The airline CEO putting Starlink on every plane

Summary

• United is using Starlink to turn a temporary connectivity advantage into lasting customer loyalty. Kirby accepts a window of two to three years before competitors catch up with low Earth orbit connectivity, potentially including Amazon. The plan is to make passengers try United for Wi-Fi, then stay for its broader technology and service: “Free was never the value proposition. Fast is the value proposition.”
• Starlink’s payoff depends on predictable coverage and rapid certification. At recording, 48% of flights had Starlink, approaching 50%, with roughly three aircraft being equipped daily; smaller aircraft came first because they were certified first. Los Angeles is targeted to become the first fully covered hub in the first half of next year, while Boeing’s 787 certification work slows other markets. Kirby says the financial “numbers weren’t that big.”
• United’s pandemic fleet order rested on shrinking competition as much as recovering demand. Kirby ordered 200 wide-bodies plus several hundred narrow-bodies while other global long-haul airlines permanently retired aircraft. Buying assets with 30-year lives during the downturn secured favorable pricing and positioned United for recovery: “We don’t have to be right. We just have to be not totally wrong.”
• United’s technology advantage began with infrastructure spending that lacked a conventional payback calculation. A project costing several hundred million dollars, started in 2020, moved systems off mainframes and modernized databases. Kirby’s subsequent AI enthusiasm centers on programming and faster product delivery, rather than document summaries; his constraint on technology investment is now “how much can we manage? It is not money.”
• Better disruption handling combines clearer communication with attempts to prevent the disruption itself. Kirby thinks a weather-radar messaging experiment moved NPS from negative 60 to positive 60 for passengers delayed an hour. He also says AI-supported routing could probably reduce Denver’s summer weather delays by 50%–75%, subject to FAA cooperation: “You can’t use weather as an excuse.”
• Air traffic control is the structural ceiling on airline growth in Kirby’s account. He calls it “90% of the issue”: modernization takes five to 10 years, but annual appropriations repeatedly interrupt execution. Despite praising current progress, he expects flight limits to spread as airports reach the capacity the existing system can manage.
• Kirby sees airline profitability as a route-network discipline problem before a product problem. He expects six of 10 US airlines to lose money this year and estimates American’s Chicago loss at roughly $1.5 billion. Low-cost carriers face their own mismatch: he contrasts New York airport charges of $52 per passenger with Spirit’s former $58 average fare. The rule is to “stick to your knitting,” including accepting limits to growth.
• Credit-card revenue overstates the loyalty business’s contribution if redemption costs disappear from the analysis. Kirby says more than 50,000 passengers fly United daily using credit-card miles, creating a substantial carrying cost against the disclosed revenue. He calls the popular accounting interpretation “blown way out of proportion”; his preferred loyalty investments include club space, already more than doubled since before COVID and planned to grow another 50%.

Deep dive

1. United bought long-lived capacity when competitors were shrinking

• Kirby’s pandemic premise was that COVID would be deep and prolonged, followed by recovering travel demand. While every other global long-haul airline permanently retired some wide-body capacity, he says, United ordered 200 wide-bodies and several hundred narrow-bodies. The wide-bodies list around $400 million each, before substantial discounts. Competitor retrenchment meant “we don’t have to be right. We just have to be not totally wrong.”

• Thompson compares the decision with Samsung investing through memory downturns. Kirby agrees with the countercyclical logic, while noting that aircraft have 30-year lives and must survive several recessions. Being effectively the only buyer secured particularly good pricing; United’s commitment also gave Boeing evidence it could take to other customers when restarting the market.

• The forecasting error, in Kirby’s account, combined extrapolating present conditions with believing surveys about future behavior. Frightened respondents genuinely intended never to travel internationally again; that did not make their intentions durable. “Experts live in the world of two standard deviations,” he argues. He likes surveys about completed experiences but ignores them for deciding what to build next.

• Collison asks what permits these large bets. Kirby credits extensive reading, a supportive board and constant communication, while dismissing Wall Street’s preference for near-term cost control. Airport fees and union contracts leave future investment as an easy budget cut. His bargain is explicit: invest for the long term, but “if it doesn’t work, you ought to lose your job.”

2. Modern infrastructure made AI useful as a programming tool

• During COVID, United began a project costing several hundred million dollars to leave mainframes, modernize databases and make information accessible. “There was no ROI. There was no payback,” Kirby says. The justification was future execution speed: United can now deliver changes that competitors cannot easily implement because their underlying systems do not support them.

• Kirby’s AI epiphany came after three one-hour tutoring sessions. During a Saturday workout, with about 2.5 hours of elapsed runtime, he built a website comparing odds across four betting venues, including Kalshi. Dropdowns selected sports, dates and bet types; entering an amount sent the bet to the venue offering the best odds. His reaction to the team: “We’re all in on this.”

• Owning critical software matters independently of AI. Kirby says United owns its reservation system, also used by Copa, after an inherited arrangement eventually forced United to support it itself. Ownership is now an advantage as more systems come in-house. For one major mainframe replacement, United withheld several hundred flights from the schedule as a precaution; the transition ultimately worked without a hitch.

• The same logic reaches the seat-back screen. United is taking software control from Panasonic so maps can remember passenger preferences instead of cycling through unwanted views. Kirby also wants a live celebration when a passenger crosses two million miles, with flight attendants knowing the precise moment. “We could have never done that unless we took over the software.”

3. Customer communication must come with better operational choices

• Thompson’s pushback is personal: after internalizing ConnectionSaver as a United advantage, he expected it to protect a tight Denver connection and was particularly disappointed when the departing flight had already closed. Kirby accepts the disappointment. His answer combines better communication with continued invention: a customer’s “wow” becomes the expected baseline, and “then we owe them another ‘wow.’”

• A weather experiment illustrates the communication opportunity. On a sunny Chicago day, storms beyond the airport blocked departure routes. Half the affected passengers received the usual delay message; half saw the radar map explaining the obstruction. Kirby thinks the figures for an hour’s delay were negative 60 NPS versus positive 60—a 120-point difference, with the explanatory message outperforming an on-time flight.

• United already offers alternative flights alongside weather warnings. Kirby wants a more detailed interface ready by Christmas: an illustrative screen would show a 41% weather-impact probability for the booked flight, versus 12% earlier and 3% later. The distinction is actionable information instead of a notification that merely raises anxiety. Integrating seat-back screens with the app should make such communication more consistent.

• Collison and Thompson call this expectation management; Kirby resists because the phrase can excuse poor outcomes. For Denver’s pop-up storms, United is simulating alternatives to the FAA’s worst-case restrictions across arrival and departure routes. Using unaffected routes could “probably reduce 50%–75%” of summer weather-impacted delays, he says, but FAA help is necessary. The cultural instruction is to overcome weather’s effects, rather than invoke weather as the explanation.

4. Air traffic control determines how much the industry can grow

• Asked whether aircraft or airports constrain United, Kirby answers: “No, it’s air traffic control. Air traffic control is 90% of the issue.” Once the FAA reaches what its current system can manage, flight caps follow—as in New York and Chicago—even when physical airspace could accommodate more traffic under a different system.

• The funding mechanism frustrates him: projects requiring five to 10 years face annual appropriations. Delayed budgets, ramp-up and shutdown can leave only four productive months in a year. Kirby believes private enterprise could execute modernization in two to three years, while explicitly declining to advocate FAA privatization because it is politically toxic. He nevertheless praises current FAA and transportation leadership and progress.

• Kirby’s political diagnosis includes private-jet lobbying, which he says protects essentially free access while airlines fund the system. He also questions maintaining roughly 450 facilities when controllers largely work from screens and could be centralized with backups. Congressional concern about local jobs complicates consolidation. His conclusion is constrained optimism: more progress than during his career, but government remains poorly equipped to move at a technology company’s pace.

5. Chicago shows why duplicate hubs consume scarce capacity

• Thompson asks why a larger United presence would benefit Chicago. Kirby’s answer is network economics: United operates roughly eight daily Chicago–Milwaukee flights, largely feeding connections, while American duplicates that catchment service. Instead of roughly 16 flights, eight larger aircraft could carry the passengers and release scarce flight capacity for Paris or Tokyo. He attributes Atlanta’s and Dallas’s stronger hub metrics partly to avoiding this duplication.

• Thompson challenges whether Kirby’s rivalry with his former employer makes the argument personal. Kirby denies it and points to even larger share shifts against other competitors elsewhere. In Chicago, he says United moved from lower market share than American to roughly double its revenue share. He estimates American will lose about $1.5 billion there this year.

• Collison presses the competitive tension in grandfathered airport slots: incumbents deserve returns on infrastructure, but entrants also need opportunity. Kirby prioritizes preventing schedules that exceed airport capacity, noting he sought effective slot control at Newark. His preferred expansion mechanism is better air traffic control creating additional slots. Chicago’s separate gate-allocation process also mattered: United returned to its pre-COVID schedule, while American, focused on the Sun Belt, did not restore its schedule and lost six gates.

6. Starlink’s proposition is speed, with loyalty as the destination

• Kirby ranks schedule first in airline choice and now puts Starlink immediately behind it. Free connectivity alone does not solve insufficient bandwidth; he says recent flights on three competitors’ legacy systems were terrible. Thompson agrees that making constrained Wi-Fi free can worsen the experience. Kirby’s categorical framing: “Fast is the value proposition,” and Starlink is currently “the only option.”

• Collison asks why rivals’ decisions against Starlink might have seemed rational. Kirby describes the industry’s habitual pace: an aircraft modification can be treated as an eight-year project, making a one-year delay appear immaterial. He acknowledges United also underestimated the opportunity. Early passenger reactions changed the scale and urgency of its commitment.

• Kirby calls the advantage temporary: competitors will eventually have low Earth orbit service, whether Starlink or potentially Amazon. He accepts Collison’s window of two to three years to win customers. “They’re going to try us for Starlink,” then stay for other technology and products. He also acknowledges Asian and Middle Eastern carriers’ strengths in service and physical product; United must keep improving beyond connectivity.

7. Coverage certainty changes the value of the Starlink rollout

• Starlink costs more, but Kirby says the financial “numbers weren’t that big.” United initially planned installations during heavy maintenance checks and coverage for only 65% of the fleet, excluding retiring aircraft. Then passengers began choosing regional flights specifically for Starlink. That behavioral evidence prompted acceleration, supported by SpaceX, rather than the customary gradual retrofit.

• At recording, Starlink covered 48% of flights and was approaching 50%, with about three aircraft equipped daily. Thompson notes that seat-mile coverage is lower because small aircraft came first. Kirby says certification dictated that sequence. Boeing’s remaining 787 certification work is a particular obstacle for a hub such as San Francisco, with substantial 787 flying.

• Thompson argues that guaranteed availability matters much more than a chance of connectivity. United plans flight-specific probabilities in its app; Kirby illustrates a MAX 8 that might have 25% availability now and 36% by Christmas. Los Angeles is planned as the first hub with 100% coverage, sometime in the first half of next year, reflecting both competition and its aircraft mix.

8. Product ambition extends from new airframes to family travel

• Collison notes that the 737 still flies around Mach 0.78, as in 1967, despite burning 50% less fuel and becoming 30–40 times safer. Kirby finds the pace “maddening.” He believes Boeing and Airbus have informally avoided new-airframe investment because investors resist it. After seeking a 757 replacement for more than 20 years, his answer is blunt: “We need a third aircraft manufacturer.”

• United is investing in JetZero’s blended-wing-body approach, which Kirby compares with the B-2 bomber’s efficient shape. Thompson raises airport compatibility; Kirby acknowledges larger aircraft would require gate changes. He also sees room for better engines.

• Relax Row, planned for later this year or next year, would turn three economy seats into a place to lie down. Kirby expects profitability but says the product will barely register financially. Its value is also brand-building. With seven children, he evaluates travel through a parent managing young kids alone; seat-back entertainment and Starlink help make that journey easier.

• For Kirby, managing 120,000 employees is “99%” creating the right culture. Marketing provides his example of trust producing speed: Maggie Schmerin launched a Starlink campaign without seeking his approval. He learned about it through other people while attending a CVS Health board meeting. Multiple approval layers, he argues, drain work of speed, creativity and fun.

9. Route experimentation works when unsuccessful flights actually stop

• Collison asks whether airlines have structurally become better businesses. Kirby answers “yes and no”: six of 10 US airlines are expected to lose money this year, while he sees United and Delta as the two winners through brand loyalty. He describes an $8 billion oil-price increase for United this year against $5 billion of profit in its best year, yet still expects solid profitability below last year’s.

• Kirby credits Patrick Quayle’s team for destinations he would never have selected, including Nuuk. United expects 53 transatlantic destinations next summer versus 23 a decade earlier, with only two failures. Bergen disappointed despite Kirby liking the destination, so United stopped it. “Our assets are fungible”: the ability to redeploy aircraft makes experimentation economical.

• The sharpest discovery example is pilgrimage travel between cities in Spain and Portugal. Quayle investigated the journeys, then found fragmented itineraries: passengers flew United into one city and another airline home from the other. Adding both endpoints addressed the actual trip. The opportunity began with noticing a behavior, then finding supporting data, rather than demanding historical evidence for a route nobody had operated.

10. Low-cost airlines need low-cost airports and strategic restraint

• Kirby contrasts New York airport charges of $52 per passenger with Spirit’s former $58 average fare. That leaves almost nothing for fuel, aircraft, employees or maintenance. He says major-airport expenses have risen roughly 70% since before COVID, making these airports particularly difficult for low-cost operators.

• Thompson pushes back: did United also price low-cost carriers out through basic economy and finer seat segmentation? Kirby acknowledges the competitive mechanism. Basic economy and larger aircraft let United compete profitably on price much of the time. Once the fare advantage disappears, he argues, the low-cost carrier loses its main reason for being chosen.

• Ryanair supplies his model of discipline: secondary airports rather than Heathrow or Charles de Gaulle. US low-cost airlines moved into expensive major airports, abandoning a foundational advantage. Collison compares the problem with chasing growth beyond a successful operating formula. Kirby’s formulation is broader: “Know who you are, know what your strengths are, and stick to your knitting.”

• The mistake runs both ways. Legacy airlines failed at creating internal airlines to out-Southwest Southwest; Kirby expects Southwest to struggle at becoming more like United. Allegiant illustrates a bounded model he respects, connecting smaller cities with destinations such as Las Vegas. He counts it among the four US airlines expected to be profitable this year.

11. Network depth makes hub entry exceptionally difficult

• United’s discipline means flying through its hubs rather than building unrelated point-to-point routes. Kirby explains that on American’s Dallas–Los Angeles service, 75% of passengers connect from elsewhere. A competitor without that feed fights for just 25% of the market. “If you’re not first, you’re last” applies by market, rather than to an airline’s overall size.

• Kirby has closed three hubs and opened none during his 30-year career. United would like more southeastern presence, but he estimates competing with Atlanta and Charlotte would require 600 daily flights. Raleigh-Durham might have the city size, yet lacks sufficient gates and runways. Expanding an airport takes a decade or more; an attractive vision does not remove those constraints.

• His advice to new airline CEOs starts with route-level profit and loss. A broken network cannot be rescued by excellence elsewhere. When managers explain persistent losses as “strategic flying,” citing corporate accounts or credit cards, Kirby says to cancel the flying. “If you have bad routes that lose money, you cannot make the airline work. It is impossible.”

12. Loyalty and partnerships must earn their economics through the experience

• Kirby rejects treating disclosed credit-card revenue as equivalent to profit. More than 50,000 passengers daily fly United using credit-card miles, and carrying them has a cost. Collison recasts the popular “credit card business with airplanes attached” claim as a financial-reading error. Kirby agrees that analysts are “looking at the revenue side and ignoring the cost.”

• Frequent-flyer changes should be judged by brand loyalty, Kirby says, rather than program mechanics alone. Limited first-class seats require calibrating benefits, while overcrowded clubs call for capacity investment. United more than doubled club square footage from before COVID and plans another 50% increase. Pandemic closures created a rare opportunity to rebuild without inconveniencing normal passenger volumes.

• Alliances extend the network: Kirby thinks United operates four times as many Frankfurt and Munich flights as American and Delta combined, with about 60% of passengers connecting onward. The problem is consistency across airlines. After years of requests, he threatened to end partnerships without seat assignments for United customers. His aim is for a partner flight to feel like an extension of United.

• Thompson asks where Kirby’s enthusiasm for consolidation stops. Kirby returns to customer experience, then adds an ambition for an American flag carrier people feel proud of. He says overseas competitors supply two-thirds of long-haul seats into the US although only 40% of customers originate overseas, attributing the imbalance partly to state subsidies. He wants United able to compete with those airlines on every dimension.