Amex raised the toll on its own loop by 29% and almost nobody left. The question is who keeps the money that goes around it.
American Express is the only payments company that is issuer, network and acquirer at once — every dollar its members spend circulates back through its own closed loop. In Q2 2026 that loop ran hotter than it has in three years: spending +9.4% FX-adjusted, card fees +15.4%, revenue guidance raised. And yet the stock is 16% below its December high, because management spent the entire beat on the perks required to keep the loop spinning. Six analyst lenses, three scenarios, four horizons.
Gray line = AXP’s actual closes at dated anchors from Jan 2025 into today: the $387.49 all-time high on Dec 12, 2025, the $290.97 52-week low in March 2026, the $361.57 July rebound, and $324.69 now. Colored paths = synthesized scenario midpoints forward, probability-weighted at bear 30% · base 45% · bull 25% (bear-tilted deliberately — see the assumptions). The clay ring at the 1-year mark is the $375.96 sell-side consensus. History panel and forecast panel share one continuous time axis but different month-per-pixel density; the forecast side is linear in price at mid-year marks.
Re-weight the scenarios
Our 30 / 45 / 25 split is a judgment call, tilted bearish because the multiple — not the earnings path — is what is actually in question. Make it yours: drag to set how likely the bear and bull cases are (base takes the remainder). The blended target below, the dotted line on the chart, and the prob-weighted row of every scenario card all update live.
Visa and Mastercard only run the rails; Amex owns every link. The fee buys the perks, the perks buy the engagement, the engagement buys the spend, and the spend — at a discount rate Amex sets itself, because it is also the acquirer — funds the next round of perks and justifies the next fee increase. The bull and bear cases are the same loop read in opposite directions. The bull says the loop is gaining energy: fees compound faster than costs. The bear points at the clay leak on the right — the $200 fee increase went out the door to airlines, hotels and 20-plus brand partners as credits, so the member pays more, the partner collects it, and the shareholder is handed the growth story instead of the earnings.
Six analyst lenses, $275 to $409
Each lens below was run as an independent pass on the same dated ground truth, deliberately without sight of the others. They agree on the facts and disagree by $134 a share — which is the most honest picture of this stock available.
The Subscription in Disguise
A subscription business wearing a credit card’s clothes. Net card fees +15.4% with 30-plus straight quarters of double-digit growth and guidance to exit 2026 in the high teens; 12.5M new cards in 2025 with over 70% fee-paying; Gen Z and Millennials ~65% of new consumer accounts at an average new-Platinum age of 33. Billed business +9.4% FX-adjusted is the fastest in three years. July’s selloff was a reinvestment decision misread as a miss.
The Cash Counter
At 17.4x forward you pay less than the market paid in 2024 (21.1x) or 2025 (24.0x) for a 36% ROE franchise. $15.05B trailing FCF, $7.47B of buybacks plus $2.42B of dividends, a fresh $16B authorization, share count down 2.49%. Honest caveat: for a lender, FCF flatters — and at 6.4x book there is no asset floor underneath you.
The Beat Was Already Spent
Expenses grew +12% against revenue +10%, and revenue actually missed. Net income rose +8% while EPS rose +11% — the gap is buybacks, not earnings. Provisions fell 23% to $1.1B, a tailwind that cannot repeat, while the tax rate normalized to 23.6% from 18.7%. The $895 renewal cliff has seen one partial cycle. Fifteen of thirty analysts are Hold-or-worse, and the CEO sold into February’s dip without buying back.
Stable, Not Widening
The closed loop still lets Amex charge what nobody else can — $695 to $895, +29%, retention near 100%. But the moat’s rent is rising: ~$3,500 of perks and 20-plus partnerships mean the fee increase largely funds the coupon book. And 170M+ merchant locations closed the acceptance gap, which also removed acceptance as a differentiator — leaving status as the only defensible layer, against JPMorgan’s $795 Sapphire Reserve.
Dead Money With an Upward Bias
A 2025 multiple bubble deflating onto its own long-run mean. Price sits under both the 50-DMA ($339.96) and 200-DMA ($336.67), which are 1% apart and converging — a coil, not a trend. RSI 40.6 is neutral-weak, not washed out. Clean lower-high sequence: $387.49 → $290.97 → $363.14 → $324.69. At the 17.6x median on $16.47 trailing EPS the stock is $290 — almost exactly March’s low.
The K-Shape Is the Moat
The cycle is cutting the multiple, not the earnings. US consumer spend ran +11% while U. Michigan sentiment sat at 56.6, down 12.5% YoY — the affluent book is decoupled. The 10% APR cap would hit only the ~23–25% of revenue that is net interest income; ~55–60% is spend-driven discount revenue and ~14–15% is subscription-like card fees. The 23.6% tax rate is a one-time reset already inside the guide.
Panel mean ≈ $347, median ≈ $349 — both below the $375.96 sell-side consensus. Five of six lenses land within $70 of each other; the single wide disagreement is the exit multiple, not the earnings path. Every lens here is a synthesized analytical framework, not a real person or a real firm rating.
Wall Street 12-month price targets
What the sell-side expects over the next year, sorted low to high and colored by rating. The detail worth noticing: seven of these eleven desks carry a Hold and still publish a target above today’s price — the Street likes the company and distrusts the entry point.
Most recent published target per firm, July–August 2026; bars are zero-based, so the dashed line at $324.69 is what matters — only BTIG’s $315 sits below it. Note the rating split: the full book of 30 analysts is 9 Strong Buy, 5 Buy, 15 Hold, 0 Sell, 1 Strong Sell. A “Buy” consensus with half the coverage at Hold-or-worse is what a de-rating looks like from the inside — nobody doubts the franchise, and nobody wants to defend 20x. Firms, ratings and targets as reported by third-party rating feeds as of Sep 11–12, 2026.
Where the loop leads
Synthesized scenario midpoints, each dated to the anniversary of this analysis. Returns are price-only versus today’s $324.69 and exclude the 1.17% dividend. Every number below is an EPS path times an exit multiple; the math is in the collapsibles. These are illustrative frameworks, not forecasts.
1 Year
Sep 20272 Years
Sep 20283 Years
Sep 20295 Years
Sep 2031▸ Bull case — 25% — show the assumptions & math
▸ Base case — 45% — show the assumptions & math
▸ Bear case — 30% — show the assumptions & math
▸ Why the weights are bear-tilted — and what the blend says
Revenue, capex, free cash flow & debt ($B)
Where the money actually goes. For a closed-loop lender the interesting bar is the gap between revenue and cash — and the gap between cash and debt, because the debt is funding, not leverage.
Revenue is Amex’s own reported total revenues net of interest expense — $60.5B (2023), $65.9B (2024), $72.2B (2025), and ~$79.4B for 2026 at the raised ~10% guide. Note that several aggregators publish a series roughly $5B lower on a different adjusted basis; where they conflict we use the company’s own releases. Capex (clay) is the fastest-growing bar on the chart — $1.57B to a $3.43B trailing run-rate, as a new global headquarters and technology spend land at once — and that is precisely why free cash flow (olive) has gone sideways while revenue added $19B. Total debt (slate) runs near one year of revenue, but for a card issuer debt is the raw material: it funds a $218.1B loan book at a 36.4% return on equity, so read it against CET1 at 10.4%, not as gearing. 2026E capex, FCF and debt are extrapolated from the trailing twelve months to June 2026, not company guidance.
The EPS ladder underneath every target ($)
No target on this page was pulled out of the air. Each one is a rung on this ladder multiplied by an exit multiple — so if you disagree with a price, the honest place to disagree is here.
Gray = reported GAAP diluted EPS. Note that 2024’s $14.01 includes roughly $0.66 from the Accertify divestiture gain — strip it and 2025’s $15.38 was up ~15%, not the ~10% the headline suggests; that correction is why the 2024-to-2025 step looks flatter than the business actually was. Olive = the base case, starting from the $17.30–$17.90 FY26 guidance midpoint and compounding 12%/yr, a touch below the ~14% long-run growth consensus. The base case’s $31.02 of 2031 EPS at an 18.5x exit multiple is the $574 five-year base target — that single multiplication is the whole model.
Nothing is actually slowing
Latest reported year-over-year growth, sorted low to high. Steady core lines in olive, faster frontier lines in clay. Read the whole chart against a stock that is down 16% from its high — that disconnect is the bull case.
Every line is positive and the frontier lines are positive by multiples. The four clay bars are all downstream of one decision — the September 2025 Platinum refresh and its $895 reprice. Card fees are the subscription line, and they are growing half again as fast as revenue, with management guiding them to exit 2026 in the high teens. Resy restaurant and lodging spend are the engagement proof: the perks that the bear says are a cost are also what is pulling spend through the loop. The frontier bars run off smaller bases than the core lines; engagement figures are those management cited on the Q2 2026 call.
Bull vs. Bear
The whole valuation argument compresses into one disagreement: is the $895 fee an annuity the shareholder owns, or a coupon book the partners collect?
▲ THE BULL CASE
- The fee line is a subscription, and it is accelerating. Net card fees $2.9B, +15.4% YoY, with thirty-plus consecutive quarters of double-digit growth and management guiding it to exit 2026 in the high teens. Pre-paid, contractual, and largely indifferent to the credit cycle.
- Pricing power was tested and it held. $695 to $895 is +29%, the first increase since 2021, and retention stayed near 100%. There is no other consumer-finance franchise that can do that.
- Spending is re-accelerating, not decaying. Billed business $455.8B, +9.4% FX-adjusted — the fastest in three years. US consumer spend +11%, the best ex-pandemic reading since Q1 2018; international +12%.
- The customer base is getting younger, not older. Gen Z and Millennials are ~65% of new consumer accounts globally and now the largest share of US consumer spend; average new Platinum holder is 33, new Gold 29. 12.5M new cards in 2025, over 70% fee-paying.
- Credit is at a cycle best, not a cycle risk. Write-offs 2.0% and still below 2019, delinquencies 1.2–1.3%, provisions down 23% to $1.1B, and the lowest projected card-loss rate of any bank in the Fed’s stress test.
- The de-rating already happened. 24.0x in 2025, 19.7x trailing and 17.4x forward now — with guidance raised in between. The $16B buyback retires stock into that, and the share count is already down 2.49% YoY.
▼ THE BEAR CASE
- The beat was already spent. Revenue guidance went up; the EPS guide did not move, because H2 marketing accelerates to +10%. Management chose growth over earnings — and the stock fell 5.6% because the market read that as costs, not choice.
- Expenses are outgrowing revenue. Total expenses +12% against revenue +10%, and the revenue line itself missed ($19.637B vs $19.69B expected). Operating leverage is the thing that never arrives.
- Two of the quarter’s best numbers were borrowed. The 23% fall in provisions is a non-repeatable tailwind, and the tax rate normalized to 23.6% from 18.7%. Net income grew only 8% versus EPS +11% — the difference is the buyback, not the business.
- The $895 renewal cliff has been tested once. Existing holders were only repriced in Q1 2026; one partial cycle of ~100% retention against JPMorgan’s $795 Sapphire Reserve is not proof that the ceiling is higher still.
- The book has never seen a downturn. If 65% of new accounts are Gen Z and Millennial at an average age of 33, the 2.0% write-off rate is a statement about the cycle, not about underwriting.
- The Street is not actually onside. Fifteen of thirty analysts are Hold-or-worse and the low target is $315, below spot. Meanwhile the CEO and senior executives sold into February’s selloff with no discretionary buying.
- There is no valuation floor. At 6.4x book, the downside is a multiple, not an asset. The ten-year median is 17.6x and the 2022 trough was 15.0x — the stock can lose a fifth of its value without a single bad quarter.
Risk map — likelihood × impact
Where each risk sits over a three-to-five-year horizon, not just how big it is. The telling shape here: Amex’s single likely-and-high-impact risk is not credit and not competition — it is the multiple.
- Open-banking churn
- Rewards & acquisition cost inflation
- Multiple de-rating to the median
- Commercial share loss
- 10% APR cap
- Interchange caps abroad
- Credit normalization
- The $895 renewal cliff
- Funding / deposit shock
- Agentic commerce severs the loop
Multiple de-rating to the median
What breaks: nothing operational. 19.7x trailing drifts to the 17.6x ten-year median and the stock is $290 — March’s low — with earnings still rising.
Rewards & acquisition cost inflation
What breaks: the $200 fee increase funds ~$3,500 of perks and 20-plus partnerships rather than EPS. Expenses +12% vs revenue +10% is this risk already happening.
Credit normalization
What breaks: the 2.0% write-off rate is a cycle low. Every extra 50bp on a $218.1B book is roughly $1.1B of pre-tax provision, and the 23% provision tailwind flips to a headwind.
The $895 renewal cliff
What breaks: retention slips below ~100% at the first full renewal cycle, card-fee growth falls out of the high teens, and the annuity narrative — the premium multiple’s whole justification — dies with it.
10% APR cap
What breaks: the ~23–25% of revenue that is net interest income. Survivable because Amex is spend-led, not rate-led — but it also caps the loan-growth engine that added +11% NII.
Interchange caps abroad
What breaks: EU and Australian regulators compress merchant discount rates, taking the high-margin Global Merchant & Network Services line with them.
Agentic commerce severs the loop
What breaks: if an AI agent picks the cheapest rail at checkout, premium rewards stop buying the transaction. Amex joining the x402 Foundation in July 2026 is a hedge, not an answer.
Funding / deposit shock
What breaks: a $218.1B loan book needs funding. CET1 at 10.4% and the best stress-test loss rate of any bank make this remote, but a rate or confidence shock reprices the liability side fast.
Commercial share loss
What breaks: Ramp and Brex keep taking middle-market and corporate card volume. A slow leak in a minority of billed business, not a rupture.
Open-banking churn
What breaks: CFPB Section 1033 data portability, effective April 1, 2026, makes the underwriting-data half of the two-sided advantage shareable and nudges switching costs down.
The jargon, decoded
Hover the dotted terms anywhere above, or scan the desk’s working definitions here.
- Closed loop
- Amex is the card issuer, the payment network and the merchant acquirer all at once. Visa and Mastercard only run the network, so Amex keeps the whole merchant fee and sees both sides of every transaction.
- Billed business
- The total dollars charged on Amex cards. The headline demand gauge — $455.8B in Q2 2026, +9.4% adjusted for currency.
- Discount revenue
- The cut Amex takes from merchants on each transaction. Because Amex is also the acquirer, it sets and keeps that rate — roughly 55–60% of revenue.
- Net card fees
- Annual membership fees, net of amortized acquisition costs. The subscription line: $2.9B a quarter, +15.4%, and the closest thing in finance to contractual revenue.
- Net interest income (NII)
- Interest earned on card balances minus the cost of funding them. ~23–25% of revenue — the slice a 10% APR cap would hit.
- Net write-off rate
- The share of loans written off as uncollectable each year. Amex’s 2.0% is the best in the industry and still below its 2019 level.
- Provisions for credit losses
- Money set aside for loans expected to go bad. It fell 23% in Q2 2026 — a real profit tailwind, and one that cannot repeat indefinitely.
- ROE
- Return on equity: annual profit as a percentage of shareholders’ equity. Amex’s 36.4% is exceptional; most large banks run 10–15%.
- CET1 ratio
- Common Equity Tier 1 — the regulatory capital cushion against risk-weighted assets. 10.4% here; regulators require well under that.
- FCF yield
- Free cash flow ÷ market cap, ~6.9% here. Treat it with suspicion for a lender: loan growth sits in investing activities, so the number flatters.
- Forward P/E
- Price divided by the next twelve months of expected earnings. 17.4x here, against a 17.6x ten-year median — the crux of the whole debate.
- Exit multiple
- The P/E assumed at the end of a forecast. Multiply it by projected EPS to get a target price; it is where almost all the disagreement on this page lives.
- Prob-weighted
- Each scenario’s price times its probability, summed into one expected value across bear, base and bull. The clay numbers on this page.
- Agentic commerce
- Purchases initiated by AI agents rather than people. The tail risk: if software picks the payment rail, rewards stop influencing the choice.