American Express Company AXP

Key data

The business

American Express runs a closed loop. It issues the card, it signs the merchant, and it owns the network in between, so it collects a fee on nearly every dollar its cardholders spend, what it calls billed business, without splitting that fee with an outside network the way a Visa-branded bank does. On top of the spending fee it charges an annual card fee, and those fees, growing 15%, behave like a subscription: paid up front, renewed out of habit, lifted every time the company refreshes a Platinum or Gold card. The customers are affluent, which is the whole game. They spend more, they travel more, and they pay their bills. Amex also lends, and card balances grew 8% last quarter, in step with spending, with the interest on those balances the third leg of the model. The brand, the closed loop, and a wealthy base that will not switch are the moat.

The numbers

Revenue here is reported net of interest expense, the company's own convention, so it is not a gross figure. The last five quarters:

The most recent quarter grew revenue about 9% and profit about 8% against a year earlier, and earnings per share 11%, the gap between the two the work of buybacks. Earnings have come in ahead of what analysts modeled in three of the last four quarters. The dip in Q4 2025 earnings is seasonal, the fourth quarter carrying the year's heaviest marketing and rewards spend.

Earnings per share compounded about 11% a year from 2021 to 2025, faster than net income's roughly 8%, because the share count kept falling on $5.8 billion of stock bought back in 2025 alone. Analysts model close to $23 in earnings by 2028, mid-teens growth from here, and it is against that number that today's stock costs only about 14 times. The engine underneath is worth naming plainly:

What this memo believes that the market discounts is that the net card fee line is a durable annuity, not a cyclical loan book, and the print that settles it is double-digit fee growth holding through a spending slowdown.

Management

Management returns cash steadily and honestly: $5.8 billion of stock repurchased in 2025 and $6.0 billion in 2024, shrinking the count while return on equity sits around 34%. The guidance record is good, with earnings ahead of estimates in three of the last four quarters. On the other side of the ledger, insiders sold about $41.8 million across fifteen sales in the last year and bought nothing, the largest from Anna Marrs and Denise Pickett in early February. Plan status is not disclosed, so read them as officers trimming after results rather than a signal either way.

How it fails or surprises you

Credit normalizes. The affluent base is the defense, but balances grew 8% and the loans are unsecured. If the US consumer softens and net write-offs climb from the low twos toward 3% or more, reserve builds hit earnings directly. The print that shows it first is the quarterly net write-off rate set against balance growth.

Spending stalls. The closed loop earns on billed business, so a genuine recession cuts discount revenue the moment cards go quiet, with no lending offset. Watch billed-business growth split into travel and entertainment versus everyday goods, and the international line, for the first crack.

The fee annuity re-rates (right tail). Net card fees growing 15%, product refreshes, and younger cardholders signing on could prove the recurring fee stream is worth a network multiple, not a lender's. If the market accepts that, the multiple expands. The print is double-digit fee growth plus rising cards-in-force through a soft patch.

Closing thoughts

This is a business the market understands well, so the edge is thin and worth naming who is on the other side. The sellers who took the stock down about 16% from its high, even as second-quarter earnings came in ahead and grew 11% per share, are betting a consumer cycle the trailing numbers do not yet show is coming, and that reserve builds on those 8%-growing balances will interrupt the run. That is exactly the prior watch-item, a very good lender priced for something more, and it has partly resolved: the multiple has eased from the top of its range toward 20 times while earnings kept climbing, so the gap between price and proof is narrower than it was a month ago. My read is that the affluent base and the fee annuity make these earnings more durable than a cyclical lender multiple implies, but that view is modest, because everyone can see the same 15% fee growth. The fatter tail is a real spending recession that would cut discount revenue and balances at once.

The bet is still that wealthy Americans keep charging their Amex cards and keep paying the annual fee to hold them. It breaks if that base stops paying its bills: watch the net write-off rate against the 8% growth in card balances, and if losses climb while balances keep rising, the durable-annuity story is wrong and this reprices as the plain lender the sellers already take it for.