Card Fees Inflect and the EPS Floor Rises, but AXP Falls a Third Straight Time: The Debate Has Moved to the Top Line

Results vs. Consensus

Q3 2026 Scorecard

Consensus EPS estimates ranged from $4.47 to $4.51 and revenue estimates from $20.12 billion to $20.13 billion; the table uses the most widely cited figures.

Year-over-Year Comparison

Operating expenses (salaries and employee benefits, professional services, data processing and Other, net) include the $982 million GBTG gain as a credit, along with the $336 million restructuring charge and the higher legal reserves.

Sequential Comparison (vs. Q2 2026)

The Adjusted EPS Bridge

Adjusted pretax and the ~$4.66 figure are derived from the company's reconciliation and the per-share amounts given on the call. The $211 million is the reserve release that management attributed to CECL model enhancements.

Quality of Beat

• Revenue: Revenue grew 9% reported and 9% FX-adjusted, so currency was neutral. Management said growth would have been about a point higher without the small-business co-brand exits, which puts the underlying rate near 10%, unchanged from Q2. The composition still favors the most recurring lines: net card fees grew 16% and net interest income 10% against balance growth of 8%. Discount revenue grew 8%, in line with billings. Service fees and other revenue were flat. The 0.3% miss is about $58 million, and it is spread thinly across card fees, interest income and non-interest revenue rather than concentrated in one line, which reads more like a slightly optimistic Street model than a demand problem.
• Margins: Total expenses grew 9%, in line with revenue, but only because the GBTG gain sits inside them as a credit. Variable customer engagement costs (rewards, business development and Card Member services) rose 15.5% to $8,891 million, taking the ratio to 44.3% of revenue from 41.8% a year ago. Card Member services, the line that carries the Platinum benefits, grew 36% after 50% in Q2, and Q4 2025's $1,951 million base means growth there should fall to low single digits if the line simply holds at Q3's level. Rewards grew 10% against 8% billings, consistent with faster growth in reward-rich travel categories. Underlying operating expenses grew 3%, by management's measure, after removing the gain, the restructuring and the legal items. Marketing grew 11%, as guided.
• EPS: The $4.68 is a fair print. The tax rate of 18.1% against 24.1% a year ago and the $211 million model-driven reserve release both flattered it, and the $0.51 of legal reserves offset them almost exactly. On the per-share amounts management gave, the three items net to about $0.02, so roughly $4.66 is the underlying figure, up about 13% and still about 3.6% above consensus. A 3% lower share count supplied about a fifth of the growth. The quality concern is the pretax line, not EPS: adjusted pretax grew 1.6%, and if the reserve release sits in the segments, segment pretax excluding it grew about 5%.

Segment Performance

Segment revenue and pretax income sum to the consolidated figures through Corporate & Other. GMNS is shown on total network volumes. The Corporate & Other pretax loss narrowed from ($598) million a year ago.

U.S. Consumer Services

The segment that carries the thesis had its best profit quarter of the five shown in the supplement. Revenue grew 11% to $9,840 million and pretax income grew 24% to $2,295 million, with provisions down 33% to $494 million and the principal-only write-off rate down to 1.6%. Billings grew 9.4%, down from Q2's 11.4% but in line with Q1, and management pinned part of the sequential slowdown on Prime Day and other retailer promotions shifting into Q2. Variable engagement costs in the segment rose 16% to $4,801 million and marketing only 3%, so the refresh still shows up as a cost the revenue line is catching up to.

"In fact, Q3 is the third highest quarterly growth rate in three years. One year after the U.S. Consumer Platinum refresh, spend retention remains at 99% for the portfolio." — Christophe Le Caillec, CFO

Proprietary cards-in-force grew 4% to 49.7 million, and average spend per proprietary basic card grew 4% to $5,522. The segment's net interest income grew 10% to $3,549 million against card balance growth of 7%.

Assessment: A 24% pretax gain on 11% revenue in the largest segment is the second consecutive quarter of operating leverage where the money is being spent. Ninety-nine percent spend retention a full year after a $200 fee increase is the strongest single proof of the premium moat this cycle has produced. The 2-point billings slowdown is real but mostly calendar, and it should not be read as the refresh fading.

Commercial Services

Commercial is where the co-brand exits land, and it is also the weakest segment underneath them. Revenue grew 6% to $4,553 million while pretax income fell 6% to $1,029 million, as segment expenses grew 14% with marketing up 29%. Billings grew 3%, or 5% excluding the exits for a second straight quarter, and proprietary cards-in-force fell 11% to 13.7 million as the transferred portfolios left. Average spend per card rose 13%, mostly a mix effect from losing lower-spend co-brand accounts. Net interest income grew 12% and provisions fell 16%.

"We are seeing strong take-up on the Graphite Business Cash Unlimited Card that we launched earlier this year. Looking ahead, our new Corporate Cashback Card and integrated expense management capabilities will give us additional opportunities to drive engagement and spend growth." — Christophe Le Caillec, CFO

In the quarter the company launched the next generation of Amex Corporate with integrated expense-management software and a mobile expense app, introduced Corporate Cashback Cards, and added a Business Savings account for small businesses.

Assessment: A 29% increase in marketing against 5% underlying billings growth is a deliberate catch-up spend in the middle market, not yet a payoff. The product set is now competitive on paper. The numbers will not show it until 2027, and this segment is where the thesis has the least to show.

International Card Services

Billings grew 11% reported and FX-adjusted, the 22nd straight quarter of double-digit growth, but down from 13% reported in Q2. Revenue grew 10% to $3,682 million and net interest income 20%. Pretax income fell 2% to $432 million, with segment expenses up 14% and variable engagement costs up 17%. That answers part of our Q2 question about the sequential drop from $781 million to $477 million: the supplement shows Q1's segment operating expense line at $630 million against $884 million in Q2 and $883 million in Q3, so Q1 was the outlier, not Q2. It does not answer the bigger question, which is why a segment growing revenue 10% is not growing profit.

"Look, it's 22 straight quarters of double-digit growth rate in international. We went from 13% to 11%. We'll see how that plays out, but 11% growth is still very strong growth." — Stephen Squeri, Chairman and CEO

Management tied international acquisition to the same younger demographic driving the U.S. book and pointed to merchant coverage of more than 190 million locations, with locations accepting the card outside the U.S. more than doubling over four years.

Assessment: International is still in the invest phase of the cycle the U.S. is now leaving, and the top line continues to justify the spend. But two consecutive quarters of pretax growth of 3% or less on double-digit revenue growth is no longer a timing story we can wave through. "We'll see how that plays out" was also the least committal answer of the call. The pillar stays AT RISK.

Global Merchant & Network Services

Revenue grew 6% to $2,082 million and pretax income 4% to $1,082 million on network volumes of $517.9 billion, up 8%. Non-interest revenues grew 8%, while the segment's net interest income fell 18% to $155 million. Expenses grew 7%.

"Merchant coverage now reaches over 190 million locations globally, up from 160 million a year ago." — Christophe Le Caillec, CFO

Assessment: GMNS tracks volume with a lag and needs no narrative. The strategic content this quarter is in the network partnerships discussed in Q&A: management has stopped courting large bank issuers and is signing smaller partners, which widens reach without making the network depend on a competitor.

Key KPIs

The VCE ratio is rewards, business development and Card Member services divided by revenue. ROE is annualized quarterly net income over average equity, as the company reports it; on adjusted net income of $3,214 million the same calculation gives about 36.7%.

Key Topics & Management Commentary

Overall Management Tone: Controlled and itemized. Management opened by walking through every discrete item before it could be found, delivered the regulatory order with an explicit statement of accountability rather than a minimization, and spent most of the prepared remarks on an AI strategy built to answer the competitive question investors had been asking. The posture was less combative than Q2's pre-emptive defense of reinvestment, and the one place it went soft was international, where the answer to a 2-point slowdown was to wait and see.

1. The Quarter Had Five Moving Parts, and Management Netted Them in Public

The GAAP-to-adjusted bridge removes two items, the GBTG gain and the restructuring charge. The more useful disclosure came after it: three further items stay inside adjusted EPS, and the CFO sized each one per share and said they cancel. That converts a print that could have been read as tax-and-reserve-assisted into one the market could take at face value, and it is the main reason we treat $4.68 as the quarter's real earnings power.

"Adjusted EPS includes three other notable items this quarter. First, an increase in legal reserves of $0.51, which includes the remaining portion of the OCC civil money penalty that was not previously reserved for, as well as a reserve for the anticipated settlement of an unrelated matter. The second item is a discrete tax benefit of $0.29 per share. Finally, a $0.24 benefit from the most recent updates to the CECL reserve models. Combined, these three items had no meaningful impact on the adjusted EPS this quarter." — Christophe Le Caillec, CFO

The arithmetic holds: $0.29 plus $0.24 less $0.51 is $0.02. What it does not resolve is that the "unrelated matter" behind part of the legal reserve was not identified on the call.

Assessment: Full itemization is a credibility asset, and management spent it well this quarter. A market that still marked the stock down on quality-of-beat concerns was reacting to the headline tax rate, not to the disclosed arithmetic.

2. A 5% Headcount Restructuring Funded by the GBTG Gain

The GBTG stake sale closed in the quarter, and management said it is dropping two-thirds of the gain to the bottom line. It used the rest to fund a $336 million restructuring charge. This is the first efficiency program management has announced in our coverage, and the CEO called the charge larger than the company's usual restructuring.

"We decided to use some of the gain to invest in making the organization more efficient and better prepared for the future. As a result, we took a $336 million charge for restructuring in the quarter, which is expected to result in the elimination of approximately 5% of our positions globally." — Stephen Squeri, Chairman and CEO

The stated purpose is speed and capacity rather than margin: management described an exercise to remove layers and align teams on the critical deliverables, and said the result would give it "greater capacity to invest in our business priorities." No savings figure and no completion date were given.

Assessment: This partly answers our Bear-1 complaint that every dollar of upside is redeployed into growth spending. Some of the windfall went into making the cost base leaner instead. But "greater capacity to invest" tells us the savings will be recycled too, so investors should not model a margin dividend from it in 2027.

3. The OCC and Fed Consent Orders

On October 8, American Express National Bank consented to an OCC order and agreed to a $350 million civil money penalty, and the holding company consented to a parallel Federal Reserve order. Both resolve reviews of the company's financial-crimes compliance program. Part of the penalty had been reserved earlier, and the unreserved remainder sits in this quarter's legal charge.

"The orders do not impose an asset cap on the company, and the costs associated with the remediation are not anticipated to affect 2027 guidance. More importantly, the deficiencies cited in the consent order should not have occurred, and we are accountable for addressing them thoroughly." — Stephen Squeri, Chairman and CEO

The CFO added that the company has invested significantly in risk and controls over the past few years and that the ongoing investment is already in the run rate. The orders do not cap the balance sheet, which is the outcome that would have mattered most for a lender growing balances 8% a year.

Assessment: The financial cost is contained and the tone was right. The overhang is not the penalty but the constraint: consent orders tend to slow product launches and banking expansion while remediation is under review, and the company has just told investors that banking products are a bigger part of its future. We add this as a new bear point at EMERGING and will grade it on whether the deposit and banking roadmap slips.

4. Net Card Fees Inflected on Schedule

Net card fees reached $2,969 million, up 16% reported and 17% FX-adjusted, after 15.4% in Q2. This was the line we called a must-deliver in July, because the entire reinvestment argument depends on the repriced Platinum book amortizing into revenue as described. It did. Average fee per card rose 14% to $136, and 76% of new accounts came in on fee-paying products, the highest since the company began reporting the measure.

"Net card fees continue to be our fastest-growing revenue line. As expected, we have seen an acceleration in growth to 17% FX adjusted this quarter, supported by the very strong results we have seen on the U.S. Platinum refresh." — Christophe Le Caillec, CFO

The CEO put the quarterly fee line at $3 billion, up 17%. One detail matters: the explicit high-teens exit rate for Q4 was not in this quarter's slide deck, and when asked, management confirmed the inflection without restating the exit target, citing a policy of not giving quarterly card-fee guidance.

Assessment: The most important promise of 2026 was kept, and the two-year amortization tail means the fee benefit continues into 2027. We read the missing exit-rate language as caution about precision, not a walk-back, but a Q4 print below 17% would now be a real disappointment.

5. The Refresh Laps: Engagement Cost Growth Moderates From Here

Q3 is the quarter in which the U.S. Consumer Platinum refresh fully laps, and the cost line has started to show it. Card Member services grew 36% after 50% in Q2, and the VCE ratio slipped to 44.3% from 44.6%. Management was unusually specific that Q4 brings a sharp slowdown in VCE dollar growth.

"We are beginning to see growth in card member services moderate following the step-up from the Platinum refresh last year. Looking ahead, we expect VCE growth to moderate significantly in Q4 as we fully lap the Platinum refresh, returning to growth rates more consistent with what we saw before the refresh." — Christophe Le Caillec, CFO

Management also highlighted a new way to redeem Membership Rewards against statement balances, which it said had strong adoption and supports the program's economics. Year-to-date, the VCE ratio is 44.5%, inside the 44%–45% full-year range raised in July.

Assessment: This is the mechanical catalyst the thesis has waited for since April. With revenue compounding near 10% and VCE dollar growth falling back toward pre-refresh rates, Q4 should be the first quarter of this cycle in which operating leverage is visible in the consolidated pretax line, not only in USCS.

6. Billings Slowed to 8%, and Most of the Slowdown Is Explained

Billed business grew 8% FX-adjusted after 9% in Q2. Management attributed the gap to two things: about 60bps of drag from the co-brand exits, with the Amazon portfolio transferring mid-quarter, and a sequential timing effect from Prime Day and other retailer promotions that fell in Q2. Goods and services spending grew 7%, where those effects concentrate, while travel and entertainment grew 10%, airline spending accelerated to 13% and restaurants grew 9%.

"This quarter included about a 60 basis point year-over-year impact from the co-brand exits that I previously discussed, with the Amazon co-brand portfolio sale happening midway through the quarter. Sequentially, spend growth also saw a small impact from the last quarter's Amazon Prime Day and other promotional retail days in the U.S." — Christophe Le Caillec, CFO

The CEO added that spending at Resy restaurants rose 18% and Fine Hotels + Resorts and The Hotel Collection bookings rose 50%, which he offered as evidence that pairing cardholders with partner merchants lifts spend.

Assessment: The explanation is credible for U.S. consumer and commercial, where the co-brand mechanics are disclosed. It does not cover international, which slowed by 2 points with no co-brand excuse. Because Q4 carries the first full quarter without the Amazon portfolio, the reported spend line will look weaker before it looks better, and investors should expect that.

7. Credit: Real Improvement, With a Model-Driven Release on Top

Provisions fell 23% to $985 million and included a $211 million reserve release, which management attributed to enhancements to its CECL models rather than to a change in economic outlook. Underneath the release, the credit data improved on their own: the principal-only write-off rate fell to 1.8% from 2.0% in Q2 and 1.9% a year ago, and 30-day delinquencies held at 1.2%.

"Balances grew 8%, NII grew at a faster rate at 10%, and write-off dollars only grew at 3%. Delinquency rate remained very low and stable. For the last 15 quarters, the rate has been either 1.2% or 1.3%." — Christophe Le Caillec, CFO

The reserve fell to 2.5% of Card balances from 2.7% in Q2 and 2.9% a year ago. Management said new vintages are performing better than older ones, which it tied to the fee-paying acquisition mix.

Assessment: Separate the two things. The release is an accounting change worth $0.24 that will not repeat, and we do not model further model-driven releases. The write-off and delinquency trend is the durable signal, and it keeps credit the most reliable pillar of the thesis.

8. AI as a Membership Strategy, Not Only a Cost Program

The CEO used most of his prepared remarks to set out how the company intends to compete as AI agents start to search, compare and pay on customers' behalf. He named five areas (value propositions, acquisition, engagement, payments and service) and framed each as a way to make membership more valuable rather than to cut cost.

"It is still early in the game. Agentic transaction volumes are still small, but the direction is clear. AI is fundamentally changing how consumers and businesses discover, compare, decide, pay, and manage their finances and their lives. But while AI agents are able to act on customers' behalf, that does not change who the customer is." — Stephen Squeri, Chairman and CEO

The concrete items: an OpenAI partnership that already includes Resy in ChatGPT and will add an Amex Offers plugin, an Agentic Commerce Experiences developer kit published in Q2, an Intent Intelligence framework released this month, and an Agent Purchase Protection product in development. Management also said its first conversational AI voice agents have launched in servicing.

Assessment: The strategy is coherent, and the closed loop is a real advantage in a world where a transaction has to carry the customer's intent and authorization. It is also pre-revenue and defensive in origin: the speech was built to answer the rewards-optimization concern before Q&A raised it, which it then did twice.

9. The Middle-Market Rebuild in Commercial

Commercial billings grew 3% reported and 5% excluding the co-brand exits, the same underlying rate as Q2. Management named middle-market companies as the sub-segment where it has under-grown, and the product response this quarter was substantial: next-generation Amex Corporate with integrated expense management, Corporate Cashback Cards, and a Business Savings account.

"Turning to commercial, spend was up 3% this quarter. Excluding the impact of the co-brand portfolio exits, spend growth was 5% for the second consecutive quarter." — Christophe Le Caillec, CFO

Management named the fintech competitors directly in Q&A and said the capabilities it has assembled were needed to compete with them, with agents to be added to the platform and the tools to be rolled out to existing customers over time.

Assessment: Commercial's growth has stopped falling, and a full product response is now in market. That is progress, but it has not yet produced a number, and a segment whose pretax fell 6% on 29% higher marketing spend is the clearest place where the reinvestment still has to prove itself.

10. Partnerships: Airbnb, Marriott and a New Network Strategy

The quarter brought an exclusive partnership with Airbnb that adds more than one million listings to Amex Travel, an extension of the long-term Marriott partnership including the U.S. co-brand cards, and new and larger Centurion Lounges. On the network side, management described a deliberate turn away from large bank issuers toward smaller and mid-sized partners.

"We know that card members enjoy the exceptional variety and scope of Airbnb's network of unique locations, and we are excited to bring over 1 million listings globally into our Amex Travel program, as well as expanding travel benefits for eligible U.S. card members." — Stephen Squeri, Chairman and CEO

The company also ranked first in the J.D. Power U.S. credit card satisfaction study for a seventh consecutive year.

Assessment: Renewing Marriott removes a renewal risk on a large consumer co-brand portfolio at a time when the company is already absorbing two smaller exits. Airbnb is the kind of benefit that is cheap to add and hard for a bank issuer to replicate, which is the membership-model answer to rewards optimization.

11. Capital Return and the Long-Run ROE Defense

The company returned $2.2 billion in Q3: $1.6 billion of repurchases and $0.6 billion of dividends, down from $2.9 billion in Q2, when repurchases were $2.2 billion. CET1 rose to 10.9% from 10.4%, helped by the GBTG gain booked at quarter-end. The CFO used a 16-year ROE chart to answer the recurring margin question.

"As you can see, we have increased the ROE from an average of 25%-28% ten years ago to around 35% more recently. Over this period of time, we are on track to more than double the size of our revenue and triple EPS. At the same time, we have also reduced the share count by over a quarter." — Christophe Le Caillec, CFO

Shares outstanding fell to 671 million at quarter-end, down from 689 million a year earlier, and book value per share rose 14% to $51.02.

Assessment: The ROE argument is the best long-run defense of the reinvestment model and it is backed by data. The slower buyback in a quarter when the stock was cheaper, with capital rising, was not explained. With the consent orders now signed, we would expect the repurchase pace to recover in Q4.

Guidance & Outlook

The guide changed in a narrow way. Revenue growth of 10% was reaffirmed. The adjusted EPS range lost its bottom $0.30, moving the midpoint to $17.75 from $17.60 while the $17.90 top end held. The GAAP range sits about $0.70 above the adjusted range, reflecting the after-tax gain net of the restructuring charge ($1.10 less $0.37 in the company's reconciliation, which notes that amounts may not sum due to rounding). The CEO framed the full year as "another year of 10% revenue growth and mid-teen EPS growth," and the adjusted midpoint is 15.4% above FY2025's $15.38.

Implied Q4: Nine-month adjusted EPS is $13.49, so the range implies Q4 adjusted EPS of $4.11–$4.41, with a $4.26 midpoint. That is 16% to 25% above Q4 2025's $3.53. A correction to our July framing belongs here: we argued then that holding Q2's $4.53 for two quarters would land near the top of the range, and that ignored seasonality. Q4 is the company's weakest earnings quarter (Q4 2025 EPS was 15% below Q3 2025), and applying the same step-down to Q3's $4.68 gives about $3.99, below the new floor. The guide therefore asks for a better-than-seasonal Q4. The VCE lap makes that plausible, because Card Member services ran at $1,951 million in Q4 2025 and its growth should fall from 36% to low single digits if the line holds near Q3's level. On revenue, Q4 needs about $20.85 billion (+9.8%) for exactly 10% growth for the year, or about $20.49 billion (+7.9%) to round to it.

Street at: Full-year consensus sat at about $17.65–$17.69 before the print, so the new $17.75 midpoint is slightly above it. That is almost entirely the Q3 beat passing through. The Street's implied Q4 was about $4.28–$4.37, against a guide midpoint of $4.26, so the guide did not raise the bar for Q4.

Guidance style: Consistent with the two prior quarters: conservative on the outer bound and explicit about what is being reinvested. This is the first EPS raise of the year, and it came only from the floor. The pattern still says management would rather beat a held top end than raise it.

Analyst Q&A Highlights

How the model responds if customers optimize rewards across cards

The opening question went to the concern management had pre-empted in its prepared remarks: whether customers, increasingly helped by AI tools, will route spending to whichever card pays the most rewards, and how the company would respond. Management answered with two structural points and one operational lever.

Q: "I guess one of the concerns that I have been hearing quite a bit about from investors is just how people might gamify rewards and allocate their spending to the highest rewards benefit. To the extent that that would happen, how would you guys adjust your model? Or is there something that you guys can do offensively to stay on the offensive?"
— Sanjay Sakhrani, KBW

A: "Our strategy really has not been to really compete simply on points or rational rewards. I think the other thing that is important to understand is we really do not have a breakage model. So we do not benefit when people do not use our rewards... we have levers within our model, and we have shown this with the Platinum card, where we can get to spend thresholds that require those benefits to be made, which then sort of eliminates that ability to game the system."
— Stephen Squeri, Chairman and CEO

Assessment: The no-breakage point is the strongest answer available and it is underappreciated: an issuer that already prices for full redemption loses less when redemption becomes more efficient. The spend-threshold lever is real but it trades benefit value for cost control, so it can only be used sparingly before it erodes the product.

VCE moderation and the path to mid-teens EPS in 2027

A question tied the expected VCE slowdown to the 2027 earnings outlook and asked how confident management was in mid-teens EPS growth next year. The answer gave the clearest Q4 cost guidance of the call and declined to give a 2027 number.

Q: "Then given Christophe's comments regarding the moderating of VCEs, can you maybe just talk big picture about the potential for mid-teens EPS into next year and your confidence in it?"
— Ryan Nash, Goldman Sachs

A: "As you think about going forward, and Q4 is going to be a good example for that, you should expect that VCE ratio to stabilize in this kind of range. But the growth rate in terms of the VCE dollars to moderate significantly and either to be more consistent with what we saw pre-refresh. We are not going to get into guidance for 2027 at this stage... That is going to help us just maintain that mid-teens EPS ambition that we have been using for years to think about the performance of this company. I feel good about that."
— Christophe Le Caillec, CFO

Assessment: The ratio stabilizing while dollar growth slows is the combination that produces operating leverage, and the CFO linked it to the mid-teens ambition explicitly. "I feel good about that" is as close to a 2027 commitment as this management gives in October.

Whether the high-teens card-fee exit still stands

An analyst noticed that the Q4 high-teens card-fee exit rate had disappeared from the slide deck and asked whether anything had changed. Management confirmed the inflection and declined to restate a quarterly target.

Q: "Just on the card fee front, I know you noted last quarter that card fees should grow in the high teens in the fourth quarter. You did not seem to put that in your slide deck this quarter, but I believe, Steve Squeri, you just mentioned that you should continue down that path in the, you mentioned the 17% the third quarter. Is there any change there in how you are thinking about the fourth quarter..."
— John Pancari, Evercore ISI

A: "We saw that inflection point this quarter. You saw it like we were at 15% in Q2, and this quarter we are at 17% FX adjusted. That is the inflection point that we were expecting and that we saw. I am not going to give you guidance about how to think about this 2027 card fee. Again, you are going to have to wait until January for that. We also do not give quarterly guidance on card fee."
— Christophe Le Caillec, CFO

Assessment: The answer is consistent with policy but it is a softer formulation than July's, when the CFO volunteered the high-teens exit. 17% is already at the bottom of "high teens," so the risk is small. A company confident in 18%–19% would probably have said so.

How much conservatism is left in the reserve rate

A question asked whether the reserve coverage ratio, cited at 2.6% on Card balances and other loans combined, still embeds conservatism, or what a normal-environment ratio would be. Management attributed the decline to portfolio mix and said its downside scenarios have not been softened.

Q: "I guess the question is, how much does the 2.6% reflect any conservatism with respect to the broader environment? Or maybe to ask the question another way, is there any color you can give on what you think a normal environment allowance coverage ratio might be?"
— Rob Wildhack, Autonomous Research

A: "The key dynamic here is the premiumization of the portfolio. As we focus more on fee-paying card members, as we focus more on VCE, we are attracting card members that have a great credit profile, and that slowly improving the credit metrics across the portfolio... When you look at by vintage, the new vintages are performing better from a credit standpoint than the past vintages... And we haven't changed the kind of economic scenarios that we use in terms of the conservative scenario over the past years."
— Christophe Le Caillec, CFO

Assessment: Unchanged downside scenarios plus better new vintages is a coherent explanation for a falling reserve rate. The question of what "normal" coverage is went unanswered, which leaves investors unable to size how much more release is possible. We assume none.

Whether premium competition has intensified

Premium competition was a recurring line of questioning, framed around large banks adding marketing dollars and a newly closed-loop competitor willing to sacrifice near-term profit. Management's position was that the market has been this competitive for a decade and that the competition is enlarging the premium pool.

Q: "Does it feel any different today, particularly with players like Capital One now operating a closed loop, generating higher returns, and having historically demonstrated a willingness to step up investments to prioritize their long-term positioning, even if that means compromising near-term profitability?"
— Multiple analysts incl. Bill Carcache, Piper Sandler; Don Fandetti, Wells Fargo

A: "Look, I think Citi, Capital One, Chase, everybody is sort of playing in this space. To be honest, we really don't see much different than what we've seen. It's highly competitive. It's been highly competitive, and our assumption is it will remain highly competitive. It's highly competitive because it's a very profitable segment of the card base."
— Stephen Squeri, Chairman and CEO

Assessment: The data support the answer this quarter: 99% spend retention, a record fee-paying acquisition mix and accelerating card fees are not what share loss looks like. But "we really don't see much different" is the answer every incumbent gives until the numbers turn, so we keep watching the fee-paying mix and retention rather than the rhetoric.

Whether AI agents will be allowed into the system

A question asked whether the company would let third-party AI assistants access its systems, unlike banks that block them, and whether investors can measure the financial effect of agents helping members extract more value. Management drew a clear access line and declined to size the effect.

Q: "I guess I'm wondering, given that you're planning to be at the forefront of AI, does that lead us to conclude that you, unlike some financial institutions that are blocking access from the bots, you will allow AI parties into the system and therefore allow your members to help them maximize their experience?"
— Erika Najarian, UBS

A: "We're just not going to let assistants run rogue through our systems. That's not going to happen. We need to protect our customers, merchants, and we're only going to allow agents access to our systems that are either approved through API or MCP channels... I think it's a little too early in the game to determine what the impact will be on VCE."
— Stephen Squeri, Chairman and CEO

Assessment: Gated access through approved channels is the right policy and it keeps the company in control of where its benefits are consumed. Calling the VCE effect "a little too early in the game to determine" is honest, and it means the AI-optimization risk cannot be ruled out on current data. That is part of why our conviction is lower.

What progress in the commercial middle market should look like

A question asked for milestones on commercial, where management has conceded under-growth. The answer located the problem precisely, named the competitors, and set a slow timeline.

Q: "Can you just retouch again on the competitive landscape around commercial for a moment? I know you mentioned a bit more about premium, but just revisiting your initiatives, what kind of progress and milestones you would like to see in terms of some element of an acceleration potential in the overall segment volume..."
— Darrin Peller, Wolfe Research

A: "I think where the competitive piece is really within the middle market piece there. That is where I think from our perspective, we have not grown the way we want to grow, and that is why we did the Center acquisition... These are things that we needed to more effectively compete with Ramp and with Brex... I think it will continue to take a little bit of time"
— Stephen Squeri, Chairman and CEO

Assessment: Naming the competitors and conceding that the company has "not grown the way we want to grow" is candid. "A little bit of time" gives investors no milestone to grade, so we will use the ex-co-brand billings rate, 5% for two quarters, as our own.

The network strategy and the card-network settlement

A question asked whether the pending Visa and Mastercard merchant settlement, which would relax honor-all-cards rules and widen surcharging, changes the company's network thinking, given a run of network partnership deals this year. Management described a shift in partner strategy and dismissed the settlement's practical effect.

Q: "I am just wondering if you are rethinking your network business and the implications of the settlement."
— Mihir Bhatia, Bank of America

A: "we are not looking to partner with the bigger banks. I think we made a lot of efforts to attempt to partner with some of the bigger banks. That did not really pay off for us. I think whether we are now partnering with Synchrony or we are partnering with Bread Financial or we are partnering with others, I think it is allowing us to reach more segments of the population that we traditionally would not reach."
— Stephen Squeri, Chairman and CEO

Assessment: Conceding that large-bank network partnerships "did not really pay off" is a strategic admission, and the pivot to smaller issuers is sensible. On the settlement, management's view that surcharging has not hurt anyone and that selective card acceptance will provoke consumer backlash is plausible, but it is an opinion, and the settlement remains an unpriced tail risk to discount revenue.

What They're NOT Saying

1. What the restructuring saves. A $336 million charge to remove about 5% of positions came with no savings estimate, no completion date and no split between reinvestment and margin. The only guidance is that it creates "greater capacity to invest," which suggests the savings are already spoken for.
2. The high-teens card-fee exit rate. July's explicit Q4 target was dropped from the deck and not restated when asked. Management confirmed the inflection and cited a no-quarterly-guidance policy it had not applied in July.
3. Why international is not growing profit. International pretax fell 2% on 10% revenue growth for a second weak quarter. No analyst asked and management did not volunteer an explanation for the profit gap. The billings slowdown, the only international question asked, got "we'll see how that plays out."
4. TheFork. The proposed European dining acquisition, introduced in July with second-half deal and integration costs, was not mentioned at all: no closing status, no cost, no timing.
5. The "unrelated matter." Part of the $0.51 legal reserve covers the anticipated settlement of a matter management did not name. Investors can see the cost and not the exposure.
6. Why buybacks slowed. Repurchases fell to $1.6 billion from $2.2 billion in a quarter when the stock was lower and CET1 rose to 10.9%. The consent orders are the obvious candidate, and the call did not address it.
7. What consent-order remediation costs. Management said the ongoing investment is in the run rate and will not affect 2027 guidance, but it gave no figure for the risk and controls spend.
8. A normal reserve rate. Asked directly what a normal-environment allowance coverage ratio looks like, management explained the decline and gave no anchor, so investors cannot size what further releases are possible.

Market Reaction

• Pre-print setup: AXP closed at $325.17 on October 22, down 12.1% year to date against an 11.9% gain for the S&P 500, down 8.4% over trailing twelve months, and up 7.7% over the trailing thirty days. The 52-week closing range entering the print was $292.27 to $384.89, so the stock entered the print in the lower half of its range after a 7.7% gain in the month before it.
• Pre-market: Shares traded about 1.6% to 2.4% lower before the open as the release, the restructuring and the revenue miss circulated.
• Reaction session (October 23, before-open reporter): Opened at $315.00, a 3.1% gap down, recovered to nearly $321 within the first half hour and traded as high as $321.20 (down 1.2%), then faded to close at $313.22, down 3.7% or $11.95, near the $311.40 low.
• Volume: 4.1 million shares against a 3.3 million 30-day average, 1.2 times normal, well below Q2's 2.1 times.
• Market and peers: The S&P 500 fell 0.5%. Visa and Mastercard each fell 1.5% and Capital One fell 1.0%, so about two points of the move were specific to American Express.
• Since coverage: The stock is down 1.7% since our April initiation and down 4.0% since our July note, against gains of 7.2% and 2.9% for the S&P 500 over the same periods. This is the third consecutive print-day decline: 4.3% in April, 4.3% in July and 3.7% now.

A different objection this time. In April and July the stock fell on capital allocation: the EPS guide was held while upside was reinvested. This quarter the EPS guide went up, which removes that complaint, and the stock fell anyway on the top line. Revenue missed for a second quarter, billings and network volume were light, and the reported growth rate has stepped down from 11% to 10% to 9% across the year. Put the restructuring headline and a tax-rate-assisted net income line beside that, and a stock that had gained 7.7% in the month before the print gave back about half of it.

Low conviction selling. Volume at 1.2 times average is the softest print-day turnover of the three quarters we have covered, and the recovery to nearly $321 in the first half hour shows that buyers were present at the open. The fade into the close came on a weak day for financials and the card networks. We read the day as a positioning unwind on a quarter without a clean headline, not as a verdict on the franchise. A third consecutive decline does tell us that the market will not re-rate this stock on management's framing. It needs to see consolidated operating leverage in a reported number, and Q4 is the first quarter that can show it.

Street Perspective

Debate: Was the beat real?

Bull view: Management itemized every discrete item and showed that the tax benefit and the reserve release were offset by legal charges. Underlying EPS near $4.66 still beat consensus by more than 3%, credit improved on its own, and the adjusted floor rose by $0.30.

Bear view: The beat came from a provision line about $325 million below the Street and an 18.1% tax rate, while every revenue line that matters came in light. Adjusted pretax income grew less than 2%, and an earnings model that needs accounting help in the quarter it laps its biggest cost step-up deserves a discount.

Our take: The bulls are right about EPS and the bears are right about pretax. The disclosed items net to zero, so $4.68 is fair. The pretax weakness is real but it reflects the refresh-cost peak, with VCE up 15.5% on 9% revenue. That peak ends in Q4, and the bear argument has a short shelf life if VCE growth slows as guided.

Debate: Is the top line decelerating, or is it co-brand noise?

Bull view: Excluding the small-business co-brand exits, revenue growth is near 10%, unchanged from Q2. U.S. consumer billings at 9.4% are the third best in three years despite Prime Day timing, T&E is growing 10%, airline 13%, and card fees accelerated. The exits carry negligible profit, so the reported slowdown costs nothing.

Bear view: Two consecutive revenue misses, light card fees and interest income, a network-volume miss, and an international slowdown with no co-brand excuse add up to a pattern. Banks are pouring marketing dollars into premium cards, and the company is spending more to grow at the same rate.

Our take: Mostly noise in Q3, with one real signal in international. The harder point for bulls is timing: Q4 carries a full quarter without the Amazon portfolio, so reported growth will look worse before the exits lap in mid-2027. Investors waiting for a clean revenue acceleration will wait two more prints, and that is a reason we lowered conviction rather than the rating.

Debate: Does AI-driven rewards optimization erode the premium model?

Bull view: The company does not depend on breakage, controls access to its systems, owns a closed loop that lets it match members with merchants, and has shown it can add spend thresholds when benefits are gamed. Its membership benefits, such as Resy, Fine Hotels + Resorts, Airbnb and lounges, are harder to optimize away than points.

Bear view: If agents route every transaction to the highest-earning card, premium issuers converge on rewards cost and differentiation shrinks. Management itself said it is too early to size the VCE effect, and competitors with closed loops and bigger marketing budgets are entering the same space.

Our take: This is the debate that sets the multiple, and neither side can prove it yet. On current data the bulls lead, with spend retention at 99% and the fee-paying mix at a record. But a risk management cannot size is a risk investors will price, which is part of why we cut the multiple in our fair value.

Debate: What re-rates the stock?

Bull view: At 17.6x the adjusted guide midpoint and about 15x next year's earnings, a 35%-plus ROE franchise with best-in-class credit is priced like an ordinary lender. The Q4 VCE lap is mechanical, and the January 2027 guide is a dated catalyst.

Bear view: The stock has fallen on three straight prints and the market has shown it will not pay for management's framing. Add a consent order, a restructuring with no disclosed savings, and a revenue line that keeps missing by a little, and the multiple can stay in the mid-teens for a long time.

Our take: The catalyst is now on the calendar: the Q4 print and 2027 guide in late January. If it shows consolidated operating leverage and frames 2027 in the mid-teens, a high-teens multiple on next year's earnings is defensible. We underwrite 16.5x–17.5x, below our initiation multiple, and still see low-double-digit total return from here.

Model Update & Valuation Framework

Valuation impact: We roll the framework to FY2027 and set fair value at $335–355, which is 16.5x–17.5x our $20.30 FY2027 estimate. That is the same dollar range we initiated with in April, reached a different way: then we used 19x–20x on FY2026, and we now use a lower multiple on a year-later earnings base. We cut the multiple because the market has declined to pay a high-teens multiple for three quarters while revenue growth stepped down, and because a consent order and an unsized AI-rewards risk have been added. At $313.22 the stock trades at 17.6x the FY2026 adjusted midpoint and 15.4x our FY2027 estimate. The $345 midpoint implies about 10% upside, plus a 1.2% dividend yield on the $0.95 quarterly dividend, for a 12-month total return near 11%.

Thesis Scorecard Post-Earnings

Overall: Thesis unchanged on the business, weaker on the path. The promise that mattered most, the card-fee inflection, was kept on schedule, the VCE lap that should produce consolidated operating leverage is one quarter away, and credit keeps improving. Against that, Bear-2 has moved from contained to emerging, international still cannot convert growth into profit, a regulatory order has been added, and the market has now sold three consecutive prints. Our April call has trailed the S&P 500 by about nine points since initiation. The fundamentals have done roughly what we said they would, and the stock has not.

Action: Hold the position at a smaller size than conviction 7 justified, and treat the late-January Q4 print and 2027 guide as the decision point. We would move to Hold on any of four signals: a 2027 EPS framework below low-teens growth or a retreat from the mid-teens ambition; Q4 net card fee growth below 17%; Q4 VCE dollar growth that does not fall to high single digits or below after the lap; or any expansion of the consent orders, including an asset cap, new penalties, or remediation costs that management says will affect 2027.