The industry has spent years arguing about which pipe wins. The bigger change is what is being rebuilt around and beneath the pipes.
The travel distribution industry has spent a decade asking whether NDC will replace the GDS. It may have been asking the wrong question.
The question assumes a contest between two pipes, and the 2026 evidence describes something else. Travelport has signed EGYPTAIR to a single agreement covering both EDIFACT and NDC. Amadeus is bringing JAL’s NDC content to agencies in 26 countries and regions from this month. Sabre says NDC is about 5% of its distribution volume and still growing. Companies supposedly on opposite sides are carrying each other’s channel, and GO7’s Peer Winter says “For too long, the industry has been arguing GDS vs NDC.”
Underneath that argument, a different fight is under way. It is about who controls the offer, the customer, the transaction, the data and the servicing relationship, and who gets paid for sitting between supply and demand. That fight, not the fate of any one pipe, is what makes this the most important transition since the GDS arrived.
NDC Was Never the Destination
NDC changed the conversation by giving airlines a standard way to send richer content to sellers: bundles, ancillaries, prices that can vary by shopper, connectivity that does not depend on the old fare-filing model. IATA itself describes the industry’s core artifacts, from the PNR to the e-ticket, as more than half a century old. NDC was the first serious attempt to replace what sits between airline and seller.
But connectivity only solves a connection problem. An airline that can send a richer message still has to decide what goes into it: which bundle, at what price, for which customer, and what happens when that customer changes the trip. Accelya’s chief product and technology officer put it plainly in May: “Modern airline retailing does not stop when an offer is created.”
That is the shift from distribution to retailing. Distribution asks how a fare reaches a seller. Retailing asks who decides what is sold, to whom, at what price and in what form, and who looks after the traveller afterwards. It also raises the stakes of the channel. This is interpretation rather than documentation, but once an airline controls the offer, the channel is where that control either survives or gets diluted. Turkish Airlines’ new Travelport agreement includes access to NDC-only fares and a reduced surcharge for NDC. Pricing the channel has become part of retailing strategy.
Offer and Order Changes the Architecture
The traditional airline sale moves through a chain of separate artifacts. A schedule is published, a fare is filed, a booking creates a PNR, a ticket is issued as its own document, and any later change touches both, usually with reconciliation afterwards. The chain carries most of the world’s air travel today. It was designed to sell seats, not to retail products.
Offer and Order replaces the chain with a different logic. The airline constructs an offer, the customer accepts it, and the accepted offer becomes an order that holds what was bought, what was paid and what has been delivered. Servicing means changing the order rather than repairing a booking and a ticket. The commercial consequence is that a flight, a seat, a bag and eventually a non-air product can be sold and changed as one thing. Arguably, whoever holds the order holds the relationship, because every party touching the transaction, from intermediary to payment provider to settlement system, must work from that record instead of copies of it.
The transition is real but early. IATA’s 2025 surveys of 20 IT providers and 14 first-mover airlines show leading airlines live with Orders around 2026 to 2027, and IATA cites Riyadh Air as a native Offers and Orders airline. An Accelya and Atmosphere Report Group study found that fewer than a third of airlines have started Offer and Order development. IATA’s Airline Retailing Maturity index now replaces its earlier NDC and ONE Order certification registries, covering everything from Shop and Order to Pay, Account and Settle. NDC has been folded into a larger frame.
The GDS Is Changing Jobs, Not Disappearing
Any fair account starts with what the GDS still does. It connects airlines to very large numbers of agencies, supplies the workflow those agencies already use, handles servicing, and sits alongside settlement systems that matter: IATA says its EASY PAY is available in over 120 BSPs and supported across both GDS and NDC channels. It also aggregates multiple sources, which is what Travelport pitches in its Turkish Airlines announcement: normalised, curated content, intelligent orchestration and search, and consistent servicing.
The GDSs are adapting in ways the old binary hides. Amadeus now sits on both sides of the transaction, selling Altea NDC to airlines such as British Airways and Royal Air Maroc while distributing NDC content to agencies. In the first half of 2026 its Air IT Solutions revenue grew 8.7%, against 1.1% in Air Distribution. Sabre has announced a shopping product that unifies EDIFACT, NDC, LCC and private content.
So the interesting question is not whether the GDS survives. It is what a GDS becomes when distribution is heterogeneous. One reading is that it moves from being the single source toward being the arbiter of many: deciding which source to query, how to compare the results and how to service the booking afterwards. Whether that role holds depends on how evenly airlines price their own channels, which returns to the surcharge question above.
The Rise of Orchestration
Follow the logic and a problem appears. If airlines expose offers through different channels, and each implements standards its own way, a travel seller faces dozens of new connections on top of the old one. A seller might need GDS content, NDC from several airlines, low-cost carrier content, direct airline APIs, hotels, rail, ancillaries and a payment route for all of it. Someone has to make that commercially usable.
The market’s answer is already visible. TPConnects’ Iris platform lets agencies access NDC content alongside EDIFACT, LCC and aggregator content in a single interface, and TPConnects reports that in a poll of attendees at a joint webinar with Finnair, managing multiple airline connections was the number one challenge. That is a vendor poll, not an industry survey, but the pattern holds across the market. GO7 describes a central hub connecting full-service and low-cost carriers. Different companies, one function.
The strongest evidence comes from the most advanced airline. Riyadh Air, built around Offer and Order, named Verteil as its launch NDC aggregator and also brought its NDC content to TPConnects’ Iris. Even an airline designed from scratch for modern retailing relies on orchestrators to reach sellers.
Orchestration reaches past content, too. The IATA Financial Gateway is pitched as an orchestration layer for airline payments across all sales channels, and a UATP white paper argues that settlement infrastructure lags the retailing build-out. My reading is that value migrates toward whoever can normalise, compare, service and settle across sources, because that is what keeps a seller productive as supply fragments. Whoever does that would sit in the data path of a growing share of transactions, and that is where leverage is likely to accumulate. It is a hypothesis, not yet a documented shift in revenue.
Who Controls the Commercial Relationship?
This is where the threads meet. Each participant wants the same thing under a different name. Airlines want control of the offer and the customer, and Emirates’ full integration of its NDC API with Expedia Group in March 2025 shows the direct route to the largest OTAs. Sellers want broad content without rebuilding their stack for every carrier. Technology providers want to be the infrastructure between supply and demand. GDSs want to remain central to the transaction. OTAs want scale and differentiated content.
Each position is legitimate, and none can deliver the others’ needs alone. Sabre argues that direct-connect models have often proved harder to scale, with higher servicing costs and fragmented customer experiences. That is a GDS’s view, but it points at a real constraint. An airline can own the offer and still depend on infrastructure it does not own for servicing, payment and comparison. My analysis is that whoever handles the change request at 2am, whoever sees the payment data and whoever decides which source answers first may each hold a piece of the relationship. On that reading, control is becoming layered, not held, and the real negotiation is over who owns which layer and what they charge for it.
When the Interface Becomes an Agent
Now add a new front end. In October 2025, Expedia, Booking.com and Tripadvisor launched AI-powered integrations inside ChatGPT. Mindtrip has launched a flight booking experience powered by Sabre’s agentic-ready APIs and PayPal’s agentic commerce services, and Sabre says its active agentic partners doubled from 30 to 60 in the second quarter.
The twist is what this does to the control question. The industry is fighting over control of the distribution interface just as the interface itself may be changing. Every party above has assumed the customer relationship would be held by an airline, an OTA or a travel agency. An agent that assembles the itinerary could hold it instead, because it sees the preferences, runs the comparison, chooses the offer and takes the payment. None of that is established yet. It is the scenario the current positioning has to survive.
One camp says the agent bypasses everything beneath it: airlines can expose MCP interfaces directly to agents, skipping aggregators. Perhaps for search. But the same analysis cautions that search is not booking, and PYMNTS notes that most AI travel assistants still run on bespoke integrations and stitched-together APIs.
My argument is that the agent hides complexity from the traveller without removing it. To compare flights across airlines fairly, an agent needs offers in comparable form, availability that is real, a price that holds, a payment that settles, and a way to change the booking later. Those are the chores orchestration layers already perform. A thinner interface may make the infrastructure beneath it more important, not less. It also raises an uncomfortable neutrality question: if the agent ranks the offers, whose interests does it serve? PYMNTS warns that the locus of power may shift again.
The Questions That Remain
None of this can be predicted with confidence. The structural questions can at least be named. Who owns the offer, the airline that builds it or the system that assembles and ranks it? Who owns the order, and does every party work from the same record? Who owns the customer when the traveller meets an agent first? Who controls merchandising once offers are compared across sources? Who handles servicing when a change touches three suppliers? Who orchestrates the sources, and on what terms? Who owns the payment relationship? And who becomes the infrastructure layer between suppliers and intelligent travel interfaces?
From the GDS to NDC, from Offer and Order to orchestration and now the agent, each stage has asked the same thing in a different form: who controls the transaction? The debate we started with asks which pipe will carry the sale. These questions ask who defines the offer, who holds the order, who stands between a thousand sources and one interface, and who gets paid for doing it. The industry has been arguing about the plumbing while the building changes shape. When it settles, will anyone still ask which pipe won, or only who ran the system that chose it?
