The unit is one proprietary card. In FY2025 the average card generated about $849 of revenue and $162 of pretax income (inferred from FY2025 10-K).
Per $100 of card spending: AmEx keeps $2.24 from the merchant, earns $0.60 in card fees, $1.04 in net interest and $0.45 in other fees, for $4.33 of revenue. Rewards, benefits and partner payments take $1.85; marketing, credit losses and operating costs take a further $1.65. What remains is $0.83 pretax.
| Industry | Card payments and card lending; US bank holding company, Category II since Q2 2026 |
|---|---|
| How it makes money | Discount revenue 52% of FY2025 revenue, net interest income 24%, card fees 14%, other fees 10% |
| What protects it | Closed loop, US anti-steering rules upheld in 2018, near-universal US acceptance, premium brand |
| Earnings drivers | Premium card fees; spend from younger and international members; loan growth |
| Watch | Benefit costs vs revenue; merchant take rate; three-party regulation |
| Cycle exposure | Medium: volume and T&E are cyclical, fees and variable costs cushion |
AmEx is issuer, network and merchant acquirer at once, so it is paid on both sides of the same transaction. When a card member spends $100, it pays the merchant $100 less its discount and keeps about $2.24; it then bills the card member, who either pays in full or revolves at a net interest yield of 8.1%. Annual fees are recognised over twelve months, and balances are funded mostly by $152.5 billion of deposits (FY2025 10-K).
At end-FY2025 there were 86.6 million proprietary cards, each basic card spending $25,453 a year on average, plus 66.2 million cards issued by partner banks on the network (FY2025 10-K). The table on spend economics carries a key point: rewards consume about half of discount revenue and all variable customer engagement costs about 83% of it (inferred). The merchant fee mostly funds the card member proposition; card fees and interest are what make the franchise profitable.
| Segment, FY2025 | Revenue $m | Pretax $m | Billed business |
|---|---|---|---|
| US Consumer Services | 34,814 | 6,810 | $707.5bn |
| Commercial Services | 16,926 | 3,668 | $541.9bn |
| International Card Services | 13,000 | 1,603 | $418.0bn |
| Global Merchant & Network Services | 7,759 | 3,968 | n/a |
What AmEx exits tells you what it values. Costco and JetBlue went in 2016, cutting US cards-in-force 18% (FY2016 10-K); the Amazon and Lowe's small business cobrands are being sold now, costing about a point of revenue growth from Q4 2026 with negligible pretax effect (Q2 2026 call). Network licensing in the EU and Australia was abandoned because interchange caps made it uneconomic (FY2025 10-K). The direction is toward proprietary, fee-paying relationships AmEx owns outright.
The richest link in the card chain is the network toll: Visa earned a 59.9% operating margin on $40.0 billion of FY2025 net revenue and Mastercard 57.6% on $32.8 billion, with almost no credit risk (company filings, web). Bank issuers carry rewards and losses instead; JPMorgan's consumer bank earned a 32% ROE, Citi's US Personal Banking 13.2% RoTCE, and Capital One's card book charged off 4.93% in Q4 2025 (web). AmEx's 33.9% ROE on a 2.0% write-off rate reflects collecting network-like merchant fees and issuer economics on a spend-centric, low-loss customer base.
The US market is a network oligopoly: in 2024 Visa carried 61.1% of card purchase volume, Mastercard 25.8%, AmEx 11.1% and Discover 2.0% (Nilson Report, web). The binding barrier is two-sided scale, and Capital One's $35.3 billion purchase of Discover shows its price: a year after closing it has moved debit onto Discover but says migrating credit cards depends on building international acceptance (Capital One Q2 2026 call, web). AmEx reports acceptance at over 170 million locations.
Three assets are hard to copy together. Card members spend about three times as much per card as on other networks, a company claim that the 2013 court record in Ohio v. American Express supports at about 3.1 times (derived, web). The Supreme Court upheld AmEx's anti-steering rules 5–4 in 2018, so US merchants cannot push its card members toward cheaper cards. And its commercial franchise is, in the CEO's words, "still 3x larger than anybody else" (Q4 2025 call).
The merchant premium is real but narrowing. Australian data show AmEx fees around 1.3% against 0.9% for Visa/Mastercard credit (RBA 2022, web), but US posted interchange on premium Visa/Mastercard cards is already about 2.75% (Kansas City Fed, web), and AmEx's own take rate slid from 2.29% of billed business in 2023 to 2.23% in Q2 2026 (10-K; 10-Q). In premium cards, rivals compete on benefits: Chase Sapphire Reserve went to $795 and Citi Strata Elite launched at $595 within weeks of AmEx's Platinum moving to $895 in 2025 (press releases, web).
"We have the Card Member, we have the network, and we have the merchant."CEO Stephen Squeri on AmEx's position in agentic commerce, Q1 2026 earnings call
| AmEx claim | Outside evidence | Verdict |
|---|---|---|
| ~3x spend per card | 2013 court record implies ~3.1x; no current independent figure | Supports (dated) |
| Merchant premium justified | Upheld in 2018; contested by merchants, RBA and CJEU | Partially supports |
| Closed-loop data advantage | Capital One paid $35.3bn for the same model; unquantified | Partially supports |
| US acceptance parity | Company-sourced only | Unverified |
The threats target the same advantage. The Visa/Mastercard merchant settlement, preliminarily approved in June 2026, would let merchants decline premium cards as a tier and surcharge more; the EU Court of Justice ruled in April 2026 that AmEx's co-brand payments can count as interchange; and the RBA is reviewing three-party networks (web; Q2 2026 10-Q). Winners in premium payments combine affluent spend scale, both sides of the transaction, the capacity to out-fund benefits, and freedom from price caps. AmEx has all four today, and the fourth is the most exposed.
Revenue compounded about 8.2% a year from FY2016 to FY2025, to $72.2 billion. Billed business grew 7.6% a year, but discount revenue only 5.9% because the take rate drifted down, while net card fees compounded 14.8% and net interest income 13.0% (inferred from 10-Ks). Growth has come from charging card members more and lending them more, not from charging merchants more. Acquisitions are immaterial ($633 million of cash in 2025), so reported growth is effectively organic.
Net card fees +18% in FY2025 and +15% in Q2 2026. The Platinum refresh added $200 to the fee; a quarter of US consumer Platinum members had paid it by Q1 2026 with "no change to our very high retention rates" (Q1 2026 call). 75% of new accounts are now fee-based.
Millennials and Gen Z were 65% of new US consumer accounts in Q2 2026, with Gen Z spend +38%. ICS billed business grew 14% in FY2025 from about 6% share in priority markets (calls; FY2025 10-K; Bernstein conference).
NII +12% in FY2025 on 9% loan growth and an 8.1% yield, helped by deposit costs falling from 3.7% to 3.3%. Both sides move with rates.
CS billed business +3% in FY2025 and +5% in Q2 2026, with CS cards −5% and small business write-offs up to 2.6%.
Down 5–6 basis points since 2023 as acceptance extends to smaller merchants and abroad.
| Period | Reported | FX-adjusted | Inorganic items |
|---|---|---|---|
| FY2025 | +10% | +9% | Center acquisition; immaterial |
| Q2 2026 | +10% | +10% | Swisscard consolidated from Jan 2026 (size undisclosed) |
| From Q4 2026 | ~−1pt | — | Small business cobrand portfolio sales |
Pretax income was 19.1% of FY2025 revenue (inferred). Rewards, card member services and partner payments were 42.8% of revenue and are guided to 44–45% in 2026 because refreshed Platinum benefits are being used heavily; card member services grew 50% in Q2 2026. These costs flex with spend, protecting margins in a downturn but capping them in good times. The $16.5 billion Membership Rewards liability is the key estimate: +25 basis points of redemption adds about $229 million of expense (FY2025 10-K).
Marketing ($6.3 billion, up ~75% since 2019) is the discretionary lever management uses to absorb upside, and ~$16 billion of salaries and other costs are the semi-fixed base (22% of revenue). For a bank, cash conversion is capital conversion: a 10–11% CET1 target means each dollar of balance growth absorbs about ten cents of equity, and a ~34% ROE leaves most earnings free to return.
| $ millions | FY2016 | FY2019 | FY2022 | FY2025 |
|---|---|---|---|---|
| Revenue net of interest expense | 35,438 | 43,556 | 52,862 | 72,229 |
| Net card fees | 2,886 | 4,042 | 6,070 | 9,993 |
| Net interest income | 5,779 | 8,620 | 9,895 | 17,364 |
| Pretax income | 8,042 | 8,429 | 9,585 | 13,795 |
| Diluted EPS ($) | 5.61 | 7.99 | 9.85 | 15.38 |
| Return on equity | 25.8% | 29.6% | 32.3% | 33.9% |
From FY2016 to FY2025 cash went, in order, to buybacks ($42.9 billion, shares −24%), dividends ($15.7 billion, $1.22 to $3.28 per share), capex ($16.2 billion) and small acquisitions (~$3.3 billion net). Returns equalled ~84% of cumulative net income and 86% in H1 2026. Management funds balance growth first, holds CET1 near the bottom of its range, and returns the rest. Since the FY2025 figures it has agreed to sell its ~30% GBTG stake for a pretax gain of roughly $945–975 million, bought out Swisscard, and proposed acquiring TheFork on undisclosed terms (Q2 2026 10-Q; Bernstein conference).
In 2020 billed business fell 19%, T&E 61% and airline 76%; revenue fell 17% and net income 54%, but ROE held at 14.2% because card fees still grew 15% and rewards and service costs fell 23% and 45% (FY2020 10-K). In 2009 billed business fell 9% and lending write-offs hit 8.5%, below the Fed's 10.54% industry card peak (AmEx 2009 Annual Report; FRED, web). Today the lending book is 1.7 times its 2019 size and net interest income is 24% of revenue, up from 16% in 2016, so a credit-led downturn would bite harder than 2020's.
AmEx is near its own cycle highs: 9% FX-adjusted spend growth in Q1 and Q2 2026 is a three-year high, write-offs (2.0%) and delinquencies (1.2%) sit below 2019, and Q2 2026 ROE of 36.4% exceeds the 2024 peak with reserves released to 2.7% of balances. These are peak-cycle returns on normalised credit. The downside mechanism is a recession cutting spend and forcing a CECL reserve build just as merchants gain tools to decline premium cards.
| Monitor | Why | Where published |
|---|---|---|
| Discount revenue ÷ billed business | Merchant pricing power | 10-K/10-Q revenue tables |
| VCE costs as % of revenue | Fees vs benefit arms race | Quarterly calls; 10-Q |
| Fee per card; fee-based share of new accounts | Retention at $895 | 10-Q Table 1; presentation |
| Write-offs and 30+ day delinquency | Credit turn on a larger book | 10-Q credit tables |
| CET1 and buyback pace | Capital under Category II | 10-Q capital section |
| V/MC settlement, CJEU follow-on, RBA review | Three-party pricing regulation | Court dockets; ACM; RBA |
Not answered by the sources: AmEx's independently measured US discount rate; payments to Delta; current spend per card by network and 2025 US share; Swisscard's revenue contribution and TheFork's price; Platinum retention in numbers; acquisition prices for Kabbage, Resy, Tock and Center; pre-2012 data in the folder.