Booking Holdings 10-K Analysis: What GAAP Earnings Hide About the AI Bet

Booking Holdings — the world's largest online travel agency with $186 billion in annual gross bookings — just generated $9.1 billion in free cash flow, the best in its history. GAAP earnings fell 8%. The market sold off 8%. The 10-K explains why both reactions miss the point.

The disconnect is hiding below the operating line. Booking's segment adjusted EBITDA grew 20.5% to $9.85 billion — by far its strongest year — but $2.7 billion in accounting artifacts turned that record performance into a GAAP earnings decline. Chief among them: $1.4 billion in foreign exchange losses on euro-denominated debt, a $457 million write-down of the KAYAK metasearch subsidiary, and $360 million in non-cash convertible note charges. None of these affected cash flow. None reflect the operating business.

The thesis: the filing reveals a self-funding cash flow machine whose operating economics are the strongest in the online travel agency sector, but the market must decide whether to value it on GAAP ($166 EPS, 32x P/E) or adjusted ($228 EPS, 18x P/E) — and only if the deepening merchant moat can defend against the AI disruption that already triggered a $457 million write-down of Booking's own subsidiary.

The Business That Grew 20% While Earnings Fell 8%

Booking Holdings is a company whose financial statements tell two opposite stories. Revenue grew 13.4% to $26.9 billion. Operating income grew 17%. Segment adjusted EBITDA — the measure management uses to run the business — grew 20.5% to $9.85 billion. Free cash flow hit $9.1 billion, a 34% margin that funds every dollar of growth, capital return, and debt reduction without touching external capital markets. By every operating metric, FY2025 was Booking's best year.

But GAAP net income fell 8% to $5.4 billion. Pre-tax income fell 6.3%. The bridge between those two realities reveals exactly where the $1.67 billion operating improvement went:

Booking Holdings' segment adjusted EBITDA grew 20.5% to $9.85 billion in FY2025 — its strongest year ever — but GAAP net income fell 8% because $2.7 billion in below-the-line accounting artifacts consumed the entire $1.67 billion operating improvement. Every dollar of the record-setting operating gain was absorbed by below-the-line items — foreign exchange losses, an impairment charge, and non-cash interest charges — that had nothing to do with booking hotel rooms.

The cash flow tells the real story. The $9.1 billion in FCF self-funds everything Booking does:

This is a business that funds its $700M AI investment, its marketing, its Connected Trip expansion, and every other operating cost before generating $9.1 billion in excess cash — and then returns 84% of that cash to shareholders while paying down debt and maintaining A-/A3 investment-grade ratings. No equity issuance. No net debt increase. The 4.2% annual share count reduction mechanically boosts per-share metrics: FCF per share grew 20.1% to $278.41 even though total FCF grew 15.1%.

"We remain focused on our key priorities, especially advancing our use of Generative AI to enhance the value we deliver to both travelers and partners. Savings from the Transformation Program will help us continue to reinvest strategically in the business in 2026 to support sustained, long-term growth."

The CEO's framing is telling. Management sees the $550M in transformation savings not as margin expansion but as investment fuel — the money gets recycled into the $700M AI bet. For investors, the question is whether you value Booking on the operating reality (Adj EBITDA +20.5%, FCF $9.1B) or the GAAP headline (NI -8%). At ~$4,070 per share, the gap between those frameworks is the difference between 17.8x adjusted earnings and 32.2x GAAP earnings — a valuation spread that tells you more about accounting than about the business.

The $1.4B Accounting Artifact That Distorts Everything

The single largest driver of GAAP earnings decline is a number that has nothing to do with travel bookings: $1,428 million in foreign exchange losses on EUR-denominated debt. To understand why this number will haunt Booking's GAAP earnings for decades, look at the debt schedule in the 10-K.

Of Booking's 26 outstanding debt tranches, 22 are denominated in euros. Only two are USD: a $1 billion note due June 2026 and a $500 million note due March 2028. After those mature, the entire long-term debt portfolio will be euros — notes extending all the way to May 2046 at rates ranging from 1.8% to 4.5%.

Booking Holdings carries 22 of 26 debt tranches in euros with maturities extending to 2046, creating a structural FX loss of $1.4 billion in FY2025 that management excludes from adjusted earnings as "non-recurring." But after the two remaining USD notes mature by 2028, the entire long-term debt portfolio will be EUR-denominated, making this loss permanently recurring. The $1.95 billion year-over-year swing — from a $526M gain to a $1,428M loss — is a pure function of dollar-euro exchange rates, not hotel bookings.

Management's response is to strip this out entirely. In the 8-K reconciliation, FX on EUR debt accounts for $42.28 per share — 68% of the total $62.49 GAAP-to-adjusted EPS gap. The other adjustments are more defensible: the $457M KAYAK impairment is genuinely non-recurring, the $360M in convertible note charges disappear now that those notes have been retired, and the $204M in intangible amortization is a standard non-cash exclusion.

The honest truth sits between the two extremes. If you include the structurally recurring FX losses but exclude the genuinely one-time items (KAYAK impairment, convertible charges, transformation costs), adjusted net income is approximately $5,588M — roughly $171 per share. That's 3% above GAAP, not 38%. Management's preferred frame systematically smooths what will be a recurring ±$1 billion annual swing.

One structural positive: the convertible notes matured in May 2025, ending the $523M annual non-cash amortization charge. In FY2026, GAAP interest expense drops to the ~$1,094M adjusted level. This creates a mechanical ~$16/share GAAP EPS tailwind that will make year-over-year comparisons look dramatically better — but it's a one-time accounting cleanup, not operational improvement. Investors who mistake the FY2026 GAAP "recovery" for business acceleration will be misreading the filing.

The AI Paradox: $700M Bet Meets $457M Write-Down

Booking's 10-K contains the most explicit AI disruption language of any major online travel agency filing — and simultaneously the largest AI investment commitment in the sector. The same company making these statements is spending $700M to build what it acknowledges might cannibalize its existing business.

"Other large technology platforms and AI-native competitors are developing Gen AI-powered assistants and agents that can search, compare, recommend, and facilitate travel and dining reservations directly within their search engines, operating systems, messaging platforms, or 'super-apps.' These offerings may reduce consumers choosing to visit dedicated online travel platforms."

This isn't hypothetical risk disclosure language. The 10-K provides filing evidence that this disruption is already happening — inside Booking's own portfolio:

"The impairments were primarily driven by a reduction in the forecasted cash flows for KAYAK, reflecting its meta-search business being impacted by expected increases in customer acquisition costs."

"Expected increases in customer acquisition costs" is the filing-safe way of saying AI Overviews and AI agents are eroding KAYAK's metasearch traffic funnel. The total write-down: $180M in goodwill plus $277M in intangibles, leaving only $203M in remaining KAYAK goodwill.

Booking Holdings wrote down its KAYAK metasearch subsidiary by $457 million because of "expected increases in customer acquisition costs" from AI competition, while simultaneously committing $700 million in AI operating expenses for FY2026 — a bet that nets +$250 million after $550 million in transformation savings offset the cost.

"Given the stronger-than-expected early results of the Transformation Program, in the third quarter of 2025, we raised our expectation for the ultimate annual run-rate savings to a range of $500 to $550 million... We have enabled approximately $550 million in annual run-rate savings and we expect to realize these run-rate savings by the end of 2026."

The net math is what matters. The $700M AI investment creates a ~$300M EBITDA drag after ~$400M in expected incremental revenue. But the $550M transformation savings — which exceeded the original $400-450M target — more than offset the drag, yielding a net +$250M favorable impact. Management guided FY2026 adjusted EBITDA growth of 10-14%, implying they expect margins to hold or expand despite the AI spend.

The competitive context makes this more urgent. Every major OTA is investing in AI: Expedia is building AI-powered recommendations, Airbnb is executing an AI-native app rebuild, and Trip.com's TripGenie is already deployed in 200+ countries with 200%+ user growth. AI investment is industry-wide. But only Booking simultaneously wrote down its own AI-vulnerable subsidiary — the KAYAK impairment is the filing evidence that separates acknowledged disruption from everyone else's aspirational press releases.

The Moat Deepens Even as Growth Slows

There are two stories in Booking's operating metrics, and they pull in opposite directions.

The first story is a deepening moat. Merchant bookings reached 70% of gross bookings in FY2025 — up 7 percentage points from 63% in FY2024. That means Booking now processes payments directly for roughly 900 million room nights annually, giving it customer data, payment data, and the ability to bundle flights, hotels, and attractions into a single managed experience.

"The mix of our total gross bookings generated on a merchant basis across the company was 70% in 2025, an increase from 63% in 2024 due to the ongoing shift from agency to merchant bookings at Booking.com... However, this results in additional expenses for personnel, payment processing, chargebacks (including those related to fraud), and other expenses related to these transactions."

Management is unusually candid about the cost. The 10-K quantifies $381M in incremental merchant transaction expenses — about 2.1% of merchant revenue. But operating margins expanded 100 basis points to 32.8% despite this drag, meaning efficiency gains elsewhere more than absorbed the merchant cost increase.

Booking Holdings' merchant mix reached 70% of gross bookings in FY2025, up 7 percentage points from 63% in FY2024, giving the company payment processing control over roughly 900 million room nights — but room night growth is decelerating from 9% to a guided 5-7% in Q1 2026, meaning the moat is deepening even as volume narrows.

The Connected Trip strategy — bundling hotels, flights, attractions, and car rentals into a single AI-powered experience — is the merchant model's strategic payoff:

"We continue to grow our Connected Trip verticals in 2025, including 37% year-over-year flight ticket growth, and about 80% attraction ticket growth off a small base."

Flight tickets reached 68 million (+37%), attractions grew roughly 80%, and multi-component bookings — where a traveler books multiple services in one session — grew in the "high 20% range." The merchant model enables all of this because Booking controls payment processing across every component of the trip.

The second story is deceleration. Q1 2026 room night guidance of 5-7% represents a meaningful step down from 9% in Q4 2025 and 8% for the full year. Revenue is decelerating less — Q1 guidance is 14-16% — partly because of a ~7% FX tailwind and the take rate expansion from 14.33% to 14.46%.

The deceleration is not unique to Booking. Expedia guided 6-9% revenue growth for FY2026, also slowing. But Trip.com Group is still growing at 16%+ with no deceleration, and Airbnb is stable at 8-10%. The volume slowdown is concentrated in Western markets — the OTA model itself is not structurally decelerating, which means Booking's growth regime change reflects market maturation, not competitive displacement.

What Would Prove This Wrong

The thesis — that Booking's operating business is the strongest in the OTA sector and the GAAP decline is an accounting illusion — depends on four testable assumptions. Each has a specific metric and threshold that would trigger revision.

If merchant mix plateaus below 75%, the moat deepening thesis weakens — the merchant model enables Connected Trip bundling and the data flywheel that powers AI personalization. If AI incremental revenue fails to reach $200M by H2 2026, the $700M becomes pure margin destruction rather than a rational investment. If the GAAP-to-adjusted gap widens beyond $70 per share — meaning FX losses persist or new impairments emerge — the market may shift to GAAP framing despite unchanged operations. And if the credit rating agencies flag the negative equity or leverage exceeds 2.5x, the capital return machine slows.

Booking Holdings' thesis is testable: if merchant mix plateaus below 75%, AI incremental revenue fails to reach $200 million by H2 2026, or the GAAP-to-adjusted EPS gap widens beyond $70 per share, the investment case shifts from a cash flow machine with temporary GAAP noise to a business facing structural disruption with permanent accounting headwinds.

The peer comparison reveals why this matters. Booking's three distinguishing features among major OTAs — structural FX losses on EUR debt ($1.4B), negative equity from buybacks (-$5.6B), and an AI impairment of its own subsidiary ($457M) — are all GAAP-depressing factors that don't affect operating performance. On the metrics that reflect actual business quality — operating margin (32.8% vs EXPE 12.7% and ABNB 20.8%), SBC efficiency (2.05% of revenue vs ABNB 13.0%), and SBC-adjusted FCF margin (31.6% vs ABNB 24.5%) — Booking leads every major competitor.

At ~$4,070 per share, Booking trades at 17.8x adjusted EPS ($228), implying continued mid-teens earnings growth. The filing supports this: adjusted EBITDA grew 20.5%, management guides mid-teens adjusted EPS growth for FY2026, and the self-funding flywheel mechanically boosts per-share metrics 4-5% annually through buybacks alone. But the GAAP P/E of 32x on declining earnings creates a framing trap. If the market shifts to GAAP valuation — which it could if EUR weakness persists or if adjusted-metric skepticism grows — the stock re-rates sharply despite unchanged operations.

The four metrics above will tell you which frame is right. Watch merchant mix for moat depth. Watch AI revenue disclosure for whether the $700M bet is paying off. Watch the GAAP-to-adjusted gap for whether the accounting noise is truly temporary. And watch the credit rating for whether the capital return machine has limits. This is not a permanent conviction — it's a time-bound hypothesis that the next two to three quarters of data will confirm or revise.

Frequently Asked Questions

Why did Booking Holdings' GAAP earnings decline while revenue grew 13%?

Three below-the-line items consumed the entire $1.67 billion operating improvement: FX losses on EUR-denominated debt of $1,428M (versus a $526M gain in FY2024 — a $1.95 billion swing), KAYAK impairment of $457M, and convertible note non-cash charges of $360M. Operating income actually grew 17% and segment adjusted EBITDA grew 20.5% — the GAAP decline is entirely below the operating line. The largest single factor was the FX swing, which alone turned what would have been 8% GAAP earnings growth into an 8% decline.

Is the $1.4B FX loss a real cost or just accounting noise?

Both. It's "real" in that Booking will eventually need euros to repay 22 of 26 debt tranches — bonds extending all the way to 2046. But it's "noise" in the current period because: it's unrealized with no cash changing hands until maturity, Booking's 90% international revenue creates a natural cash flow hedge, and it reverses when the dollar weakens. In FY2024, the same debt generated a $526M gain. The FX line will swing ±$1 billion annually — it tells you about exchange rates, not about the business.

What is Booking Holdings' $700M AI investment for?

It's incremental operating expense (not capital expenditure) for FY2026, representing a 4.1x increase from approximately $170M in FY2025 AI-related spend. It covers personnel, cloud computing, and technology development for AI-powered personalization, Connected Trip enhancements, and automated customer service. Management expects approximately $400M in incremental revenue and roughly $300M in net EBITDA drag — but this is substantially offset by $550M in transformation program savings, yielding a net +$250M favorable impact on adjusted EBITDA.

Why did Booking write down KAYAK?

The 10-K reveals the specific trigger: "reduction in forecasted cash flows for KAYAK, reflecting its meta-search business being impacted by expected increases in customer acquisition costs." This is a euphemism for AI disruption. KAYAK's metasearch model depends on acquiring traffic from Google and converting it to OTA referrals. AI Overviews and AI agents are reducing the value of this traffic funnel. Total impairment: $180M goodwill plus $277M intangibles equals $457M. Remaining KAYAK goodwill is only $203M.

Is room night growth decelerating?

Yes. Q1 2026 guidance is 5-7% room night growth, down from 9% in Q4 2025 and 8% for FY2025. However, revenue is decelerating less — Q1 revenue guidance is 14-16% (7-9% constant currency) because of take rate expansion (+13bps to 14.46%) and FX tailwinds of approximately 7% in Q1. The growth deceleration is shared with Expedia (guided 6-9% for FY2026) but not Trip.com (16%+ growth with no deceleration). Easter timing may also shift some Q1 bookings to Q2.

How does Booking Holdings compare to Airbnb and Expedia?

Booking leads on operating margin (32.8% vs Airbnb 20.8% vs Expedia 12.7%), EBITDA margin (36.9% vs 35% vs 23.8%), and SBC efficiency (2.05% of revenue vs 13.0% vs 2.7%). Airbnb leads on headline FCF margin (38% vs 33.8%), but Airbnb's SBC-adjusted FCF margin (24.5%) falls well below Booking's (31.6%). Expedia is the cheapest OTA on every valuation metric (7.3x EV/EBITDA vs 13.2x for Booking) but has weaker margins, slower growth, and brand fragmentation. Booking's unique features — EUR debt FX losses, negative equity, KAYAK impairment — are all GAAP-depressing; on operating metrics, Booking clearly leads.

Can the buybacks continue at $6.4B per year?

Yes, and they're self-funding. The $9.1 billion in FCF covers $6.4B in buybacks, $1.2B in dividends, and $1.3B in debt repayment with $160M surplus. No external capital is needed. At the current price of approximately $4,070, the buyback yield is 5.0% — above the weighted-average EUR debt cost of roughly 3.5-4.0%, meaning buybacks are value-accretive. The $21.8B remaining authorization supports approximately 3.4 years at the current pace. Credit ratings (A-/A3) are stable. The constraint isn't cash flow — it's whether management maintains discipline at elevated prices. FY2025 average purchase price was $4,960, above the current market.

What does Booking Holdings' negative equity mean?

Booking's total equity is -$5,578M because $54.3 billion in cumulative treasury stock overwhelms $40.7 billion in retained earnings. This makes traditional metrics (ROE, P/B, debt-to-equity) meaningless. But negative equity is a feature of the capital return strategy, not a sign of distress. The company has $17.3B in cash, A-/A3 credit ratings, net debt of only $1.5B, and net debt/EBITDA of just 0.15x. Booking could stop all buybacks today and rebuild positive equity within three years from retained earnings alone.

What is the Connected Trip strategy?

Connected Trip is Booking's strategy to bundle multiple travel components — hotels, flights, attractions, and car rental — into a single AI-powered booking experience. FY2025 progress: flight ticket sales grew 37% to 68M, attraction tickets grew approximately 80% (off a small base), and multi-component bookings grew in the "high 20% range." Direct channel room night mix reached the "mid-fifties percentage." The strategic rationale: bundling creates switching costs, enables data-driven personalization, and increases revenue per customer. The merchant model (70% of bookings) enables this because Booking controls payment processing across all components.

What is the stock split about?

The 25-for-1 stock split, effective April 2, 2026, reduces the share price from approximately $4,070 to approximately $163. There is no fundamental impact. At around $4,070, Booking was one of the highest-priced stocks in the market, limiting retail investor access. The split enables potential Dow Jones Industrial Average inclusion (the Dow is price-weighted, so a $4,000+ stock would dominate the index). Historically, stock splits signal management confidence and can modestly improve trading liquidity.

Methodology

Data sources: Booking Holdings Inc. FY2025 10-K (filed February 18, 2026, CIK 0001075531, accession 0001075531-26-000009). Booking Holdings Inc. Q4 2025 8-K Earnings Press Release (filed February 18, 2026, accession 0001075531-26-000008). Peer data: Expedia Group (EXPE) FY2025 actuals from 8-K earnings press release, Airbnb (ABNB) FY2025 actuals from earnings press release, Trip.com Group (TCOM) estimated from Q1-Q3 2025 actuals plus Q4 consensus (FY2025 results expected February 25, 2026).

Analysis pipeline: BigQuery core metrics (191+ calculated metrics per company), Filing Intelligence 5-pass analysis (narrative quality, accounting quality, hidden liabilities, risk landscape, segment performance), 8-K earnings extraction, raw 10-K XBRL viewer footnote extraction (R52, R70-R73, R75-R79, R88-R90, R93-R95, R97, R99).

Limitations: Trip.com FY2025 not yet reported (reports February 25, 2026) — comparison uses Q1-Q3 actuals and Q4 consensus estimates. Airbnb SBC at 13% of revenue distorts headline FCF comparisons — SBC-adjusted FCF computed for fair comparison. Booking's adjusted metrics exclude structurally recurring FX on EUR debt — an "honest" adjusted EPS (~$171/share including FX) was computed alongside management's preferred $228. Expedia's gross margin (~90%) uses a different cost structure definition than OTA peers — comparison focused on operating margin and EBITDA margin. Market cap varies significantly by date ($174B at December 31 vs ~$130B at February 21) — current price used for valuation metrics.

Disclosure: This analysis is for informational purposes only and does not constitute investment advice. MetricDuck Research holds no positions in Booking Holdings Inc. (BKNG), Expedia Group Inc. (EXPE), Airbnb Inc. (ABNB), or Trip.com Group Limited (TCOM). All data sourced directly from SEC filings or public financial disclosures.