AAL: a leveraged view on the price of oil

Record revenue, eleven cents a share, and a barrel that lands on a sliver of equity

American Airlines reported the highest quarterly revenue in its history on July 23, 2026 — $16,735 million, up 16.3 percent year over year — and earned eleven cents a share on it. The same day it guided full-year adjusted earnings to a range of ($0.65) to $0.65 a share, a range that straddles zero. Both of those statements are true at once, and the reason they are both true is the whole case.

At the September 1, 2026 close of $12.95, American's equity is worth $8.6 billion. The enterprise it sits on top of — counting debt and finance leases, operating lease liabilities, and the net pension and postretirement obligation, less unrestricted cash and short-term investments — is worth $37.7 billion. The equity is 22.7 percent of it. Net obligations on that same basis are $29.1 billion, or 3.4 times the entire market value of the company. Stockholders' equity on the June 30, 2026 balance sheet is negative $4.0 billion.

The reason that ratio matters right now is on the crude tape. WTI averaged $64.51 a barrel in February 2026, the last full month before hostilities between the United States and Iran began at the end of that month. It peaked at $102.13 in May 2026 and averaged $83.90 in August 2026, the most recent published month — still 30 percent above where it sat before the war, having given back roughly 48 percent of the spike. American closed February 2026 at $13.32 and closed September 1, 2026 at $12.95. Crude is up roughly a third over that span and the equity is slightly lower, which is the observation this case is built to examine rather than the conclusion it reaches. All three crude figures are monthly averages of daily EIA spot prices at Cushing and carry their own measurement months; the widely reported $67 settle of February 27, 2026 and the print above $119 a week later are daily marks from press accounts and are used nowhere in this case.

The path between those two endpoints matters more than the endpoints, and it is the strongest evidence against reading this equity as a simple inverse of crude. The stock did not fall through the war. It bottomed on the shock, at a $10.74 close in March 2026, and then rallied hard — through April and May, while crude was still climbing to its peak, and again in June, when crude fell 19 percent and the stock rose 21 percent in the same month. It reached a 52-week high of $18.79 inside that window. From there it has round-tripped: $12.95 today is 31 percent below the high, and 28 percent below the $18.07 close of June 30, 2026, the highest month-end mark in the series. Crude did not do that. It averaged $84.81 in June against $83.90 in August 2026 — essentially flat. The equity gave back a third of its value while the commodity stood still, which is not what a levered inverse position on crude does. Whatever the market repriced in June, it was not only oil, and it did not stay repriced. Section 1 takes that apart month by month.

That ratio is what makes this equity a leveraged view on the price of oil, and it is where the case now opens. American burns roughly 4,821 million gallons a year, so ten cents on the gallon is about $482 million of annual pre-tax income — 5.6 percent of the entire market value of the company, and fifty cents is 28 percent. No competitor is more exposed on that axis, and the reason is not consumption but the equity underneath it: a dollar of enterprise effect lands on a base 4.4 times smaller than the enterprise. Section 1 sizes that lever and then tests whether the market has ever paid it, which is a different question with a different answer. The Institute takes no view on the direction of crude, and states one conditional in its place: we do not know when crude falls, or whether it falls from here at all, but if it does, it is a tailwind for this company and a disproportionately large one. That is arithmetic about earnings, not a forecast about the tape, and the two are kept apart throughout.

Two assets are carried at zero throughout and should be visible before any multiple is read. The enterprise value above deducts cash and nothing else: it assigns no value to the federal tax carryforwards, on which American had recorded a deferred tax asset of $4,095 million at December 31, 2025 — $6.19 a share gross against a $12.95 stock, or 48 percent of market capitalization — and no value to AAdvantage beyond the $11.0 billion of debt secured against it. Neither is netted into any multiple in this case. Both are argued on their own terms, the carryforwards in section 11 and the loyalty program in section 13, and in both cases the honest answer is narrower than the bull version. They are printed here because a reader comparing American's multiple to Delta's should know what the numerator leaves out.

What that structure does is convert an ordinary operating result into a violent equity outcome in both directions. American's second-quarter operating margin was 2.7 percent. United, running 95 percent of the revenue in the same quarter under the same fuel curve and the same demand environment, earned $805 million of net income against American's $71 million. The fuel shock did not create that gap. It exposed a margin that was already too thin to absorb one.

The liquidity picture is the part most readers get wrong, and it is the reason this case is written from the cash flow statement rather than the income statement. American generated $4.7 billion of operating cash flow in the first half of 2026 against a net loss of $311 million. That looks like a company whose earnings understate its cash generation. It is not. Roughly 72 percent of that operating cash flow is the growth in customer float — money collected for flights not yet flown and miles not yet redeemed. The float balance itself is permanent, as deferred revenue is in any business paid before it delivers; what enters the cash flow statement is only the change in it, and a change is not earnings. Full-year 2025 free cash flow, on the same definition, was negative $680 million.

The Institute's conclusion, reached by following the evidence rather than the premise: the discount is largely deserved, but not for the reason the headline suggests. The commercial franchise is working — record revenue, corporate share gains, premium demand. What is not working is the relationship between the capital structure and the margin available to service it.

Valuation at a glance

Every figure below is derived, sourced and qualified in the section named beside it. The table is a signpost, not a substitute for the sections — in particular the last four lines are an arithmetic sensitivity and not a target, and the two $0 lines are a statement of what the enterprise value deliberately excludes rather than a view that those assets are worthless.

Two clocks, both stated, and on the crude row three. Equity prices are struck at the September 1, 2026 close; balance sheet claims are measured at June 30, 2026 as filed on Form 10-Q. The crude figures are EIA Cushing spot, monthly averages of daily prices; February 2026 is the last full month before hostilities between the United States and Iran began at the end of that month, and August 2026 is the most recent published month, released September 2, 2026. The crude and equity endpoints are therefore one calendar day apart and are not a matched pair. Net obligations are debt and finance lease obligations, plus all operating lease liabilities, plus the net pension and postretirement obligation, less unrestricted cash and short-term investments. The last two rows hold net obligations constant at the June 30, 2026 figure and assume no debt paydown and no change in share count; they show what the equity is worth at each peer multiple, not which multiple is right, and they run in both directions.

The fuel lever, sized and then tested

Every section to this point has treated the fuel shock as the thing that exposed the capital structure. It is also the most commonly proposed cure for it, and that proposition deserves the same treatment as the others in this case: size it first, then test whether the market has actually paid it. Those two exercises give sharply different answers, and the gap between them is the finding.

The Institute takes no view on the direction of crude oil and does not forecast one here. What follows is the transfer function — how much of a fuel move lands on this particular equity — and the evidence on how much of it the tape has historically delivered. The reader supplies the oil view; this section supplies the multiplier and its limits.

There is one position the Institute will state, and it is deliberately a conditional rather than a call. We do not know when crude falls, or whether it falls from here at all, and nothing in this case attempts to time the barrel — the timing question is where most oil-linked equity theses go wrong, and it is not answered here. Conditional on crude falling, however, the effect on American is a tailwind, and an unusually large one measured against the size of the company’s equity: at roughly 4,821 million gallons a year, a dime off the gallon is about $482 million of annual pre-tax income, or 5.6 percent of the entire market value of the company, for every year the lower price persists. That is a statement about the company’s earnings and it is arithmetic.

The benefit does not stop at earnings, and the second leg is the one that matters more for an equity this thin. Because a fuel saving lands in EBITDA, it deleverages the company without retiring a dollar of principal. Net obligations of $29,119 million against trailing EBITDA of $3,090 million are 9.42 times. A dime off the gallon carries EBITDA to $3,572 million and the ratio to 8.15 times; a quarter takes it to 6.78 times; fifty cents to 5.29 times. Leverage is what has compressed this multiple, the covenants are measured against EBITDA, and the 2028 refinancing will be priced off the same ratio — so the identical dollar of fuel saving works on the earnings and on the leverage at once. That is one dollar producing two effects, not two benefits to be added together, and the case does not add them.

Two qualifications belong with that, and both cut the same way. The deleveraging is a ratio effect only: the $28,594 million of scheduled principal does not shrink because jet fuel got cheaper, the $7.3 billion coming due in 2028 does not move, and the interest bill is unchanged. And the arithmetic runs in reverse with equal force — crude rising by the same increments raises the ratio just as fast, which is a fair description of what has happened since the war began. A lever is not a direction, and this one is bolted on in both directions.

Whether the equity re-rates on those earnings is a separate question, and it is the one this section actually tests. The answer the record gives is not encouraging: across 66 months the correlation of the stock to jet fuel is -0.10, to WTI -0.16, and to the crack spread — presently 44 percent of the jet-fuel barrel — +0.004, which is no relationship at all. The Institute holds both readings at once and asks the reader to do the same. The operating tailwind is real, it is sized here, and it does not require anyone’s oil forecast to be right in order to arrive as cash. The re-rating is not something the five-year record supports assuming, and a reader who owns this equity for the fuel lever should understand that the cash and the multiple have historically been two different events.

The lever, measured

American consumed $4,881 million of fuel in the June 2026 quarter at $4.05 a gallon, against $2.29 in the same quarter of 2025 — a 77.1 percent increase in the price of the input. That is roughly 1,205 million gallons in the quarter, or about 4,821 million gallons annualized at the same volume. The consumption figure is what converts a price move into a dollar figure, and the dollar figure is what has to be measured against the equity rather than against the enterprise.

Institute calculation on 4,821 million gallons a year, being the June 2026 quarter’s consumption of $4,881 million at $4.05 a gallon, annualized at constant volume. Pre-tax is the correct basis: the federal net operating loss carryforward set out in section 11 means a marginal dollar of pre-tax income is not currently taxed in cash. Market capitalization is $8,572 million at the September 1, 2026 close. This holds capacity, fares, hedging and the crack spread constant and is therefore a sensitivity, not a forecast of earnings.

Ten cents a gallon is roughly $482 million a year of pre-tax income, or about 5.6 percent of the entire market value of the company, every year it persists. Fifty cents is 28 percent. No competitor is more exposed on this axis, and the reason is not that American burns more fuel than Delta or United — it is that American has less equity underneath the same barrel. The common is 22.7 percent of enterprise value, so a dollar of enterprise effect lands on an equity base roughly 4.4 times smaller than the enterprise itself. Operating leverage on the income statement, multiplied by financial leverage on the balance sheet. That is the whole of the structural claim, and it holds whichever way fuel moves.

The share price and the jet fuel curve

Fuel is the second largest line in an airline's cost base and the one management controls least — American carries no hedges against it, a point developed below — so the intuition that American's share price should move inversely with the jet fuel curve is a reasonable prior. It is also testable. The chart below plots the Gulf Coast jet fuel spot price against American's month-end close over 67 months, and the statistics beneath it test the relationship rather than illustrate it.

Jet fuel is the U.S. Gulf Coast kerosene-type jet fuel spot price, EIA series EER_EPJK_PF4_RGC_DPG, retrieved September 2, 2026 from the Federal Reserve Bank of St. Louis mirror MJFUELUSGULF, series last updated August 5, 2026; August 2026 is not yet published and the series could not be independently cross-checked against EIA’s own monthly page, which returned no data. The share price is the month-end close from Stooq, retrieved September 2, 2026 and verified against an independent September 1, 2026 print of $12.95 and against annual December closes for each year in the window. The two clocks are not identical: the fuel figure is a monthly average of daily spot prices, while the share price is a single month-end observation. Correlations on monthly percentage changes use the 66 overlapping change observations.

The relationship is not there in the ordinary case. Over the full window the monthly-change correlation is -0.10, a t-statistic of -0.79, and an R-squared of 0.010 — fuel explains about one percent of the variance in American’s monthly return, which is indistinguishable from zero. Shifting the fuel series forward or backward by up to three months does not rescue it; the strongest of the seven lags reaches only 0.176. A reader who buys or sells this stock on the fuel tape in a normal month is trading noise.

What the data does support is narrower and more useful. The sign is correct throughout — fuel is a cost, and the coefficient never turns positive — and the relationship strengthens sharply as the fuel move gets larger. Restricted to months when fuel moved more than ten percent the correlation is -0.38, and above fifteen percent it is -0.49. In 2026 alone, when Gulf Coast jet fuel ran from $2.03 in January to $3.94 in May, a rise of 94 percent, the correlation is -0.74 and survives dropping any single month. That 2026 window is not a random draw. It is the war window: crude averaged $64.51 in February 2026, the last full month before hostilities with Iran, ran to $102.13 in May 2026 — 58 percent above the pre-war average — and has since settled back to $83.90 in August 2026, which is -18 percent off the peak and still 30 percent above where it started. Roughly 48 percent of the spike has come out. The measured relationship is strongest inside exactly that window, which is what makes it worth printing and also what makes it fragile: a single event supplies most of the variation.

The month-by-month path is printed below rather than summarized, because it is the part of this evidence that argues against the Institute's own lever and a reader is entitled to see it unaggregated.

Crude is the EIA Cushing monthly average and the equity is the month-end close, from the same two series used for every correlation in this case. The final row is deliberately mixed-clock and should not be read as a matched pair: it sets the August 2026 crude average against the September 1, 2026 equity close, one calendar day apart, because those are the most recent published observations of each. The last column tests only the sign of the two moves, not their size.

Three of those six months move opposite to crude and three do not, and the sign test flatters the relationship even at three. March fits and fits well: crude rose 42 percent and the equity fell 19 percent. June fits emphatically. But April and May do not — the stock rose 9 and 25 percent while crude climbed to its wartime high — and July does not, because both fell together. August counts as opposite only on sign: a 4.3 percent move in crude against a 15 percent move in the equity is not a crude explanation of anything.

The round trip makes the same point at the level of the whole window. The stock reached a 52-week high of $18.79 inside the war and now trades at $12.95, 31 percent lower; measured from the $18.07 close of June 30, 2026, the highest month-end mark in the series, it is 28 percent lower. Crude did not do that. It averaged $84.81 in June against $83.90 in August 2026, a move of 1 percent. A commodity that went nowhere cannot account for an equity that gave back a third of its value, and the July 23 guidance reduction is the obvious candidate for what did. The 52-week high is a trailing statistic measured at the September 1, 2026 close and the Institute has not independently pinned the session on which it was set; the June 30 month-end close is dated exactly, and both are printed for that reason.

The Institute also tested the version of this that a reader would reasonably propose next: that the equity anticipates crude rather than following it — that the April and May rally was the market pricing an expected peace before the barrel registered it. On the same 66 months, the correlation between the stock's return and crude's return one month later is +0.09, and two months later it is -0.02. An anticipation signal would be negative at those horizons — the stock rising now against crude falling later — and neither is. A reading of +0.25 appears at three months, but its sign is wrong for the peace story, since it pairs a rising stock with rising crude, and it is one cell out of 7 tested, which is the pattern multiple comparisons produce on their own. The lever set out earlier in this section is structural and it holds. What this window does not support is the further claim that the tape reliably pays it.

That is a tail relationship, not a trading rule, and the 2026 window carries an explicit caveat: seven observations is a small sample, and the same seven months contain the record June quarter and the July guidance reduction, either of which could be doing the work the fuel price appears to be doing. Read carefully, though, the result corroborates rather than contradicts the argument the rest of this case makes. Fuel does not drive this equity in the ordinary course because the gap between American and its peers is not a fuel gap. It becomes visible only when the shock is violent enough to overwhelm a margin that had no room to absorb it — which is the point. The fuel spike did not create the problem. It made an existing one legible.

Why the relationship bound in 2026: the absence of hedges and a closed strait

The statistics above establish that the relationship exists and when it binds. They do not explain why. Two facts do, and neither appears anywhere else in this case: American carries no fuel hedges at all, and the 2026 fuel curve is not a commodity cycle but a war.

Take the hedging first, because it is the one American controls. The FY2025 Form 10-K states that at December 31, 2025 the company had no fuel hedging contracts outstanding, that its policy is not to enter into transactions to hedge fuel consumption, and that absent a change of policy it will “continue to be fully exposed to fluctuations in fuel prices.” That is a deliberate choice rather than an oversight, and for two decades it was the right one — hedging cost the industry real money across a long stretch of falling and flat crude, and American is not alone in having abandoned it. But an unhedged carrier converts a commodity shock into an earnings shock with no lag, no floor, and nothing to smooth it. Every cent of the move lands in the quarter it happens.

The same 10-K sized the risk using its own three-year band: spot jet fuel ranged from roughly $1.83 to $3.82 per gallon between January 1, 2023 and December 31, 2025, and the filing warns that even a relatively small move can be material to operating results and liquidity. American realized $4.05 per gallon in the June 2026 quarter. Both clocks matter here: the $1.83–$3.82 band is a December 31, 2025 measurement of the prior three years, and the $4.05 is a realized average for the three months ended June 30, 2026. The company's own illustration of the risk was overrun to the upside within six months of filing it.

The cause is not a demand cycle. It is the Strait of Hormuz.

Source: International Energy Agency, Oil Market Report — April 2026, published April 14, 2026, and Oil Market Report — August 2026, published August 12, 2026. Every level above carries its own measurement date because they were measured on different ones; the April figures describe the first closure and the July figures describe conditions after the mid-June ceasefire broke down and the strait was effectively closed a second time in early July. mb/d is million barrels per day; kb/d is thousand barrels per day.

That table is the missing half of the fuel finding. The rise of 94 percent charted above is not a price move in search of an explanation — it is the refined-product consequence of the largest supply disruption the IEA has recorded.

The bypass corridors are real, and they are the reason the loss was thirteen million barrels a day rather than twenty. Pipeline capacity out of the Gulf that avoids the strait — the Saudi east–west line to the Red Sea, the Emirati line to Fujairah, the Iraqi line to Ceyhan — nearly doubled its measured throughput, from under 4 mb/d before the war to 7.2 mb/d by early April 2026. Several further expansions have been announced across the region and are frequently quoted in capacity terms. The Institute does not price them here. An announced capacity is a forward target set by its sponsor, not a measurement, and this case anchors only on flows that someone has actually observed. What the observed flows establish is the shape of the constraint: the bypass routes are a partial substitute, they were already carrying incremental volume within weeks, and even so they replaced roughly a third of what the strait had been moving.

The second point is the one an airline analyst is most likely to miss: crude has largely retraced and jet fuel has not. North Sea Dated traded around $130 per barrel in mid-April 2026, about $60 above pre-conflict levels; it ended July at $96.80 and was trading near $92 when the IEA went to press on August 12, 2026. Over the same stretch Atlantic Basin refining margins set all-time highs, with jet fuel, diesel and gasoline cracks surging on depleted stocks and lost product trade. The binding constraint migrated from crude supply to refining capacity and seaborne product trade, and jet fuel sits at the sharp end of both. A reader tracking Brent as a proxy for American's fuel bill would have concluded in August that the shock was two-thirds over. American's realized price says otherwise.

Price, not volume: what actually happened to the fuel bill

The company's own June-quarter disclosure permits an exact decomposition, and it is worth doing because the result is not what the headline suggests. Fuel expense of $4,881 million at an average of $4.05 per gallon implies roughly 1,205 million gallons consumed in the quarter; the prior-year figures, $2,663 million at $2.29, imply 1,163 million. Consumption therefore rose 3.6 percent while price rose 77.1 percent. Splitting the $2,218 million increase between the two gives a price effect of about $2,121 million and a volume effect of about $97 million — 96 percent of the increase is price. American did not fly its way into this. It was handed to the airline.

Gallons consumed are derived by the Institute from American Airlines Group Inc., Form 8-K Exhibit 99.1 furnished July 23, 2026, dividing aircraft fuel and related taxes by the disclosed average price per gallon; the two computed effects reconcile to the reported $2,218 million increase without residual.

Two cautions belong here, both of which cut against the argument rather than for it. First, the 2026 correlation of -0.74 still rests on seven observations, and nothing above changes that. What the mechanism does is make the sign and the timing credible rather than coincidental; it does not manufacture sample size. Second, and more uncomfortably, this is the part of the case with the shortest shelf life. Every market and balance-sheet figure elsewhere runs on the September 1, 2026 clock, but the strait was still effectively closed at the IEA's August 12, 2026 report and the situation has been reversing direction every few weeks. If flows resume, the tail relationship compresses and the fuel line normalizes faster than anything on the liability side of the balance sheet can. The capital structure is the durable finding. The fuel shock is the one that will date first.

The two legs are not the same size

A correlation is symmetric by construction and cannot say whether the stock falls on fuel spikes, rises on fuel collapses, or both. For a holder that distinction is the entire question, so it is measured separately: months are sorted by the size of the fuel move, then split by its direction, and the stock’s own return is summarized in each bucket.

Institute calculation on the same 66 monthly observations. Returns are month-over-month log changes in the month-end close, expressed in percent. The bucket sizes are small and are printed for that reason: the strongest row rests on 4 months. Nothing here is a distributional claim; it is a description of what the sample contains.

The asymmetry runs in the direction a holder would want and against the direction the correlation alone would suggest. In the 5 months when fuel rose more than fifteen percent, American returned -2.7 percent on average. In the 4 months when it fell more than fifteen percent, American returned +7.4 percent, and rose in 75 percent of them against a base rate of 48 percent across the whole sample. The unconditional mean month over this window is -0.2 percent, so the down-leg bucket is not inheriting a rising market — it is beating a falling one. The cleanest single observation is June 2026: jet fuel fell 26 percent and the stock rose 21 percent in the same month.

Two cautions belong with that, and neither is rhetorical. The bucket contains 4 months. And the arithmetic of a 23-percent equity sliver is symmetric whatever the historical buckets show — the same structure that would carry a fuel decline into the equity carries a further increase into it just as efficiently.

The naive version of the trade, and why it is wrong by an order of magnitude

There is a version of the fuel argument that is arithmetically seductive and empirically indefensible, and because it is the version a reader is most likely to construct unaided, the Institute prints it rather than leaving it to be constructed. Take one dollar a barrel. At forty-two gallons to the barrel and a constant crack spread, that is about $115 million a year of pre-tax income on American’s consumption. Capitalize it at the 7.37 times the market is already paying for the enterprise and it is roughly $846 million of enterprise value. Land that on an $8,572 million equity base and one dollar a barrel is worth about 9.9 percent of the common.

Now measure what the tape actually paid. One dollar a barrel is about 0.70 percent of the jet fuel price at the July 2026 monthly average of $3.403 a gallon. Run that through the 2026 slope of -0.60 — the most fuel-sensitive regime in the entire sample, chosen deliberately to give the mechanical case its best opponent — and one dollar a barrel is worth about 0.42 percent of the common. The mechanical figure is roughly 23 times the observed one.

Institute calculation from the $12.95 close of September 1, 2026. The mechanical column capitalizes the fuel saving as permanent EBITDA at 7.37 times with net obligations held constant. The observed column applies the 2026 regression slope of -0.60 to the same move expressed as a percentage of the July 2026 jet fuel price. Neither column is a price target. The mechanical column is shown because it is wrong in an instructive way; the observed column rests on 7 months and is not a forecast either.

The market is right and the arithmetic is wrong. A spot fuel move is not permanent EBITDA. Capacity adjusts, fares adjust, competitors’ costs move identically so none of it is a share gain, and the crack spread does not hold. Capitalizing a spot price as though it were a perpetuity is the single most common error in commodity-levered equity analysis, and the 23-times gap between the two columns is the price of making it. A reader who builds the mechanical case and then waits for the market to agree is waiting for something that has not happened in 67 months of data.

Crude, jet fuel, and the half of the bill nobody prices

Everything above measures the stock against jet fuel, which is what American actually buys. That is not the price the market quotes. The market quotes crude, and jet fuel is crude plus a refining margin — the crack — that moves on its own schedule. Over this window the crack has ranged from $6.61 a barrel in 2021-06 to $70.64 in 2023-01. It is not a constant, and treating it as one is the assumption that makes a crude view and a fuel-cost view look like the same view when they are not.

The distinction is not academic at today's prices. At the July 2026 monthly averages, WTI was $80.46 a barrel while jet fuel was $3.403 a gallon, or $142.93 a barrel. The crack between them was $62.47, against an average of $27.48 across 2021 through 2025. Crude is 56 percent of what American pays for a barrel of jet fuel today, against 75 percent on the 2021 to 2025 average. Nearly half the fuel bill is now refining margin, and a crude decline that leaves the crack where it is delivers less than half of what the barrel math implies.

That cuts both ways, and the second direction is the more interesting one. A crack that simply returned to its own 2021 to 2025 average, with crude held exactly where it is, would take jet fuel from $3.403 to $2.570 a gallon — a saving of $0.833 a gallon, or roughly $4,015 million a year pre-tax at the June 2026 quarter's consumption. That is 1.7 times the largest row in the sensitivity table above, and it requires no view on the price of oil whatsoever. There are two independent routes to a lower fuel bill here, and on this evidence the refining margin is the more stretched of the two.

So which price does the stock respond to?

A holder's instinct is that the market watches the headline rather than the input, and that instinct is testable. Jet fuel is crude plus crack; if the stock tracks American's true cost it should respond to both components, and if it tracks sentiment it should respond to crude and ignore the crack. The crack varied by a factor of more than ten across this window, so there is no shortage of movement to detect. The three correlations, on the same 66 months of log returns, are below.

Institute calculation. WTI is the EIA spot price FOB at Cushing, Oklahoma, monthly average of daily prices, retrieved from the FRED mirror MCOILWTICO on September 4, 2026; that series was last updated September 2, 2026 and publishes through August 2026, but the window is truncated at July 2026 so every crude month has a matching jet-fuel month and a matching stock close. None of the first three rows is statistically significant at conventional thresholds, and they are printed with their t-statistics for that reason. The conditional rows rest on very few months — the WTI bucket has 9 — and are descriptive, not inferential.

Two readings survive this table and a third does not. The one that does not survive is any claim that the relationship is statistically established: across 66 months neither crude nor jet fuel clears a conventional significance threshold against the stock, and the difference between them does not either. The first reading that does survive is that crude has the larger coefficient of the two, -0.164 against -0.098, and separates further when the move is large — -0.584 on months crude moved more than ten percent against -0.382 for jet fuel. The second is the crack row, and it is the sharpest number in the section: +0.004. Not a weak relationship — a flat one. Roughly half of what American pays for fuel has moved violently for five years and the equity has not registered it at all.

That is what a sentiment channel looks like as distinct from a cash-flow channel. If the market were pricing American's fuel bill, the crack would matter as much per dollar as the crude does, because a dollar is a dollar in the income statement. It does not. The market appears to price the number it can see quoted. The Institute states this as a supported reading and not a finding, for two reasons: it rests on correlations that are individually insignificant, and it runs the other way in 2026 taken alone, where jet fuel (-0.737) explains the stock better than crude (-0.554) and the crude result is fragile besides — dropping any single month moves it between -0.258 and -0.756, against a jet-fuel range of -0.617 to -0.822. Both things are in the data and both are printed.

A holder should take the risk in that asymmetry seriously rather than only the opportunity. If the crack normalizes and crude does not, American collects roughly $4,015 million a year of pre-tax relief through the component the market has never once paid for. The cash would be real and the re-rating, on five years of evidence, might not come.

What the lever does reach, and what it does not

One correction belongs here, because the obvious way to write this paragraph is wrong. Fuel is an operating cost, so it moves EBITDA — and EBITDA is the denominator of both multiples this case has used to value American. A fuel move therefore runs straight through the leverage and valuation ratios rather than leaving them alone. It is worth seeing how far.

$ in millions. Institute calculation holding enterprise value and net obligations constant at the September 1, 2026 measurement and adding the annualized pre-tax fuel saving to trailing EBITDA of $3,090 million. This is a ratio illustration and not a forecast: it assumes the saving is permanent, which the preceding subsection has already shown is the wrong assumption for a spot price. It is printed to locate which ratio the lever moves, not to value the equity.

The lever closes one gap almost exactly and barely touches the other. American trades at 12.20 times EBITDA against Delta's 8.72 times, and the whole of that discount is worth about 26 cents a gallon; 54 cents would reach United's. On leverage the same move accomplishes almost nothing: fifty cents a gallon takes net obligations from 9.42 times EBITDA to 5.29 times, still 2.3 times United's 2.30 times. The multiple gap is an EBITDA problem and fuel is an EBITDA lever. The leverage gap is a debt problem and fuel is not a debt lever.

That is the qualification that connects this section to the verdict. What the price of a barrel does not move is the numerator: the principal balance, the contractual interest burden, and the $17.5 billion of committed aircraft spending through 2030 are fixed in dollars whatever fuel costs. The 3.4-times net obligation against the market capitalization is a stack of contracts, not a margin. A fuel decline makes the condition the verdict rests on easier to meet. It does not meet it.

The premise tested: relative price position

The case was commissioned on the observation that the stock is trading near its 52-week low. That is true in the loose sense and misleading in the precise one, and the distinction matters because it determines whether the question is “why is this airline cheap” or “why are airlines cheap.”

Source: quoted market data at the close of September 1, 2026, 4:00 PM EDT. Trailing twelve-month revenue through the June 2026 quarter. The measurement date on every figure in this table is the same day; ranges are the trailing 52 weeks ended that day.

American is 28.3 percent above its 52-week low. United is 23.6 percent above its own. On that measure American is not the one closest to its floor — United is. What separates them is not proximity to a low but the multiple of revenue the market will pay: 0.15 times sales for American against 0.54 times for United and 0.73 times for Delta. American's revenue is 85 percent of Delta's and its market capitalization is 17 percent of Delta's. That gap — five-sixths of the revenue for one-sixth of the equity value — is the fact the rest of this case explains.

So the premise is restated before it is tested: American is not distinctively near a low. It is distinctively cheap against its own revenue, and the whole of that cheapness lives on the right-hand side of the balance sheet.

The June 2026 quarter: record revenue, eleven cents of earnings

The second quarter of 2026 is the cleanest available test of the thesis, because it holds the operating environment constant across all three network carriers and lets the capital structure do the differentiating.

Sources: 2026 column from American Airlines Group Inc., Form 8-K Exhibit 99.1 furnished July 23, 2026; 2025 column from the same company’s Form 8-K Exhibit 99.1 furnished July 24, 2025 — both for the three months ended June 30 of their respective years. Adjusted earnings per share excludes $28 million of after-tax net special items in 2026 and $29 million in 2025. Percentage changes on the lines the 2026 release does not present comparatively are computed by the Institute from the two filed columns; the operating margin line is stated in percentage points rather than as a percentage change. $ in millions except per-share.

Read the top and the bottom of that table together. Revenue set a company record. Fuel expense rose $2,218 million on a 77.1 percent increase in price per gallon. Operating income fell 60.7 percent to $446 million, and net interest expense of $409 million consumed 92 percent of it. What survived to the equity was $71 million against $599 million a year earlier, a fall of 88.1 percent on revenue that rose 16.3 percent.

The prior-year column carries a detail worth pausing on, because it cuts in American’s favor and it is easy to miss. Net interest expense did not rise. It fell, from $433 million to $409 million, a decline of 5.5 percent — the deleveraging described below doing exactly what it is supposed to do. The interest burden still consumed almost the whole of operating income, but not because the burden grew. It consumed it because operating income collapsed underneath a fixed charge that was actually getting smaller. That is the difference between a company with a debt problem and a company with an earnings problem carried on a debt-heavy balance sheet. The revenue side, meanwhile, was strong across every line: passenger revenue rose 15.9 percent, cargo 29.4 percent and other revenue 18.0 percent, while salaries, wages and benefits — the largest controllable cost — grew only 5.9 percent. Operating margin still fell 5.2 points, from 7.9 percent to 2.7 percent. Fuel did that, and nothing else in the table did.

That last sentence is the capital structure speaking. On a $16.7 billion revenue quarter, the interest bill alone is nearly the whole of operating income. There is no operating improvement of ordinary size that changes that arithmetic; it takes either a much larger margin or a much smaller balance sheet.

Commercial performance in the quarter

It is important not to let the balance sheet conclusion contaminate the read on the business. The demand and revenue evidence in the quarter is strong and it is broad.

Source: American Airlines Group Inc., Form 8-K Exhibit 99.1 furnished July 23, 2026. Cabin and entity figures are passenger revenue per available seat mile. Managed corporate revenue growth of 26 percent was the fifth consecutive quarter of double-digit growth.

Premium cabin unit revenue outgrew Main Cabin by 4.6 points. Every geographic entity grew unit revenue. Managed corporate revenue rose 26 percent, the fifth consecutive quarter of double-digit growth, which is the single most meaningful number in the release because corporate share is the metric American lost during its 2023 distribution strategy and has been rebuilding since. Unit costs excluding fuel rose under three percent against 5.4 percent capacity growth. None of this is the profile of a company with a demand problem or a cost-control problem.

The problem is that a 2.7 percent operating margin, on this balance sheet, is not enough. Consider the same quarter at United.

Sources: American — Form 8-K Ex. 99.1, July 23, 2026, and Form 10-Q for the quarter ended June 30, 2026. United and Delta — respective second-quarter 2026 earnings releases. Delta reports adjusted net debt and adjusted operating cash flow; those are the company's own non-GAAP definitions and are not identical to American's presentation, which is why they are labelled rather than compared directly. American's figure is debt and finance leases less unrestricted cash and short-term investments; Delta does not reconcile its adjusted net debt to that basis in the cited release, so the two are printed side by side and are not differenced. “n/d” means the figure is not carried in the Institute’s model on a basis comparable to the rows around it; where a carrier discloses the item on its own definition, that disclosure is described in this note rather than placed in the cell. Market capitalizations are struck at the September 1, 2026 close. $ in millions.

American ran 95 percent of United's revenue and produced 8.8 percent of United's net income. Both carriers grew revenue about 16 percent. Both absorbed the same fuel curve. Neither release discloses enough to attribute the gap in net income line by line. What the balance sheets do show is the difference in what each carrier owes, and section 5 sets that out.

Enterprise value on a claims-inclusive basis

Before any multiple can be quoted, the enterprise has to be measured on a basis that includes every claim ahead of the common. For an airline that means three things the headline “net debt” figure typically omits: operating lease liabilities, which finance aircraft the airline flies and does not own; the net pension and postretirement obligation, which is a funded claim on future cash; and restricted cash, which is not available to offset debt and is therefore excluded from the cash credit.

Sources: share count and balance sheet from American Airlines Group Inc., Form 10-Q for the quarter ended June 30, 2026; price at the September 1, 2026 close. Restricted cash and restricted short-term investments of $709 million are deliberately excluded from the cash offset because they are not available to retire debt. The pension line is the balance sheet carrying amount at June 30, 2026; the underlying actuarial funded status is measured at December 31, 2025. The four lines below the multiple are shown for completeness and do not feed the enterprise value used anywhere else in this case; see the discussion immediately below.

Two things follow immediately. First, at 22.7 percent of enterprise value the common stock is a thin residual claim, which means every point of change in the value of the enterprise moves the equity roughly four and a half points. That cuts both ways and is the honest case for owning it as well as the honest case against.

Second, a valuation multiple quoted on market capitalization tells you almost nothing here. American trades at 0.15 times trailing revenue on its equity. On enterprise value it trades at 0.65 times the same revenue — a 4.4-fold difference. Anyone reasoning from the equity multiple alone is measuring the wrong thing.

The operating lease line deserves its own explanation, because the objection to it is a fair one. ASC 842 does not put a liability on the balance sheet by itself. It puts a liability there and a matched right-of-use asset against it, and in American's case the asset is the larger of the two: $6,919 million of right-of-use asset at June 30, 2026 against $6,802 million of lease liability, a net position of $117 million in the company's favor. Adding the liability to enterprise value while ignoring the asset can therefore look like counting one side of a matched pair.

The Institute adds the liability and does not net the asset, deliberately and for the same reason enterprise value does not net owned aircraft in property and equipment against the debt that financed them. The right-of-use asset is an operating asset. It is the aircraft American flies, and it is already generating the revenue and the EBITDA that sit in the denominator of every multiple quoted off enterprise value. Subtracting it from the numerator while its earnings remain in the denominator would credit the same aircraft twice. Enterprise value asks what it would cost to acquire the operating business free of its financing claims; it does not ask what the business would fetch in liquidation. The peer figures below are computed on the same convention, which is the other reason not to depart from it here.

The arithmetic the other way is shown in the four lines below the multiple above so the reader can weigh it directly: netting the right-of-use asset would put enterprise value at $30,772 million, net obligations at $22,200 million, 2.6 times market capitalization rather than 3.4 times, and the common at 27.9 percent of the enterprise rather than 22.7 percent. It is a meaningfully gentler picture. It is not, on this reading, the right one — but the case for it is real enough that the reader should not have to take the Institute's word for the convention without seeing what the alternative produces. The alternative basis is shown for American only. Delta and United do not disclose right-of-use assets in the second-quarter releases this case works from, so the peer comparison cannot be restated on the netted basis, and it is not.

The debt architecture and the 2028 concentration

American's borrowings are not a single number; they are roughly two dozen instruments with different collateral, different covenants and very different maturities. The tranche detail below is the December 31, 2025 schedule from the Form 10-K, which is the most recent instrument-level disclosure. Balances have moved since — total debt and finance leases were $28,927 million at June 30, 2026 against $29,007 million at December 31, 2025 — but the architecture is unchanged and it is the architecture that matters.

Source: American Airlines Group Inc., Form 10-K for the year ended December 31, 2025, Note 4. Ordered by maturity, nearest first, rather than in the order the 10-K presents them; instruments shown as a year range sort on their first year, and repaid instruments carry no maturity and sort last at a zero balance. Principal outstanding, $ in millions. “var” denotes a floating rate; the rate shown is the effective rate at December 31, 2025. Enhanced equipment trust certificates carry a weighted average rate of 3.95 percent; equipment loans and other notes 5.57 percent. Em dashes indicate the prior-year balance is not separately presented.

Source: American Airlines Group Inc., Form 10-K for the year ended December 31, 2025, Note 4. Ordered by maturity, nearest first. The PSP notes are the CARES Act Payroll Support Program promissory notes. The 6.50 percent convertible senior notes were repaid during 2025 using proceeds of the 2025 AAdvantage Term Loan. $ in millions.

Two clocks run on those two tables and only one of them is printed in the header. The schedule is the most complete issue-by-issue disclosure American files, and it is filed once a year, at December 31. Three of its lines have since gone to zero. In the first quarter of 2026 American prepaid in full the $1,000 million of 8.50 percent senior secured notes due May 2029 — the most expensive fixed-rate secured paper on the schedule — and paid in full the $629 million senior short-term term loan due January 2026; the $583 million of 5.50 percent AAdvantage notes reached maturity in April. That is $2,212 million of principal extinguished in six months, and it is the reason the 8.500 percent notes do not appear in the bond quote table below. A reader working only from the December 31, 2025 schedule would carry all three forward. The June 30, 2026 balance-sheet total, which does reflect them, is the figure used everywhere else in this case.

Three observations about the composition. Roughly 85 percent of the principal is secured, which means the collateral — aircraft, slots, gates, routes, and above all the AAdvantage program — has already been pledged. That constrains the options available in a stress scenario, because the assets a distressed airline would normally monetize are already encumbered. Second, the rates on the secured paper American issued during 2026 run from 5.46 to 6.04 percent fixed, against enhanced equipment trust certificates averaging 3.95 percent; refinancing older aircraft debt at current spreads raises the interest bill even when it lowers the principal. Third, the 10.75 percent notes retired during 2025 tell you what American was paying at the bottom of its credit cycle, and the coupons the company is issuing at today prices as non-investment-grade paper. The Institute does not print an agency rating here: a rating is a measurement made by a third party on its own action date, and no such action is carried in the filings this case works from.

The net debt reduction and its composition

The deleveraging is the part of the American story that management leads with, and it is real. It is also more interesting than the headline. Below is net debt by year on American's own published definition — debt and finance leases, plus operating lease liabilities, plus pension obligations, less unrestricted cash and short-term investments — held constant across every date so the years are comparable.

$ in millions. The table opens at December 31, 2021, the gross-debt peak — the most American has ever owed on this definition, $50,371 million. Read the opening row as the top of the gross arc and nothing more: net debt peaked a year earlier, at $40,412 million on December 31, 2020, because the 2021 balance sheet carried $12,431 million of cash against the larger gross balance. The 2019 and 2020 year-ends are discussed below but are no longer printed as rows. Definition, verbatim from American's reconciliation: “Total debt is defined as debt, finance and operating lease liabilities and pension obligations, and net debt is defined as total debt, net of unrestricted cash and short-term investments.” Rows without a dagger are the company's own published figures, from the Q1 2023, Q1 2024, Q4 2024 and Q4 2025 earnings materials and the July 2026 GAAP to non-GAAP reconciliation. † denotes an Institute reconstruction on the identical definition, built from the filed balance sheet for that year plus the pension-only funded status from that year's Form 10-K pension note; American did not publish the table for 2021 or 2023. The reconstruction method was validated at December 31, 2022, where it reproduces the company's published 2,153 pension line and 36,875 net debt exactly. The pension line is pension-only funded status and excludes the separately disclosed retiree medical obligation, which was 1,141 at December 31, 2025.

Two dates matter and they are not the same date. Gross total debt peaked in 2021 — $50,371 million at that December 31, and higher still at mid-year, which is the “peak levels in mid-2021” management refers to. On the debt-and-lease measure, which can be computed at interim dates because it does not need the pension note, American carried $48,297 million at June 30, 2021 against $46,655 million three months later and $35,729 million today — a fall of $12,568 million. But net debt peaked a year and a half earlier, at $40,412 million on December 31, 2020, because the borrowing American did in 2021 arrived alongside the cash it raised. At June 30, 2021 the company was sitting on $17,950 million of cash and short-term investments. Gross debt and net debt were telling opposite stories in the same quarter.

That is why the table opens where it does. Measured from the gross peak, the total obligation fell from $50,371 million to $35,875 million, a reduction of $14,496 million or 28.8 percent, and net debt fell from $37,940 million to $28,105 million, $9,835 million or 25.9 percent. Every figure in the decomposition below runs from the first row of the table above, so the reader can check it against a row that is actually printed. Here is what did the work.

$ in millions. A figure in parentheses reduces net debt; a figure without them increases it. Institute calculation from the table above. The cash line carries no parentheses because cash and short-term investments fell from 12,431 to 7,770 over this window, and spending down a cash balance raises net debt rather than lowering it. The pension line is a remeasurement of a funded status, not a payment: it fell from 4,219 to 146 principally because discount rates rose, and the plans were frozen effective November 1, 2012.

Two of these four lines are not what the deleveraging headline implies. Debt and finance-lease principal came down $9,133 million, which is real, was funded, and is the single largest contributor. But 41.4 percent of the fall in net debt is the pension line collapsing from $4,219 million to $146 million, and that is not cash out the door. A frozen defined-benefit plan's funded status is the present value of a fixed stream of promises discounted at market rates; when rates rise the number shrinks without anyone paying anything, and it swells back if rates fall. Running the same decomposition from the net-debt peak at December 31, 2020 rather than from the gross peak makes the point more sharply still — on that start date net debt fell $12,307 million and the pension line alone accounts for 49.6 percent of it.

The fourth line is the one a holder should sit with. American did not only repay debt over these five years; it also spent $4,661 million of the cash it raised during Covid, and that drawdown worked against the net-debt figure by the same amount. The gross reduction of $14,496 million is therefore the better measure of what was actually retired, and the net reduction of $9,835 million is the smaller number precisely because the liquidity buffer is thinner than it was at the peak.

The comparison that frames the equity case, though, is against 2019 rather than against either peak, and it runs off the table above because the pre-Covid year-end is no longer printed as a row. The figures are the company's own and are stated here in full so nothing is hidden by the shorter table: at December 31, 2019 American carried $24,315 million of debt and finance leases, $9,129 million of operating leases and $5,461 million of pension obligations against $3,826 million of cash, for $35,079 million of net debt. American's net debt on its own definition is now $6,974 million below where it stood at the end of 2019, before any of this happened — $28,105 million against $35,079 million. Both of those endpoints are the company's own published figures, so there is nothing to argue about. What is worth arguing about is the composition. Against 2019, borrowings and finance leases are $4,612 million higher, not lower. The improvement comes from $5,315 million off the pension line, $2,327 million of operating leases run off, and $3,944 million more cash on the balance sheet. An airline that borrowed more than it had before Covid and shows less net debt than before Covid is telling you where to look.

The maturity ladder and the 2028 concentration

Source: American Airlines Group Inc., Form 10-K for the year ended December 31, 2025, Note 4. Amounts are contractual principal maturities as scheduled at that date and do not reflect refinancings completed during 2026, including the second-quarter transactions that the company states addressed its only meaningful 2027 maturity.

Calendar 2028 carries $7.3 billion — 26 percent of all scheduled principal, and more than either adjacent year on its own — 64 percent above 2027 and 81 percent above 2029. Three instruments account for most of it: the 2013 Term Loan Facility maturing February 2028, the 7.25 percent senior secured notes that are interest-only until February 2028, and the 2021 AAdvantage Term Loan, whose March 2025 amendment cut quarterly amortization to roughly $6 million and pushed the balance into a balloon at the April 20, 2028 maturity.

That wall has a price, and the number should be built on what American is actually paying for new money rather than on the coupons it paid in 2023. Every secured borrowing disclosed in the first half of 2026 is in the table below. The range is 5.46 to 6.04 percent fixed, not the 7.25 to 8.50 percent an earlier reading of this balance sheet would have used — and the top of that older range, the 8.50 percent notes, was prepaid in full and no longer exists.

Source: American Airlines Group Inc., Forms 10-Q for the quarters ended March 31, 2026 and June 30, 2026. The 2026 Engine EETCs refinanced the $228 million of 2019-1 Engine EETCs at maturity and raised $339 million of incremental borrowing. The floating rate shown is the average as of June 30, 2026. $ in millions.

Applied to the 2028 wall, that changes the answer materially, and it changes it in American’s favor. The three instruments the filings name — the 2013 Term Loan Facility, the 7.25 percent notes and the 2021 AAdvantage Term Loan, $3,984 million between them — already carry a weighted 6.31 percent. Refinancing those at 2026 rates does not cost money; it saves roughly $11 million to $34 million a year. The headwind is confined to the remaining $3,340 million, predominantly aircraft paper carrying the 3.95 percent weighted average coupon on the equipment trust certificates, which reprices upward by roughly $50 million to $70 million. Net, the 2028 wall refinanced at the rates American is paying today adds on the order of $17 million to $59 million of annual interest, against trailing operating income of $1,008 million.

The Institute is revising its own number here and says so plainly. An earlier construction of this illustration applied a 425 to 550 basis point spread across the entire $7.3 billion, which implicitly treated all of it as three percent paper and produced a figure near $300 million to $400 million. More than half the wall is not three percent paper; it is six and seven percent paper, and the correct spread on that half is negative. The revision cuts in the equity’s favor, which is the direction that warrants the most scrutiny given the position disclosed at the end of this case, so the components are printed above rather than netted. Two caveats hold it honest: the two term loans are floating-rate, so for them the refinancing question is credit margin rather than base rate, and the $3,340 million balance is not disclosed instrument by instrument, so the equipment trust certificate average stands in for it. It remains an arithmetic illustration on disclosed rates, not a forecast of where American will actually clear.

That amendment is worth pausing on, because it is the clearest example in the filings of a pattern that runs through this balance sheet. Reducing near-term amortization to a quarter of a percent per quarter is genuinely useful — it frees roughly $220 million a year of cash relative to the prior schedule. It does not reduce the obligation. It relocates it to a single date, and it does so on the same asset that secures the rest of the loyalty financing. American has bought itself time repeatedly and has generally bought it well; the ladder shows where the bill for that time comes due.

Revolving capacity and the liquidity covenant

Source: American Airlines Group Inc., Form 10-K for the year ended December 31, 2025. Revolving commitments were upsized from approximately $2.9 billion to $3.0 billion on April 21, 2025. “Other facilities” includes a $350 million revolving facility maturing March 2027 with a one-year extension option and $47 million of available cargo-receivables borrowing base capacity expiring December 2026. No borrowings were outstanding under any facility at December 31, 2025. $ in millions.

Undrawn revolvers are the healthiest fact on this page and they are also the reason a specific covenant deserves attention. A significant portion of American's debt agreements requires the company to maintain at least $2.0 billion of unrestricted cash and cash equivalents plus undrawn revolver capacity. With $3,397 million of undrawn capacity alone, that test is not close to binding today. But note its construction: it counts revolver availability, and revolver availability is itself subject to collateral coverage tests. In a scenario where aircraft appraisals fall, borrowing base capacity falls with them, and the liquidity covenant tightens at exactly the moment cash is scarce. That is the standard secured-credit doom loop and it is worth naming even in a period when it is nowhere near triggering.

The other covenant class is loan-to-value and collateral coverage, tested against annual or semiannual appraisals across the secured notes, term loans, revolvers and spare engine trusts. A breach does not accelerate; it obliges American to post additional collateral, permit cash to be withheld, pay the debt down, or accept a rate step-up. The AAdvantage Financing adds a peak debt service coverage ratio whose breach can force early repayment in whole or in part. American reported compliance with all of these at the most recent applicable measurement dates.

The bond market's pricing of the debt

The equity is near a 52-week low. The credit is not. American's traded paper is quoted at or around par on every issue underwritten in the current rate environment, which is the single most useful outside read on this balance sheet available to a public investor — and it does not agree with the equity.

Prices are aggregator quotes retrieved September 2, 2026 from the TradingView public bond table, which is a market-data aggregator and not a filed source, expressed as a percent of par. Two-Clock caution: the source publishes an update time for the equity quote only and does not publish a quote date for each individual bond, so these prices carry an unknown measurement lag. The AAdvantage 5.750% of 2029 was independently cross-checked on the Frankfurt listing venue at 99.46 on the same date — 8 basis points from the aggregator quote, mutually corroborating — though Frankfurt turnover that day was zero on a single price fixing, making it an indicative listing quote rather than a trade print; the real market for this 144A/Reg S bond is US over-the-counter. The 8.500% senior secured notes due May 15, 2029 (ISIN USU02413AJ82, WKN A3LRG2) appear in neither source, and that is not a sourcing gap: American prepaid the full $1,000 million in the first quarter of 2026, so the issue no longer trades. It is excluded here rather than shown at a stale price. The two equipment trust certificate lines are unweighted averages of the individual quotes.

Two things follow. The first is that the market is not pricing distress in the debt. The 7.250% senior secured notes of 2028 change hands above par at a 5.18 percent yield to worst, and the $2.75 billion AAdvantage Loyalty IP 5.750% of 2029 — the instrument secured on the asset section 6 argues is the company's best — sits within half a point of par, inside a 52-week band of 98.12 to 101.41 — a band published by the same Frankfurt listing whose turnover on the quote date was zero, and which therefore carries the same indicative-quote caveat as the price it brackets. A market genuinely worried about recovery does not quote a 2029 secured claim at 99. The equity and the credit are looking at the same company and reaching different conclusions, which is exactly what a thin equity sliver beneath a large secured stack should produce: the credit market is pricing the debt as money-good, and on that pricing the residual behaves like an option on the enterprise rather than a claim on its assets.

The second is that the discounted paper is discounted for the wrong reason to be interesting. Nine enhanced equipment trust certificates carrying coupons below 4.5 percent average 94.35, with the 2.875% of 2034 down at 88.31. Six carrying 4.5 percent or better average 97.70 — including issues maturing in 2038 and 2039, longer-dated than any of the discounted ones. Across the fifteen, price correlates with coupon at +0.69 and with remaining tenor at only +0.12. That is a rate story, not a credit story: the low-coupon certificates were written when money was cheap and are marked down to today's yield curve, not marked down because anyone doubts the aircraft. It also quantifies the refinancing problem this section opened with, and it sizes it correctly: American retires three and four percent paper and replaces it at 5.46 to 6.04 percent — a spread of roughly 150 to 200 basis points on the aircraft paper, not the 400-plus the 2023 coupons would imply — and the bond market has already repriced the old certificates to reflect it. The discounts in the table are the market marking that spread, and they are the right order of magnitude for it.

The collateral package, the covenants, and the 2029 Notes amortization

The single most consequential financing structure on this balance sheet is the 2021 AAdvantage Financing, and it deserves its own section because it is simultaneously American's most valuable asset and its most heavily encumbered one.

On March 24, 2021, American and AAdvantage Loyalty IP Ltd. — a Cayman special purpose vehicle — issued $3.5 billion of 5.50 percent senior secured notes due 2026 and $3.0 billion of 5.75 percent senior secured notes due 2029, alongside a $3.5 billion term loan facility. The collateral package is comprehensive: a first-priority security interest in the AAdvantage Agreements themselves and all payments under them, the intellectual property licenses, the deposit and reserve accounts, the equity of the Loyalty Issuer and the SPV guarantors, and American's rights to AAdvantage data and intellectual property.

Read that list again. The co-brand agreements, the payments under them, the data, and the brand are all pledged. Approximately $11.0 billion of financing sits against a program that carries $11.6 billion of deferred revenue on the June 30, 2026 balance sheet. Any transaction that would monetize AAdvantage — a spin, a stake sale, a joint venture — has to be negotiated through that structure first.

Source: American Airlines Group Inc., Form 10-K for the year ended December 31, 2025, Note 4(c). Amortization on the 2029 Notes is subject to early-amortization events including failure to meet a minimum debt service coverage ratio. Mandatory prepayment triggers include indebtedness issued by the SPV, pre-paid AAdvantage mileage sales exceeding $505 million, a change of control of AAG, and collateral sales above a threshold — each carrying an applicable premium.

The last row is the one to carry forward. Amortization on the 2029 Notes began on July 20, 2026 — inside the quarter American had just reported — at $250 million a quarter. That is $1.0 billion a year of non-discretionary principal repayment, secured on the loyalty program, running through April 2029. It is not a maturity that can be refinanced away in a single transaction; it is a standing draw on cash that begins precisely as the seasonal float turns.

The mandatory prepayment triggers deserve a line as well, because one of them constrains a tactic airlines habitually reach for under stress. Pre-selling miles to a co-brand partner — the maneuver that funded several carriers through 2020 — is a mandatory prepayment event here above $505 million. American can still do it. It just cannot do it at scale and keep the proceeds.

Obligations not classified as debt

Two claims sit ahead of the common that a debt-only view omits, and for an airline both are large.

Operating leases

Source: American Airlines Group Inc., Form 10-K for the year ended December 31, 2025, lease footnote. Balance sheet amounts are measured at December 31, 2025. Lease cost includes $248 million relating to Republic Airways capacity purchase agreements; American holds a 20.8 percent equity interest in Republic Airways Holdings. Operating lease liabilities of $6,802 million at June 30, 2026 are used in the enterprise value bridge above. $ in millions.

Total net lease cost of $5.3 billion in 2025 is the number to hold onto. Note its composition: variable lease cost of $3.4 billion is nearly twice the fixed operating lease cost, and variable lease cost is dominated by airport rents, landing fees and capacity purchase payments that scale with activity. That portion is a genuine operating expense rather than disguised financing, which is why this case capitalizes only the recorded operating lease liability of $6,802 million in the enterprise value rather than applying a multiple to total rent. The conservative treatment is deliberate: capitalizing variable lease cost would inflate enterprise value and make American look artificially expensive against peers whose disclosure splits differently.

Pension and postretirement

Source: American Airlines Group Inc., Form 10-K for the year ended December 31, 2025, retirement benefits footnote. The measurement date for all actuarial amounts is December 31, 2025; the balance sheet carrying amount used in the enterprise value bridge is $1,160 million at June 30, 2026. Plans are substantially frozen. The IAM National Pension Fund was certified in “endangered” status on March 29, 2019 notwithstanding a funded ratio above 80 percent. $ in millions.

The pension has improved materially and this is the one liability line trending the right way: the funded deficit narrowed from $1095 million to $539 million during 2025, helped by an actual return on plan assets of $1,229 million. The plans are substantially frozen, so the obligation does not grow with service. Minimum required contributions for 2026 are $238 million — real, but not a threat to liquidity. Combined underfunding across pension and retiree medical of $1.7 billion is material to enterprise value and immaterial to the near-term cash question. Both statements should be carried forward.

The composition of $11.3 billion of reported liquidity

The company ended the June 2026 quarter with $11.3 billion of total liquidity, and the chief financial officer said so on the call. That figure is accurate and it is also the most commonly over-read number in the release, so it is worth decomposing.

Sources: American Airlines Group Inc., Form 10-Q for the quarter ended June 30, 2026, for balance sheet amounts; Form 10-K for the year ended December 31, 2025, for revolver capacity, which is the most recent facility-level disclosure. The company stated total liquidity of $11.3 billion at June 30, 2026 on the July 23, 2026 earnings call; the reconciliation above is the Institute's and differs slightly because revolver capacity is carried at the December 31, 2025 disclosed amount. $ in millions.

The liquidity is real and it comfortably clears the $2.0 billion covenant floor. But look at the three lines beneath the total. Sitting against $7,770 million of unrestricted cash and investments is $9,551 million of air traffic liability — tickets sold for travel not yet provided — and $11,573 million of loyalty program liability. American's cash balance is smaller than the customer obligations it is holding against. That is normal for an airline and it is not a solvency observation. It is the reason the next section exists: when a business is funded by customer prepayments, the cash flow statement and the earnings statement can point in opposite directions for a year at a time, and only one of them is telling you about the business.

Customer float in operating cash flow

This is the analytical core of the case.

In the six months ended June 30, 2026, American generated $4,694 million of cash from operations, up from $3,419 million in the prior-year period, while reporting a net loss of $311 million. A reader who stops there concludes that the accounting loss is a depreciation artifact and the business is throwing off cash. Decompose it instead.

Source: American Airlines Group Inc., Form 10-Q for the quarter ended June 30, 2026, condensed consolidated statements of cash flows and balance sheets. The air traffic liability build is the change from $7,158 million at December 31, 2025 to $9,551 million at June 30, 2026. The loyalty build is the corresponding change in the current and noncurrent loyalty program liability. “Working capital and other, net” is derived as operating cash flow less net loss less depreciation and amortization and therefore includes items beyond the two float components identified. $ in millions.

Roughly 72 percent of American's first-half operating cash flow is money customers have handed over for flights they have not taken and miles they have not redeemed. Air traffic liability grew from $7.2 billion at year-end to $9.6 billion at June 30 — a $2.4 billion build — because passengers buy summer travel in the spring. That is not profit. It is a liability that gets discharged by flying the airplanes, and the seasonal portion of that build gives back over the second half every year.

The seasonality is not a hypothesis; it is visible in the full-year numbers. Across all of 2025 the air traffic liability built only $399 million and the loyalty liability $510 million — together roughly 29 percent of full-year operating cash flow, against 72 percent in the first half of 2026. The half-year figure is a seasonal peak, not a run rate.

What the float is, and what it is not

A necessary correction to how this finding is usually stated, including in much of the sell-side commentary on airline cash flow. The float is not an anomaly, a distortion, or a borrowing. It is a permanent working capital account. Customers buy travel before they take it, so the balance exists for the same reason accounts payable exist, and it belongs to the business cycle rather than to any particular year. Airlines are not unusual in this. Deferred revenue is the ordinary condition of subscription software, insurance premium finance, publishing, prepaid services, gift cards and season tickets. In every one of them the balance persists as long as the enterprise does. It does not unwind, it does not have to be repaid in cash, and it does not go to zero. There is exactly one circumstance in which it goes to zero, and that is a wind-down.

So the analytical claim has to be stated on the flow rather than the stock, and the distinction is the whole of it. The balance is permanent; what enters the cash flow statement is the change in the balance. A float that simply persists contributes nothing to operating cash flow in any period. Only a float that grows contributes, and it contributes exactly its growth. American's first-half operating cash flow is 72 percent composed of a liability that grew — part seasonally, because summer travel is bought in the spring, and part structurally, because a carrier adding capacity carries more forward bookings each year than the last. Neither component is fraudulent and neither is repayable. Both are, however, a different thing from earnings: they do not compound, they cannot service interest twice, and they stop the moment the airline stops growing. The case against American is not that the float will reverse. It is that a business generating cash from growth in a customer liability while earning nothing has no second source of cash if growth pauses.

The wind-down case is not hypothetical

Sourcing note, stated here rather than at the back: the account that follows rests on contemporaneous national press reporting, not on any Spirit filing, and it is set out because the sequence bears directly on the float argument above.

The wind-down exception acquired a live example during the period this case examines. Spirit Airlines ceased all operations on May 2, 2026, after a second trip through Chapter 11 in eighteen months and after failing to reach terms with bondholders on emergency financing, and entered liquidation. The carrier attributed the failure to the same jet fuel shock that runs through every table on this page — the price spike that followed the 2026 Gulf conflict. Roughly seventeen thousand jobs went with it.

Two things follow, and they point in opposite directions for American.

The first is the warning. Spirit's float went to zero, and it went to zero in the only way it ever can — the airline stopped flying, and the customers who had prepaid became unsecured creditors of an estate. That is what a permanent liability looks like on the day the permanence ends. It is also a reminder of what the fuel line does to a levered balance sheet. The shock that took 61 percent out of American's operating income in a single quarter did not damage the weakest carrier in the industry; it ended it. American is not Spirit — it has $11.3 billion of liquidity, a loyalty program pledged against $11.6 billion of deferred revenue, and a functioning capital markets bid. But the distance between the two is measured in exactly the quantity this case is about, which is the capacity to absorb a fuel year without needing the float to keep growing.

The second is the tailwind, and it is real. Spirit's exit removed an entire ultra-low-cost network from domestic markets in the middle of American's second quarter. American's domestic unit revenue rose 10.6 percent year over year in that quarter and system unit revenue 10.3 percent, against capacity growth of only 5.4 percent. The Institute does not claim to have measured how much of that pricing came from Spirit's disappearance rather than from industry-wide capacity discipline or from fuel being passed through, and no such attribution should be inferred. What can be said is directional and it matters: a competitor whose entire proposition was undercutting legacy fares is gone, the seats it flew are gone with it, and the fare environment American is pricing into for the back half of 2026 is structurally less contested than the one it entered the year with. That is a genuine offset to the fuel line, and it is the strongest thing in the revenue picture.

It is also, on the evidence in front of the Institute, not enough. American grew unit revenue 10.3 percent into a less competitive market and still watched operating income fall 61 percent, because fuel rose 83.3 percent and there was no hedge behind it. A carrier that cannot convert the disappearance of its cheapest competitor into operating income during the quarter that competitor disappeared is telling the reader where its problem actually sits, and it is not on the revenue line.

Strip the float and what is left of the first half is roughly $1,292 million of operating cash flow, against $1,633 million of capital expenditure and aircraft purchase deposits already spent, and against depreciation of $1,136 million. On that basis the business roughly covered its own depreciation and did not cover its capital program. That is the honest read of the first half.

Free cash flow and the second-half requirement

Sources: American Airlines Group Inc., Form 10-K for the year ended December 31, 2025, and Form 10-Q for the quarter ended June 30, 2026. Free cash flow is defined here as cash provided by operating activities less capital expenditures and aircraft purchase deposits, before sale-leaseback proceeds. The 2023 and 2024 comparatives use the same definition. $ in millions.

Full-year 2025 free cash flow on this definition was negative $680 million: $3,099 million of operating cash flow against $3,779 million of capital spending. That was in a year in which American earned $111 million of net income and posted $1.5 billion of operating income — a better year than 2026 is guided to be.

Now the second half of 2026. The chief financial officer told investors on July 23 that at the midpoint of the current guidance American expects to produce positive free cash flow for the year and to finish with lower net debt than it started. Take that claim seriously and test it.

Sources: reported amounts from American Airlines Group Inc., Form 10-Q for the quarter ended June 30, 2026; guidance from Form 8-K Exhibit 99.1 furnished July 23, 2026, with the fuel assumption built on the forward curve as of July 21, 2026; capital expenditure guidance from the same-day earnings call. The July 23, 2026 release is the only primary-source guidance the Institute has identified for 2026; reports of a subsequent August revision could not be traced to a company filing or release and are not relied upon here. $ in millions.

The arithmetic of the claim holds together, and it is worth being precise about why, because the conclusion is neither that the CFO is wrong nor that the free cash flow is good news. Positive full-year free cash flow at a zero-earnings midpoint is arithmetically available because depreciation of roughly $2.3 billion exceeds the cash cost of a $4.0 billion capital program only modestly, and because the year-over-year float position does not have to unwind to zero — it only has to unwind to a level consistent with a growing airline. A carrier adding capacity carries more forward bookings at each year-end than the last. Growth funds itself, on the way up. That is why the ex-float line above is labelled a floor rather than an estimate: stripping the entire first-half build assumes none of it is structural, which section 9 argues is too harsh. The Institute does not print a central case, because splitting the build between the seasonal part that reverses and the structural part that does not would require a disclosure American does not make. The floor is the number that can be defended from the filings, and the conclusion below is reached on the floor.

But the composition matters more than the sign. Free cash flow produced at zero net income is depreciation and float. It is not earnings, it does not compound, and it does not reduce the obligation in any durable way. Net debt on a debt-only basis did fall in the first half — from $23,171 million at December 31, 2025 to $21,157 million at June 30, 2026, a $2,014 million improvement. Note what funded it: the cash balance rose while debt was roughly flat, and the cash rose on the float. The deleveraging is real on the measurement date and it is seasonally financed.

Against that sits $17.5 billion of planned aircraft and engine expenditure across calendar 2026 through 2030, based on commitments as of December 31, 2025. Averaged, that is roughly $3.5 billion a year of committed capital against a business whose best recent operating cash flow year produced $4.0 billion. The fleet program is not optional in the way discretionary capex is; deliveries are contracted, and the financing for them is what keeps total debt from falling.

The tax carryforwards omitted from enterprise value

Enterprise value counts the debt and nets off the cash. It does not count the tax attributes, and in American's case that omission is large enough to matter to the equity. At December 31, 2025 the company held approximately $11.9 billion of gross federal net operating losses and a further $6.0 billion of other federal carryforwards — $17.9 billion in total against future federal taxable income — of which $16.3 billion carries forward indefinitely and only $1.6 billion begins to expire, in 2033. On top of that sit roughly $5.0 billion of state net operating losses expiring between 2025 and 2045.

Source: American Airlines Group Inc., Form 10-K for the year ended December 31, 2025, Note 6 and Management's Discussion and Analysis. All amounts measured at December 31, 2025 — a nine-month-old measurement date in a September 2026 document; first-half 2026 taxable income will have consumed part of the carryforward and the balance is not disclosed at the interim date. Per-share and percent-of-market-cap columns use 661.9 million shares outstanding at June 30, 2026 and the $12.95 close of September 1, 2026, a market capitalization of $8,572 million.

Read the valuation allowance line, because it is the whole argument in one number. A company that did not expect to use its losses would carry a valuation allowance against them. American carries $22 million against $8,692 million of gross deferred tax assets — essentially none. Management states that it has “determined that positive factors outweigh negative factors in the determination of the realizability of our deferred tax assets,” and the auditor treated the realizability of those carryforwards as a critical audit matter, which means it was examined rather than assumed. The company believes it will earn its way through them, and its auditor tested that belief.

The value of the carryforwards, and its limits

The practical consequence is that American's book tax provision is not a cash cost. The 10-K says so directly: the $79 million provision recorded in 2025 at an effective rate of 41.2 percent was “substantially non-cash,” and the same language covers the $308 million recorded in 2024 at 26.7 percent. An investor modelling American on after-tax earnings is applying a tax rate to cash flow that mostly is not being paid. On a free cash flow basis the shield is already in the numbers; on an earnings basis it is not, and that gap is worth understanding before comparing American's price-to-earnings multiple to anyone else's.

The carryforward deferred tax asset of $4,095 million is 48 percent of a market capitalization of $8,572 million, or $6.19 a share against a $12.95 stock. The net deferred tax asset after netting the deferred tax liabilities — principally accelerated depreciation on the fleet — is $2,359 million, or $3.56 a share. Those are the two honest bookends, and the difference between them is real: the deferred tax liabilities reverse too, and a going-concern airline that keeps buying aircraft keeps regenerating them.

Four qualifications, none of them small. First, a loss carryforward is worth nothing without taxable income to absorb it, and American's 2025 pre-tax income was roughly $190 million. At that rate the carryforward is not a near-term asset; it is an option on the earnings recovery the equity case already depends on, which means it is correlated with the thesis rather than diversifying it. Second, the attributes are worth less than face: $17.9 billion of carryforwards shields tax at the statutory rate, not dollar for dollar, which is why the recorded asset is $4,095 million rather than $17,900 million. Third, the balance is falling: the same line read $4,292 million a year earlier, so the company is consuming roughly $197 million of it a year at current profitability. Fourth, and most important for anyone thinking about the shareholder register, the attributes are exposed to section 382.

Section 382 of the Internal Revenue Code limits the annual use of net operating losses after an “ownership change” — broadly, a shift of more than fifty percentage points among five-percent shareholders over a rolling three-year window. American treats this as a live risk and says so in its own risk factors: an ownership change “may severely limit or effectively eliminate our ability to utilize our NOL carryforwards and other tax attributes.” The company's certificate of incorporation carries transfer restrictions designed to prevent one. For a $8,572 million equity carrying a $4,095 million carryforward asset, that is not boilerplate. It means a large accumulation by a new holder — an activist, a strategic, a distressed buyer — can destroy a meaningful fraction of the value it is trying to capture, and it is one reason the equity is harder to concentrate than its size suggests.

Valuation on a consistent basis

With the enterprise measured properly, the comparison to Delta and United can be made honestly — but only if the same measurement is applied to all three. Each carrier publishes its leverage on its own definition: Delta reports adjusted net debt, United reports total debt including finance leases and other financial liabilities, and American reports net debt as debt and finance leases less unrestricted cash and short-term investments. None of the three is the lease- and pension-inclusive construction used in section 4. Rather than quote three incompatible figures side by side, the Institute rebuilds all three from the June 30, 2026 balance sheets on one convention: market capitalization, plus debt and finance lease obligations, plus all operating lease liabilities, plus the net pension and postretirement obligation, less unrestricted cash and short-term investments.

That convention is not an Institute preference imposed on the peer group. Applied to United, it returns net obligations of $18,105 million — which is, to the dollar, United’s own published adjusted net debt at the same date. A construction that reproduces a peer’s own disclosed figure exactly is one the reader can trust when it is turned on the other two. That is one carrier on one date, and it is a check on the construction rather than a proof of it. Delta does not publish a reconciliation that would allow the same test, so the Delta column should be read as the Institute’s construction and not as a figure Delta has confirmed.

Sources: equity prices at the September 1, 2026 close; balance sheet claims at June 30, 2026 from each carrier’s second-quarter Form 8-K Exhibit 99.1 and Form 10-Q. Two clocks, both stated. Net obligations are debt and finance lease obligations, plus all operating lease liabilities, plus the net pension and postretirement obligation, less unrestricted cash and short-term investments; Delta’s $1,749 million of sale-leaseback financing liabilities are included in the debt leg because Delta itself counts them as debt. EBITDA is trailing twelve-month operating income plus depreciation and amortization on GAAP components, computed identically for all three; United’s own adjusted EBITDA on the same twelve months is $7,399 million, the difference being United’s non-operating and special-item adjustments. Free cash flow uses each carrier’s own published measure where one exists — Delta and United publish theirs; American publishes none, so its figure is the Institute’s, operating cash flow less capital expenditure and aircraft purchase deposits. The three definitions are not identical and the difference is disclosed rather than smoothed, which is why the free cash flow line is read as three separate measures rather than as a ranking.

American’s forward multiple of 11.4 times is the highest of the three, not the lowest — which is the correct result and the one most likely to surprise. A company earning close to zero has a large forward multiple on a small forward number; that is arithmetic, not a valuation signal. The equity multiple that looks cheap is price to sales, and price to sales is precisely the metric that ignores the capital structure. Both observations point the same way: on the equity alone, American is not measurably cheap. On the enterprise, it is levered.

The comparison on an enterprise basis

On the common convention, American carries $29.1 billion of net obligations against $8.6 billion of equity value — 3.4 times. Delta carries $20.1 billion against $50.0 billion of equity, 0.40 times. United carries $18.1 billion against $34.0 billion, 0.53 times. American’s net obligation is 1.4 times Delta’s on an equity base 5.8 times smaller — a leverage ratio to equity roughly 8.4 times Delta’s and 6.4 times United’s. Delta’s weighted average interest rate is 4.9 percent with 78 percent fixed; American issued secured paper during 2026 at 5.46 to 6.04 percent. More obligation, less equity beneath it, at a higher cost of debt. The cost-of-debt leg of that has narrowed, though: the gap is now roughly 56 to 114 basis points rather than the 250 to 350 the 2023 coupons implied — and it is not a clean comparison in either direction, because Delta’s figure is a weighted average across debt already outstanding while American’s is the rate on new money. The obligation and the equity base are the durable part of the discount, and that part is not a mispricing.

The cash line has to be read the same way. In the six months ended June 30, 2026 Delta generated $1,436 million of free cash flow and United $3,225 million, each on its own published definition. American publishes no free cash flow measure, so its $3,061 million is the Institute’s construction — operating cash flow less capital expenditure and aircraft purchase deposits. Three definitions, three constructions: the figures are printed so the reader can see each carrier’s own disclosure, not ranked against one another. And section 9 is the reason American’s figure has to be read with its composition attached rather than as a headline.

What the multiple implies

Trailing twelve-month operating income through June 30, 2026 is $1,008 million — full-year 2025’s $1,467 million, plus the first half of 2026’s $405 million, less the first half of 2025’s $864 million. Against an enterprise value of $37,691 million, that is 37.4 times trailing operating income. Add back depreciation and amortization of $2,082 million over the same twelve months and the multiple falls to 12.2 times EBITDA — before any lease-cost adjustment, and on a trailing number depressed by the fuel shock.

That is where the peer comparison bites hardest. On identical construction, Delta trades at 8.7 times and United at 6.6 times. American’s 12.2 times is the most expensive of the three on the measure that includes the debt — the exact reverse of the picture on price to sales. The reason is on the leverage line beneath it: American’s net obligations are 9.4 times EBITDA, against 2.5 times at Delta and 2.3 times at United. American’s own stated ambition is net debt to EBITDA around three turns. On the trailing twelve months it is at roughly three times that target, and the gap is closed by EBITDA recovering, not by debt being repaid.

That is not an expensive multiple in absolute terms, and it is the strongest quantitative argument available for the long case: at a normalized operating income anywhere near the $3.0 billion American earned in 2023, the same enterprise value would be about 12 times operating income, and the equity — being a thin residual — would not have to move a little. It would have to move a lot. The arithmetic of a 23-percent equity sliver is symmetric.

Symmetric is not a rhetorical flourish, so the Institute prints the arithmetic that makes it one. Hold the net obligations constant, normalize EBITDA to $5,116 million, and mark the enterprise at each peer’s multiple. The multiple implied by today’s price is 7.37 times.

Institute computation on the June 30, 2026 net obligations of $29,119 million, held constant, and 661.9 million shares outstanding at the same date. Peer multiples are those in the comparative table above, at the September 1, 2026 close. This is an arithmetic sensitivity on a normalized earnings assumption, not a forecast, and it assumes the normalization is reached without adding debt — the condition set out below.

Marking the enterprise at Delta’s multiple takes the common to $23.38, up 81 percent; marking it at United’s takes it to $7.13, down 45 percent. Same normalized earnings, same balance sheet, a swing of more than a hundred and twenty points on the multiple alone. That is what a thin residual does in both directions. Any reader who takes only the upside line has taken half the table.

The condition attached to that argument is the whole of the risk. It requires the enterprise value to rise while the debt does not, which requires operating income to recover while the fleet program is funded from cash rather than from new borrowing. Every year in the recent record shows the opposite pattern: capital expenditure exceeded operating cash flow in 2025, debt issuance of $3.8 billion offset repayments of $5.5 billion, and total debt and finance leases ended 2025 at $29.0 billion — roughly where it stands today.

The sum-of-the-parts case, and what the loyalty program is worth to a shareholder

The most-cited bull case for American is that the market is not paying for AAdvantage. The argument runs: loyalty programs at the large network carriers have been valued in the tens of billions; American’s is arguably the largest by membership; therefore the $8.6 billion equity is pricing the airline attached to a very valuable program at less than nothing.

The Institute takes that argument seriously and will not print it as a finding, because the numbers underneath it will not carry the weight. Here is what can and cannot be said.

What is filed and dateable: the loyalty program liability is $11,573 million at June 30, 2026 — $4,353 million current and $7,220 million noncurrent. That is a liability, not an asset: it is deferred revenue measuring miles sold and not yet redeemed. Approximately $11.0 billion of debt is secured on the program. Citi co-brand card spend grew 8 percent and new member enrollments grew 30 percent in the June 2026 quarter. Those are the facts.

What is not filed: any valuation of AAdvantage as a standalone business. Third-party estimates circulating in August 2026 put competitor programs at up to $30 billion, drawing on the pricing of the Air Canada Aeroplan transaction. Those are secondary-source peer estimates carrying their own measurement dates, they are not marks on AAdvantage, and American has not disclosed a program-level profit figure that would let a reader independently value it.

More importantly, the encumbrance changes the character of the asset. A loyalty program whose agreements, payment streams, data, intellectual property and issuing entity are all pledged as first-priority collateral for $11.0 billion of debt — with mandatory prepayment triggered by a change of control, by collateral sales above a threshold, and by pre-paid mileage sales above $505 million — is not an asset that can be separately monetized on a shareholder’s timetable. Whatever AAdvantage is worth, the first $11.0 billion of it is spoken for, and the structure was deliberately built so that it could not be extracted without the lenders’ participation.

The honest formulation is narrower than the bull case and still favorable: AAdvantage is a genuinely valuable, genuinely growing asset that American has already borrowed against heavily and cannot readily sell. It supports the credit. It does not, on the current structure, support a sum-of-the-parts argument for the equity.

What a buyer of this equity is buying

Position disclosure, repeated here. The author established a long position in AAL on September 3, 2026, after this analysis was completed. The author holds no position in DAL or UAL. This is educational analysis published by the Institute and is not investment advice.

The conclusion below turns on one condition — whether the fleet program can be funded without adding debt — and the case has assembled every input that condition requires. The Institute therefore runs it rather than stating it.

Sources: American Airlines Group Inc., Form 10-K for the year ended December 31, 2025, cash flow statement, Note 4 maturity schedule and risk factors. Two clocks: the cash flow is a 2025 measurement, the commitments and the ladder are December 31, 2025 measurements. Averaging is the Institute’s, and the annual pattern is not level — 2028 alone carries $7.3 billion of principal against the $4,390 million average. The mandatory AAdvantage amortization of roughly $1.0 billion a year sits inside the scheduled principal line and is not added again. $ in millions.

Read the two subtotals. The first sets the weaker of the two tests: hold total debt flat by refinancing every maturity as it comes, and ask only whether operating cash flow covers the aircraft program. On the 2025 result it does not — it falls short by roughly $401 million a year, before any allowance for the higher coupon at which the refinancing would be done. The second sets the test the verdict actually requires: fund the fleet and retire the principal rather than roll it, which needs roughly $7,890 million a year against the $3,099 million produced in 2025 — a gap of about $4.8 billion a year.

That is the arithmetic behind every qualification that follows. It does not say the company cannot refinance; the bond quotes in section 5 say plainly that it can. It says that on the most recent full-year cash generation, the fleet program is funded by the capital markets rather than by the business, and that the condition the bull case depends on is not met at the current level of operating cash flow. It is a level test on one year, not a forecast, and a recovery in operating income of the size discussed in section 12 would change the answer.

The case was built to follow the evidence wherever it led, including to the conclusion that the discount is deserved. It largely is — with two qualifications that matter.

The premise required correction. American is 28 percent above its 52-week low, not at it, and United is closer to its own floor. What distinguishes American is not proximity to a low but the price the market puts on its revenue: 0.15 times sales against 0.73 times for Delta. That distinction moves the question from timing to structure, which is where it belongs.

The business is not the problem. Record quarterly revenue, a fifth consecutive quarter of double-digit managed corporate revenue growth, premium unit revenue up 13.4 percent, every geographic entity positive, and unit costs excluding fuel contained under three percent against 5.4 percent capacity growth. A reader looking for evidence that American cannot sell a seat or control a cost will not find it in this quarter.

The capital structure is the problem, and it is not a cyclical one. Net obligations of $29.1 billion at June 30, 2026 sit against $8.6 billion of equity at the September 1, 2026 close. Interest expense of $409 million consumed 92 percent of operating income in the best revenue quarter in company history. Calendar 2028 carries $7.3 billion of scheduled principal, the 2029 AAdvantage Notes began drawing $1.0 billion a year of mandatory amortization on July 20, 2026, and $17.5 billion of committed aircraft spending runs through 2030 on commitments as of December 31, 2025. Stockholders’ equity is negative $4.0 billion.

And the cash flow is float. This is the finding the Institute would most want a reader to carry away, because it is the one most easily missed: $4.7 billion of first-half operating cash flow against a net loss is not evidence of hidden earnings power. Approximately 72 percent of it is customer money for flights not yet flown and miles not yet redeemed. It reverses seasonally. Full-year 2025 free cash flow was negative $680 million in a year materially better than 2026 is guided to be, and the guided full-year adjusted earnings range straddles zero.

So: is the equity cheap? On enterprise value against trailing operating income the answer is arguably yes — roughly 13.0 times a rough trailing EBITDA depressed by a fuel shock, on an airline with a working commercial franchise. And because the equity is only 23 percent of the enterprise, a recovery in operating income to anything resembling 2023 levels would move the equity by a multiple of the move in the enterprise.

That is a real option and it should be described as one rather than dressed as a valuation. What it requires is not a better quarter but a sustained period in which operating income recovers while the debt does not grow — which means funding the fleet program from cash. The record does not yet show that happening: capital expenditure exceeded operating cash flow in 2025, and total debt and finance leases have been roughly flat at $29.0 to $28.9 billion across the period. Until the fleet is self-funded, deleveraging is seasonal rather than structural, and the discount is the market pricing that distinction correctly.

The Institute’s conclusion, stated plainly: the discount to Delta and United is deserved on the evidence available at the September 1, 2026 close. The equity is nonetheless a high-variance instrument in both directions — the sensitivity in section 12 puts the range at -45 percent to +81 percent on the same normalized earnings, depending only on the multiple applied — and a reader who buys it should understand precisely what they are buying: not a cheap airline, but a 23-percent residual claim on a $37.7 billion enterprise, whose value turns on whether the fleet program can be funded without adding debt. That is a coherent thing to analyse as an option. It is not a value investment, and calling it one obscures the only question that decides it.

Sources and measurement dates

Every quantitative claim in this case is traceable to one of the documents below, and every figure runs on the measurement date shown rather than on the publication date of this case. Where a filed figure has been superseded by a more recent measurement, both clocks are shown.

All Form 10-K, Form 10-Q and Form 8-K references are filings of American Airlines Group Inc. with the Securities and Exchange Commission. The first row is the labelled exception: quoted market data is not a filed source, and the peer trailing revenue, net income and per-share legs rest on it. Every other figure in this case ties to a filed document.

Company statements, peer releases, secondary sources and Institute derivations

Reports circulating in August 2026 of a further revision to American’s 2026 guidance could not be traced to any company filing, press release or furnished exhibit; the July 23, 2026 Form 8-K remains the only primary-source guidance identified as of the publication date of this case, and no August revision is asserted anywhere on this page.