the friction of a human remembering a brand, accumulating points, and being too inertia-bound to switch. An agent has no inertia and no sentimentality. It will defect for a 4% better total-trip price without blinking. The loyalty programs that survive will do so by becoming actually useful — guaranteed upgrades, guaranteed availability, fee waivers that the agent can verify — rather than by being elaborate psychological traps.
Climate accounting forces the infrastructure framing
There is one more force pushing travel toward infrastructure status, and it is the one the industry talks about least honestly: carbon. With travel and tourism contributing approximately $11 trillion to global GDP in 2024, any serious global decarbonization pathway must price aviation, cruise, and accommodation emissions in a way that today's per-trip mental model cannot absorb.
By 2040 I expect carbon accounting to be embedded directly into the booking surface — not as a virtuous afterthought offered by an airline's checkout page, but as a binding constraint on what the agent is allowed to book on your behalf. Corporations already do versions of this for business travel. Extending it to leisure travel, with national or supranational allowances, is not a wild prediction; it is a logical extrapolation of the regulatory direction Europe has already set. When a 1.8-billion-arrivals world, as the UNWTO has projected for 2030, keeps growing into the late 2030s, the alternative is unmanageable.
Treating travel as infrastructure is what makes carbon accounting tractable. Roads have weight limits; grids have load limits; by 2040, regions and routes will have carbon limits, and the agent layer will be the mechanism that enforces them invisibly, the way a thermostat enforces a setpoint without lecturing you about it.
The human experience gets better, not worse
I want to be careful here, because the infrastructure framing sounds cold, and travel is the opposite of cold. So let me say plainly: I think this transition is good for travelers, and the people who will resist it loudest are the intermediaries whose margins depend on confusion.
What does a 2040 trip feel like from the traveler's seat? You say, in natural language, what you want. The agent assembles it across the long tail of supply, prices it as a single coherent journey, books it against your carbon and budget allowances, and renegotiates in real time when reality intervenes. The friction that currently consumes the first ten hours of any complicated trip — the comparison shopping, the tab-juggling, the loyalty-program optimization, the rebooking when a flight is canceled — disappears. What remains is the part that actually matters: being somewhere new with the people you love.
That is not the death of travel romance. That is the restoration of it. The romance was never in the booking flow. The booking flow has been a tax on the romance for thirty years.
Counterargument: the high-touch, irreducibly human view
The serious objection to the infrastructure thesis — and it is serious, not a strawman — runs like this. Travel is not a logistics problem to be optimized; it is a relational, sensory, deeply human practice whose value is precisely the things an agent cannot mediate. The concierge who knows your aunt has a bad knee and rebooks you to a ground-floor room. The boutique hotelier who upgrades you because you mentioned your anniversary. The taxi driver who detours past the bakery his cousin owns. These are the moments people remember, and they are constituted by human discretion, by the supplier seeing the guest as a person rather than as a row in a structured feed. Critics in this camp argue that pushing travel toward agent-mediated, machine-readable infrastructure will optimize away exactly the texture that makes travel worth doing — the way airline deregulation optimized away legroom and meals. Worse, it concentrates power in whichever two or three foundation-model providers end up running the dominant agent fleets, recreating the OTA oligopoly at a higher and less accountable layer of the stack. The human travel agent, on this view, was never the problem; the platform middleman was, and replacing one middleman with a more opaque one is not progress.
I take that critique seriously, and I think it gets the diagnosis right and the prognosis wrong. Yes, agent mediation can flatten texture — but the current system already has. When a single platform lists 3.4 million properties, with 2.9 million of them in the non-hotel category, the median traveler is not having a high-touch experience; they are picking from the first page of an algorithmic ranking they do not understand. The texture critics want to preserve is already inaccessible to most people most of the time. A well-designed agent layer can actually surface the boutique hotelier and the 12-room guesthouse to travelers who would never have found them, because it can read the long tail in a way humans cannot. The concentration risk is real and worth regulating, but the answer is interoperability standards and agent portability — the same playbook that worked for email and is being attempted for messaging — not a defense of a status quo that already failed the texture test. The romance survives infrastructure; in fact, infrastructure is what lets the romance scale to 1.8 billion arrivals.
What to do between now and 2040
If you run a hotel, an airline, a destination marketing organization, or a tour operator, the work for the next fifteen years is not to build a better website. It is to make your supply legible, verifiable, and reroutable to an agent layer you do not yet control. That means structured data, honest amenity claims, real-time availability, programmatic pricing, machine-readable carbon disclosures, and APIs that do not punish small partners. The suppliers who do this early will be over-represented in agent recommendations the way SEO-native publishers were over-represented in early Google. The ones who wait will discover, around 2032, that they have been quietly disintermediated by competitors they never noticed.
If you are a traveler, the work is smaller and more interesting: start practicing now. Use the agent tools that exist. Notice where they fail. Notice where they already outperform the human-driven workflow you have been tolerating. The 8% of travel discovery already happening through AI, per Expedia's survey, is the leading edge of a behavior that will be near-universal in fifteen years. The travelers who learn to direct an agent well will get more interesting trips at lower cost. The ones who do not will pay the inertia tax.
And if you are a policymaker, the work is the most consequential of all. Decide now what interoperability, carbon accountability, and agent transparency should look like, before the agent oligopoly is calcified. Travel is going to become infrastructure whether we plan for it or not. The only question is whether the infrastructure we end up with is open, auditable, and accountable — or whether it is three black boxes in a trench coat that everyone is forced to use because there is no longer any other way to move.
I'd rather we chose. By 2040, we will not have the luxury of pretending we didn't.
