Pre-Provision Earnings Jump 10%, ROTCE Hits 21%, and Earnings Power Comes Into View: Upgrading to Outperform

Results vs. Consensus

Three months ago we kept Capital One at Hold after a 24% EPS beat because the beat lived below the pre-provision line: the provision fell, the reserve was released, and the operating engine stood still. Q3 is the mirror image. The provision barely moved (down $73M), the reserve release shrank to $225M from $662M, and yet pre-tax income still rose 20% sequentially, because revenue grew 3% while expenses fell 2%. Strip both quarters' releases out and pre-tax income rose 38% quarter over quarter, to $4.35B from $3.16B. The headline beat is a little smaller than Q2's in percentage terms, but its composition is what the thesis had been waiting for.

Q3 2026 Scorecard

Year-Over-Year Comparison

Q3 2025 was the first full quarter with Discover, so this is the first clean like-for-like year-over-year comparison of the combined company.

Quarter-Over-Quarter Comparison

Quality of the Quarter

Revenue: Solid and better composed than it looks at 3%. Net interest income rose 4% on a 40bps jump in the margin to 8.41%, driven by a higher domestic card yield (up 50bps to 17.66%), $13 billion less average cash, and an extra day. Non-interest income was flat sequentially but up 17% year over year, with discount and interchange fees up 29% as the debit network synergies sit fully in the run-rate. Domestic card revenue came in about $0.2B ahead of what the few segment estimates expected. The year-over-year revenue growth of 6.4% is held back by two deliberate drags, the shrinking Discover card book (down 2.9%) and a personal-loan book down 17%, which is why ex-Discover card loans growing about 6% is the better read on the underlying franchise.

Expense: The quarter's real surprise. Total non-interest expense fell 2% to $8.86B, with marketing down 4% to $1.59B and operating expense down 2%. Marketing is still up 14% year over year, so this is not a trough like Q1; it is an elevated level that dipped slightly while revenue grew. The adjusted efficiency ratio fell 262bps to 48.76%. Two caveats keep us honest. First, management itself called Q3 "a seasonally strong quarter in most years," and it has guided a Q4 marketing jump similar to last year's 38%. Second, against Q3 2025 the adjusted ratio is 80bps worse, because adjusted expense grew 7.2% while adjusted revenue grew 5.5%. Sequential operating leverage is real; year-over-year operating leverage is not there yet.

EPS: Higher quality than Q2's, with one flattering item. GAAP EPS of $5.61 carried $1.08 of adjusting items (acquisition amortization $0.60, Discover integration $0.42, Brex integration $0.06). The sequential gain from $5.81 to $6.69 adjusted came from pre-provision earnings and lower charge-offs, not from a bigger release, which is the opposite of Q2. Year over year, though, the 12% adjusted EPS growth leaned on a tax rate that fell to 22.6% from 27.1% and on a share count down 4.7%; pre-tax income grew only 4.3%. The buyback is a durable driver; the tax rate is not one to extrapolate.

Segment Performance

All three operating segments grew net income sequentially, and for the first time in several quarters the gains were led by revenue and cost rather than by the provision. Credit Card's provision was flat and its expenses fell; Consumer Banking kept compounding on network revenue and auto; Commercial posted a zero charge-off rate.

Segment Results — Q3 2026

Domestic Card Detail — The Engine

Credit Card — Earnings Up 19% on Flat Provision

Credit Card net income rose 19% to $2,859M, and unlike Q2 the provision did none of the work: it was $2,479M against $2,474M. The gain came from revenue up 3% and expenses down 4%. Domestic card is the whole story. Its revenue margin widened 41bps to 17.85% on a 50bps rise in loan yield, purchase volume grew 11% year over year, and charge-offs fell 73bps to 3.98%. Balances barely moved: period-end loans rose 0.5% sequentially and 2.5% year over year, because the legacy Discover book shrank about 3% while legacy Capital One grew about 6%. Year over year, domestic card expense grew 9% on technology, marketing and the addition of Brex, which management now books in this segment.

"Third quarter purchase volume grew 11% year-over-year, primarily driven by growth in legacy Capital One purchase volume. The addition of Brex and the corporate card business in the second quarter was a modest tailwind to purchase volume growth, while legacy Discover purchase volume was a modest headwind." — Richard Fairbank, Founder, Chairman & CEO

Assessment: The engine is earning more on roughly the same balances, which is what pricing power and credit discipline look like. The limiting factor is still growth: domestic card loans are up only 2.5% a year while the Discover book shrinks. That will not change until the back-book conversion completes and the restarted Discover marketing turns into balances, which management now places beyond "the next several quarters."

Consumer Banking — Network Revenue and Auto Keep Compounding

Consumer Banking net income rose 15% sequentially and 33% year over year to $560M. Revenue is up 17% year over year, with non-interest income up 62%, the clearest view of the debit-network synergy in the P&L. Global Payment Network volume reached about $191 billion, up 25% year over year. Auto is the other leg: loans grew 4% in the quarter and 13% over the year to $92.9B, and originations rose 22% year over year to $13.1B. Auto charge-offs ticked up to 1.59%, which management attributes to a subprime mix "returning to pre-pandemic levels." Deposits fell 1%, mostly from the sale of a small IRA portfolio acquired with Discover.

"Consumer banking revenue for the quarter was up about 17% year-over-year, driven predominantly by Discover revenue synergies and growth in auto loans." — Richard Fairbank, Founder, Chairman & CEO

Assessment: Consumer Banking is where the deal's revenue synergy is most visible and most durable. The auto growth is real but worth watching: a rising subprime mix is a deliberate return to the old book, not a deterioration, but it lifts loss rates and reserve needs as the book grows, which is why this segment built $41M of allowance while the card book released.

Commercial Banking — A Zero Loss Rate

Commercial Banking net income rose 31% to $313M on revenue up 6% and a provision down to $20M. The net charge-off rate was 0.00%, criticized performing loans fell to 4.40%, and criticized non-performing loans fell 10bps to 1.22%. Loans and deposits each grew about 1%.

Assessment: A clean quarter from a book that is small relative to the whole company. Commercial remains a source of stability rather than of thesis.

Key KPIs — Combined Company

Key Topics & Management Commentary

Overall Management Tone: Management was the most forthcoming it has been since the Discover deal closed. For five quarters it answered every earnings-power question with a claim of consistency; this time it volunteered a method investors can use to see the number themselves, and it framed the quarter as proof rather than promise. The confidence came paired with an unusually blunt investment message: more marketing in Q4 and 2027, a network investment "just kind of getting underway," and an AI agenda that management said it would be a disservice to hold back. Disclosure discipline did not change on the specifics: no efficiency target, no settlement estimate, no Brex P&L.

1. The Beat Moved Above the Line

The single most important fact about Q3 is that pre-provision earnings, the measure of the operating engine, grew 10%. In Q2 the same line was flat and the beat came from a reserve release.

"Linked-quarter revenue increased 3% and non-interest expense declined 2%, resulting in pre-provision earnings growth of 10%. On an adjusted basis, pre-provision earnings grew 9% quarter over quarter." — Andrew Young, Chief Financial Officer

Revenue growth came from margin (NIM +40bps) and modest loan growth; the expense decline was broad, with marketing, salaries, professional services and "other" all lower. Occupancy edged up 2%, and one line jumped: communications and data processing rose 27% to $622M, a move the call did not address.

Assessment: Setting aside Q3 2025, when the first full quarter of Discover lifted every line, this is the first quarter since the deal closed in which pre-provision earnings grew strongly without the help of a seasonal marketing trough. It is one quarter, and a seasonally strong one, so it does not settle the operating-leverage debate. But it does show the combined company can grow pre-provision earnings at a double-digit sequential rate while carrying Brex, Hopper and a full investment agenda, which is more than we could say in July.

2. "Squint Your Eyes": Earnings Power Comes Into View

For five quarters management told investors that post-integration earnings power was "consistent with what we expected at the time we announced the deal" and declined to say what that was. This quarter Fairbank gave the recipe.

"You can squint your eyes and see it now in the first 3 quarters of 2026. You would need to normalize for the sizable allowance release we've had thus far this year and add the increased marketing we expect in the fourth quarter. Looking forward, you would add in the remaining synergies we expect to get and probably deduct the lost interchange revenue that will likely come from the Visa, MasterCard settlement. Our current view is that the remaining synergies and the Visa, MasterCard settlement are roughly offsetting in magnitude." — Richard Fairbank, Founder, Chairman & CEO

We followed the instructions. The table below is our arithmetic, not company guidance.

Assessment: This is the biggest change in the investment case this quarter. Our earlier fair-value ranges leaned on a normalized figure near $24 that management never confirmed; its own recipe points lower, to roughly $21 a share on 2026's share count. A lower number that investors can verify is worth more than a higher number they have to take on faith, and the legibility discount was the main reason the stock traded at 8x. Two caveats matter. The recipe is a run-rate at today's level of investment, and management would not say whether the coming step-ups are inside it (see the Q&A). And "roughly offsetting" rests on a settlement estimate that management itself says has "a pretty wide error bar."

3. The Visa/Mastercard Settlement Enters the Math

The revised merchant settlement with Visa and Mastercard appeared on the call for the first time as an explicit input to Capital One's earnings power, and management said it is already planning for it.

"The hearing on final approval will take place in November. I know a lot of people in the investor marketplace or maybe even analysts are saying, 'Well, we'll wait to see what happens, and then sort of put it into our estimates.' We put a pretty high likelihood on the final approval of this, and so in our planning, we have, for some time, been incorporating our best estimate that we can. This has a pretty wide error bar." — Richard Fairbank, Founder, Chairman & CEO

Management sized the hit only by reference: about equal to the synergies still to come beyond the $1.8B annualized run-rate already achieved. Against the $2.5B combined target set out on earlier calls, that implies an interchange loss in the neighborhood of $0.7B a year. Management also said the settlement would not by itself trigger tactical changes to rewards, because the strategy already assumes it.

Assessment: A new, real headwind, and one that converts the remaining synergies from upside into an offset. It is worth noting that the Street may not have it in numbers yet; management pointedly contrasted its own planning with investors who are waiting for the ruling. We add it to the thesis as a new bear point. The political overlay, renewed talk of the Credit Card Competition Act, is a second interchange risk that management argued against but cannot control.

4. Marketing: A Q4 Step-Up Now, a Harder Lean in 2027

Management was unusually specific about the near-term expense path and unusually direct about the longer one.

"As we look ahead to the fourth quarter, we expect the Q3 to Q4 percentage growth in marketing to be similar to the Q3 to Q4 percentage growth we saw in 2025, primarily driven by seasonality." — Richard Fairbank, Founder, Chairman & CEO

"Looking ahead, we expect to lean harder into marketing in 2027, driven by several factors. We're fully pursuing our origination opportunities, and the window is open across our businesses. Our biggest competitors continue to invest heavily in marketing, and we need to be at the top of our game. Also, we plan to now lean back into Discover marketing." — Richard Fairbank, Founder, Chairman & CEO

Q4 2025 marketing was $1,934M, up 38% from Q3 2025's $1,403M. The same growth on Q3 2026's $1,594M points to roughly $2.2B in Q4, about $0.6B more than Q3. Fairbank also argued that marketing will structurally take a larger share of the cost base, because the fastest-growing businesses (heavy spenders, Capital One Shopping, the national bank) are marketing-intensive by design.

Assessment: Q4 will show a higher efficiency ratio, and that is guided, seasonal and already in our earnings-power math. The 2027 message is the bigger issue: "lean harder" with no size attached is the clearest statement yet that the reinvestment bear is not going away. Our upgrade does not assume operating leverage in 2027; it assumes the earnings power holds roughly at its current level while the buyback shrinks the share count.

5. Synergies Ahead of Schedule at $1.8B

For the first time management put a run-rate number on combined synergies achieved.

"On a combined basis, we're achieving the expected synergies ahead of schedule. Our year-to-date results reflect about $1.8 billion of annualized combined synergies, primarily driven by the successful debit conversion early in 2026 and the pace of achieving expense synergies." — Richard Fairbank, Founder, Chairman & CEO

Assessment: A credible data point that matches what the P&L shows: debit revenue synergies fully in Consumer Banking, expense synergies starting to bite as platforms convert. The remaining amount is no longer upside in management's own framework, because it is offset by the settlement. Discover integration expense was still $338M in the quarter, so the cost of getting there continues into 2027.

6. NIM at 8.41%: The Structural Effects "Have Run Their Course"

Net interest margin rose 40bps, ahead of the ~8.3% the Street expected, and the CFO said the Discover-driven structural changes are now complete.

"Looking ahead, something I noted last quarter, is that all of the structural impacts from the Discover acquisition have largely played out at this point in our NIM. From a rate perspective, our NII is relatively neutral to rates." — Andrew Young, Chief Financial Officer

Young pointed to two drivers from here: seasonality (day count, revolve rates, the holiday spending that inflates the Q4 denominator) and the relative growth of assets and liabilities. He suggested this quarter's average and ending cash, about $55 billion and just under $50 billion, bracket a typical level, and that the last rising-rate cycle is a reasonable starting assumption for deposit betas in this one.

Assessment: The margin is now a level, not a trajectory. That is good for legibility and neutral for growth: further NIM gains would have to come from mix, and the mix shift management is pursuing (heavy spenders, auto) is not obviously margin-accretive. We model seasonal movement around the low-8% range rather than further expansion.

7. Credit: Charge-Offs Fall on Recoveries, Delinquencies Turn Merely Seasonal

Domestic card charge-offs fell 73bps sequentially and 65bps year over year to 3.98%, and management gave a precise explanation that includes a part that will not last.

"Our recoveries inventories are higher than normal because of higher charge-offs in the rearview mirror. This is even more true at Discover, which had significantly higher charge-offs 2 years ago than they do today. But even when we adjust for this inventory effect, we see stronger recovery rates on recent charge-offs than what we've observed historically, and we think at least some of that's driven by sustainable improvements in our recovery strategies enabled by our transformation." — Richard Fairbank, Founder, Chairman & CEO

The forward-looking signal shifted slightly. Delinquencies rose 35bps in the quarter, which management called in line with seasonality, and Fairbank noted that delinquencies have been "moving with seasonality for a few months now as opposed to beating it." The domestic card release of $250M was "partially offset by greater consideration to downside economic scenarios related to geopolitical risks," a modest re-tightening of the overlay that eased in Q2.

Assessment: Credit remains a clear strength, but the source of improvement is changing. The long run of better-than-seasonal delinquencies is over, and part of the charge-off benefit is a recoveries inventory left over from higher past losses. Charge-offs can keep falling for a few quarters on the lag, but the tailwind is maturing. We stop counting on further reserve releases.

8. The Discover Brownout: Originations Turn, Balances Lag

Management met the commitment it made in July and gave the clearest view yet of how the Discover book returns to growth.

"As of the end of September, 100% of Discover Card originations are on our tech stack. We are in the middle of our conversion of the back book, and we will complete that in the first quarter." — Richard Fairbank, Founder, Chairman & CEO

"This quarter, new originations were up year-over-year for the first time since the deal closed. As the back book converts, we are resuming and leaning into customer management programs which have been limited during the conversion. All these moves, of course, take time to turn into loans. We expect continued contraction over the next several quarters, and then for loan growth to turn positive after that." — Richard Fairbank, Founder, Chairman & CEO

Discover card outstandings were down 2.9% year over year, deeper than the roughly 1.5% decline cited in July, as management had said they would be. The personal-loan book, which Fairbank described as a "magnified brownout" because of its short duration, fell 7% in the quarter.

Assessment: The leading indicator turned: originations are growing. The lagging indicator, balances, will keep shrinking "over the next several quarters," which pushes the visible inflection toward the second half of 2027. That is a little later than a reader of July's "somewhere around the fourth quarter" bottom might have hoped, but the two statements are compatible: the rate of contraction can bottom before balances grow.

9. The Network: Credit Volume Moves Gradually, Not All at Once

Management laid out three ways to grow credit volume on the Discover network and made clear it will not repeat the debit-style cutover.

"We will also build network volume by originating Capital One cards on the Discover Network and by converting Capital One customers from their existing network to the Discover Network. And we've been conducting tests to help inform our strategy regarding which customers and products we target for moving to the Discover Network and the timing of those moves… This is not going to be a big bang thing like you saw on the debit side. But we want to make very clear this is what we're working backwards from." — Richard Fairbank, Founder, Chairman & CEO

Assessment: The strategic intent is now explicit: Capital One is working toward moving its own credit volume onto the network it owns. The pace will be set by global acceptance and the Discover brand, both of which require investment. That makes the network the largest long-run option in the stock and also a named source of the investment that management says is "just kind of getting underway."

10. Brex: "Add Water," Still No P&L

Fairbank gave one of the call's longest answers on Brex's technology and the integration plan, which he described as a targeted first phase that lets Brex keep growing.

"Brex is already experiencing some of the early tailwinds that come with our brand. We expect these to only grow stronger over the next few months. They're benefiting from the cost of funds, from moving to our balance sheet. We have already stood up a program to share high potential leads from across our businesses with Brex and are seeing very promising results at the outset." — Richard Fairbank, Founder, Chairman & CEO

The fuller payoffs (Capital One's marketing machine, spend management for small-business cardholders, the travel portal) require more integration, and bringing the travel portal to Brex is "probably a next year kind of thing." Brex integration expense fell to $47M from $96M.

Assessment: Brex is now folded into Domestic Card and remains unquantified. With the earnings-power recipe in hand, the Brex opacity matters less than it did: whatever Brex earns or loses is inside the nine-month numbers the recipe starts from. It remains an option we keep out of the base case.

11. Capital: $2.6B of Buybacks, CET1 Flat, Tangible Book Down

Capital return held its pace and the ratio held its level.

"Our common equity Tier 1 capital ratio ended the quarter at 13.8%, roughly flat to the second quarter, as net income in the quarter was largely offset by share repurchases of $2.6 billion, dividends, and growth." — Andrew Young, Chief Financial Officer

Diluted shares fell 1.9% in the quarter and 4.7% over the year, and period-end shares fell 5.4% over the year to 601.5M, net of the stock issued for Brex. The company also redeemed about $1B of preferred stock. Tangible book value per share fell 2.4% to $102.72, because the accumulated other comprehensive loss widened by $3.7B as the market value of the securities portfolio fell about $3 billion.

Assessment: The buyback is the most dependable driver of per-share growth while loans grow slowly. The net share count fell about 5% over the past year, and at $2.6B a quarter and a price near $201 the gross pace is closer to 2% a quarter. We use the lower, 5% net figure in our valuation. CET1 at 13.8% still sits well above the 11% management has called its long-term need. The AOCI hit is a mark, not a loss, but it is a reminder that tangible book is rate-sensitive in a hiking cycle.

Guidance & Outlook

Capital One again gave no formal guidance, but Q3's call was more useful than any since the deal closed. It supplied a reproducible earnings-power method, a specific Q4 marketing signpost, a statement that NIM's structural changes are done, and a Discover loan timeline. What it withheld is just as clear: the size of the 2027 marketing step-up, an efficiency target, and a settlement estimate.

Implied Q4: With marketing up about $0.6B and seasonally higher charge-offs, Q4 adjusted EPS should land well below Q3's $6.69, most likely in the mid-$4s to around $5 by our estimate, against Q4 2025's $3.86. That would put full-year 2026 adjusted EPS near $21–22, on top of the $16.88 earned through September. Q4's adjusted efficiency ratio will rise from 48.76%; the comparison that matters is Q4 2025's 53.73%.

Street at: Pre-print estimates we could find put 2026 near $19.60–19.70 and 2027 at about $23.40–24.50, both stale. The nine-month result already puts 2026 ahead of those figures. 2027 is the risk: the marketing lean-in and the settlement are both reasons to expect 2027 estimates to come under some pressure, and our earnings-power figure of ~$21 sits below the Street's 2027 even before the buyback adds its ~5%.

Guidance style: Still no numbers, but a meaningful change in posture, from asking investors to trust the outcome to showing them how to check it.

Analyst Q&A Highlights

Measuring the Return on Marketing Ahead of Loan Growth

The first question went straight to the marketing step-up: how management measures returns when balances lag the spend by quarters, and whether it will keep spending before loan growth shows. Management defended its "horizontal" measurement approach and, picking up the question's premise, conceded that marketing has grown faster than revenue for years.

Q: "As we think about it, how do you measure the return on the investment for those marketing dollars? Maybe, for example, the dollars allocated to the Discover brownout, since there is a lag between deployment of those dollars to account generation to balance growth. Should we expect you to keep leaning into marketing ahead of the loan growth materializing?"
— Terry Ma, Barclays

A: "It's striking to notice that the growth rate of marketing over the last number of years has been higher than the growth rate of revenue, the growth rate of other expenses… My point was there is a gradual mix change going on where marketing is a higher component of our growth businesses and therefore, in some sense, becoming a higher component of our business. But through it all, we continue to be as bullish about marketing and the return on our marketing as we have been really at any time in this journey."
— Richard Fairbank, Founder, Chairman & CEO

Assessment: An honest answer that confirms the bear's premise and disputes its conclusion. Marketing is structurally becoming a bigger share of costs, by design. The defense is that the businesses it builds carry lower risk and lower capital, so the efficiency ratio is the wrong yardstick. That argument works only if returns on equity hold, which is why ROTCE, not the efficiency ratio, is the metric we now anchor to.

What the Earnings-Power Statement Is Meant to Say

An analyst asked directly whether the earnings-power remarks meant returns above 20%, and whether marketing would eat into them. Management declined the number but restated the point with more conviction than in any prior quarter.

Q: "Are you saying that I tried to do math quickly, but returns should be at 20% plus, or is there something else underlying the statement?… should we expect that you'll be maintaining these returns, or do you think marketing will eat into some of them?"
— Ryan Nash, Goldman Sachs

A: "My point was with a couple of approximate adjustments, we can actually just see it right now. We don't have to hold our breath and wait for it. It's really, we can see it right now, and we're not in the precise number business."
— Richard Fairbank, Founder, Chairman & CEO

Assessment: Management would not say "20%," but it did not need to. The recipe it offered produces roughly 19–20% at today's capital and in the low 20s at the 12.5% CET1 denominator it used at deal announcement. For the first time the earnings-power claim can be checked against reported results, and it checks out.

Whether Earnings Power Already Includes the Coming Investment

The most important question of the call came twice, once mid-call and once as the final question: is the earnings power management described net of the investment step-ups it is also announcing, or will that spending come on top? Management did not answer it directly.

Q: "When you talk about the earnings power that you had projected at the deal, does that include those step-ups in marketing technology spending, or does that spending power come on top of the model, if you will? Just trying to understand if the earnings power already incorporates some of this investment spending you are talking about."
— Multiple analysts incl. Mihir Bhatia, Bank of America; John Pancari, Evercore

A: "We have already been leaning into investments significantly in our company. We have talked a lot about that. A lot is already in the numbers that you see. We are continuing to lean in as well. Some things that are probably going to see more in the future than they have are investment in the Discover Network, which is really just kind of getting underway… We are not generally going to be in the business of being in the year after year giving guidance on the earnings power of the company."
— Richard Fairbank, Founder, Chairman & CEO

Assessment: The non-answer is the answer: some of the investment is in the numbers and some is not, and management will not say how much. This is the main reason our upgrade comes with moderate rather than high conviction. We treat the ~$21 figure as a level that ongoing investment holds roughly flat, not as a floor that 2027 growth builds on.

The Direction of the Efficiency Ratio

Asked whether the efficiency ratio should improve from here, management warned against reading any single quarter and repeated that it will not guide the metric.

Q: "When we think about sort of the direction of travel on the efficiency ratio from here, how should we think about it? Should it be improving from here on out?"
— Sanjay Sakhrani, KBW

A: "There's a lot of variability quarter to quarter in efficiency ratio, so of course, I wouldn't anchor to any one quarter's results. The efficiency ratio will continue to reflect the realization of synergies and our investment imperatives… We're not giving specific guidance on efficiency ratio, and we've been more focused in our dialogue on earnings power."
— Richard Fairbank, Founder, Chairman & CEO

Assessment: Management itself told investors not to extrapolate Q3's 48.76%, which is fair given the Q4 marketing guide. We take the hint: the operating-leverage evidence we credit is the 10% growth in pre-provision earnings, not the ratio's quarterly level.

How Much Further Card Credit Can Improve

An analyst asked what is driving the card credit improvement and whether it can continue. Management credited recoveries, separated a temporary inventory effect from a lasting one, and signaled the delinquency trend has flattened to seasonal.

Q: "Just maybe unpack for us what is driving the improvement in card. As you look forward, do you think that we can continue to see this type of improvement given the dynamics you are seeing in the portfolio?"
— Ryan Nash, Goldman Sachs

A: "In the last few months, I think what we see is more in line with seasonality rather than outperforming seasonality. But we are in a very good place with respect to credit. Now, if you think about charge-offs, there are a couple things to say there. First of all, charge-offs are lagged relative to delinquencies. The charge-offs continue to improve as the delinquencies have improved."
— Richard Fairbank, Founder, Chairman & CEO

Assessment: Unusually candid. Management did not promise more improvement; it described a credit cycle that has stopped getting better at the front end while charge-offs catch up at the back. That supports a strong 2026 loss rate and argues against counting on more releases in 2027.

When Discover Card Balances Grow Again

An analyst asked how long until the Discover book returns to positive loan growth now that marketing and originations are restarting. Management gave the first concrete leading indicator and a timeline measured in "several quarters."

Q: "Then just also on the topic of Discover, just as you were restarting marketing and originations around it, do you have a sense on the timing before we can see Discover return to positive loan growth? Any way to think about the economics of new balances there versus prior?"
— Darrin Peller, Wolfe Research

A: "In the third quarter, Discover Card outstandings were down 2.9% year-over-year. By the way, it's worth noting the flip side of these pullbacks and the brownout has been strong credit performance… This quarter, new originations were up year-over-year for the first time since the deal closed… We expect continued contraction over the next several quarters, and then for loan growth to turn positive after that."
— Richard Fairbank, Founder, Chairman & CEO

Assessment: The originations turn is the first hard evidence that the brownout is ending. The balance-sheet payoff is still a 2027 event, and the question about the economics of new balances went unanswered.

The Pace of Moving Capital One's Own Cards to the Network

An analyst asked how fast Capital One credit volume will move to the Discover network. Management described a deliberate, product-by-product migration rather than a cutover.

Q: "Separately, just around the network, can you just update us on the progress of moving some of the Capital One credit card volumes onto the Discover Network and any of the pilots that you are running? How should we think about the pace of that migration and the timeframe involved?"
— John Pancari, Evercore

A: "In parallel, we will lean hard into enhancing global network acceptance and building the Discover brand. And along the way, we will slope the work as we drive acceptance and slope the work in selecting which products and customers are moved onto the network. This is not going to be a big bang thing like you saw on the debit side."
— Richard Fairbank, Founder, Chairman & CEO

Assessment: No timeline, by choice. The network option is real but will be paid for over years of acceptance and brand investment before it pays back, which is consistent with keeping it out of our base case.

Resilience Through the Next Credit Turn

An analyst asked whether the combined company is more resilient than the Capital One that traded down to tangible book in past downturns. Management listed the improvements (a spender-led mix, fewer high-balance revolvers, real-time monitoring) and then volunteered a point that cuts the other way.

Q: "While the Capital One in the past has historically gotten down to tangible book value and downturns, it would be helpful to hear how you think the resilience of the business has changed, and whether you believe the business is in a superior position today to weather the next turn in credit, whenever that comes."
— Bill Carcache, Piper Sandler

A: "I do want to say that one thing that was very wind-aided in the global financial crisis that we should point out, and that is that we had a significant amount of our balance sheet of our loans off balance sheet. Also it was not a CECL regime. So when the sort of the bad times came, our allowance build was way less than it would be pound for pound now with everything on balance sheet and everything with CECL and all that."
— Richard Fairbank, Founder, Chairman & CEO

Assessment: A useful reminder that the next downturn will hit reported earnings and capital faster than the last one did, even if the underlying book is better. It argues for the conservative capital posture management keeps, and it is why we value the stock off earnings power and buyback capacity rather than assuming the current credit tailwind persists.

What They're NOT Saying

1. How big the 2027 marketing step-up is. Management said it will "lean harder into marketing in 2027" and restart Discover marketing, but gave no amount or growth rate. That is the single most important unknown for 2027 estimates.
2. Whether earnings power is net of the coming investment. Asked twice, management said "a lot is already in the numbers" and that more is coming, including network investment "just kind of getting underway," without saying which.
3. The settlement number. Management is planning for the Visa/Mastercard settlement and called its estimate wide, but sized it only as "roughly offsetting" the remaining synergies. The $2.5B synergy target was not restated on the call; only the $1.8B achieved was.
4. Brex's economics, again. Brex now sits inside Domestic Card. The only Brex figure in the release is $47M of integration expense; revenue, loans and contribution remain undisclosed two quarters after the close.
5. The tangible-book hit. The accumulated other comprehensive loss widened by $3.7B and tangible book per share fell 2.4%. Management named neither on the call; the CFO noted only that the securities portfolio's market value fell about $3 billion. Nor did the $1B preferred redemption, the first net securities loss in the five quarters shown ($40M), or a 27% jump in communications and data-processing expense.
6. Capital return from here. No analyst asked about capital and management offered no view on buyback pace or the path toward its 11% long-term need, a notable silence with CET1 at 13.8%.
7. What AI is worth. Asked whether AI investments are producing revenue or share gains, Fairbank said "we certainly do not intend to quantify the benefits," while describing AI spending as an imperative.

Market Reaction

• Pre-print setup: COF closed at $201.56 on October 20, down 16.8% year to date against the S&P 500 up 12.7%, down 6.2% over the trailing twelve months, and down 0.4% over the trailing 30 days. The 52-week closing range was $176.10 to $257.94.
• After-hours move: Roughly +4%, with shares near $209 after the release and call.
• Reaction session (October 21): COF opened at $206.37 (+2.4%), traded as high as $211.49 (+4.9%), and closed at $206.66, up 2.5%, on 6.6M shares, 1.8x its 30-day average. The S&P 500 rose 0.6%. Card peers rose less: Synchrony +1.0%, American Express +1.7%.
• Follow-through: The gain did not hold. COF fell 1.6% to $203.43 on October 22 and 1.0% to $201.34 on October 23, essentially back to the pre-print close (−0.1%), while the S&P 500 fell 1.2% over the same span.

The tape told two stories in three days. The first-day move was a clean, if modest, endorsement of a better-quality beat: COF beat the index by about two points and its two closest card peers by 0.8 to 1.5 points on elevated volume, and the intraday high of nearly +5% shows buyers were willing to pay for it early. By the close, about half of that intraday gain had gone, and the next two sessions erased the rest as the broad market sold off.

We read the fade as skepticism, not rejection. The bear argument circulating after the call was not about the quarter; it was about the next one and the year after: a guided Q4 marketing jump, a harder 2027 lean, and a settlement that takes interchange away. A stock that is down 17% in a year when the index is up 13% has been pricing exactly that kind of open-ended investment story. What changed is that management gave investors a way to see the earnings power that investment is supposed to protect, and the stock is no more expensive than it was before that disclosure. Over the three sessions COF still held up about a point better than the S&P 500, which is the right direction but nowhere near the re-rating the quarter earned in our view.

Street Perspective

Debate: Is the Earnings Power a Floor or a Ceiling?

Bull view: Management has now shown a ~20% return on tangible equity with the investment program running, and it told investors they can "reach out and touch" the deal's earnings power. Investment is already inside the run-rate, buybacks compound it, and the market is still pricing the stock as if the deal were unproven.

Bear view: The number is a ceiling. Management announced more marketing for Q4 and 2027, a network investment that is only starting, and an AI agenda it will not hold back, then refused to say whether any of it is inside the earnings-power figure. Add the settlement, and 2027 estimates are more likely to fall than rise.

Our take: Closer to a level than either side admits. We do not assume growth on top of ~$21, and we accept that 2027 estimates may drift lower. But a ~20% return franchise that retires about 5% of its shares a year grows per-share earnings even if the base stays flat, and at 9.6x that base, the valuation already assumes something worse than flat.

Debate: Operating Leverage or Q3 Seasonality?

Bull view: Revenue up 3% with expenses down 2% and pre-provision earnings up 10% is the operating leverage the bears said would never show. Synergies are running ahead of schedule, and the expense base absorbed Brex and Hopper without growing.

Bear view: Q3 is a seasonally strong quarter, marketing dipped before a guided 38% Q4 jump, and on a year-over-year basis the adjusted efficiency ratio is 80bps worse. One good sequential quarter in a seasonal high is not a trend.

Our take: The bears have the year-over-year math right, and Q4 will look worse. But the bull point that matters is the composition: for the first time since the first full Discover quarter, pre-provision earnings grew strongly without the help of a marketing trough, and the EPS gain did not depend on a large reserve release. We are not upgrading on the efficiency ratio; we are upgrading on earnings power that is now visible and on a price that ignores it. Q4 2026 against Q4 2025 is the fair test, and we will hold ourselves to it.

Debate: Does the Valuation Discount Still Fit?

Bull view: At about 9.6x a verifiable earnings power and 2.0x tangible book, the market is valuing a 20% return on equity, the largest U.S. card book and an owned network like an average lender. The legibility discount should close now that management has shown the math.

Bear view: Cheap card lenders stay cheap. Consumer credit is late-cycle, delinquencies have stopped improving, CECL makes the next downturn hit faster, and the political and legal pressure on interchange is rising. The multiple is low for reasons that did not change this quarter.

Our take: The bears describe real risks, but most of them were just as present at the 52-week closing high of $257.94, and the stock is now more than 20% below that high with higher earnings, better credit and a clearer earnings base. We do not need the multiple to reach card-network levels. On the ~$22 figure for a 2027 share count the stock trades near 9.2x, and a move to 10–11x that, plus the buyback and dividend, is enough to beat the market.

Model & Valuation Framework

Valuation framework: We now value Capital One on the earnings power its own management has shown investors how to compute, and we deliberately use our lower reconstruction rather than the Street's 2027 figures. Our earlier ranges ($205–245 in July) applied 8.5–10x to a normalized base near $24 that management never confirmed. Management's recipe points to about $21 on 2026's share count; holding that level flat and letting the buyback retire roughly 5% of shares gets to about $22 on a 2027 share count. The October 23 close is about 9.2x that $22 (and 9.6x the $21 on 2026's count). At 10–11.4x, which we think a verifiable ~20% return on tangible equity deserves once the legibility discount closes, the $22 supports $220–250. The midpoint of $235 is about 17% above the October 23 close of $201.34, and the $3.20 annual dividend adds about 1.6%. Tangible book of $102.72 frames the downside: at 2.0x book today, a return to the 52-week low of $176.10 would mean about 1.7x for a business earning ~20% on that book, a level we think the buyback would defend. We keep Brex, the credit-volume network migration and any operating leverage in 2027 out of the base case; they are upside, not assumptions.

Thesis Scorecard Post-Earnings

We grade the standing thesis, the same eight pillars carried into this quarter (seven from initiation plus Bear-4, added in Q4 2025) and one new bear point added this quarter, against the print and the call.

Overall: The thesis strengthens, and the two pillars that gated an upgrade both moved. Bear-1 eased from Materializing to Emerging because pre-provision earnings grew strongly without the help of a seasonal marketing trough, even though the year-over-year comparison and the 2027 marketing guide keep it alive. Bear-2 eased from Emerging to Contained because management gave investors a reproducible way to see the earnings power, and that earnings power checks out near a 20% return. Bull-1 delivered its commitment, Bull-3 stayed strong but is maturing, and Bull-4 held its pace. Against that, Bear-4 escalated in rhetoric and a new Bear-5 enters on the settlement.

Action: Upgrade to Outperform, conviction 6. Earnings power is now visible at roughly $21 a share, the stock sits at about 9.6x that and 2.0x tangible book after giving back its post-print gain, and the buyback retires about 5% of the share count a year. We would revisit the rating if Q4 2026's adjusted efficiency ratio comes in worse than Q4 2025's 53.73%, if the 2027 marketing step-up proves large enough to push normalized earnings power clearly below $20, if the settlement's interchange hit proves materially larger than the remaining synergies, or if credit turns.

Bottom Line: The Number We Were Waiting For

Rating decision: We upgrade Capital One to Outperform from Hold, with a fair-value range of $220–250. In July we said we wanted one more quarter of the efficiency ratio holding and, ideally, some Brex legibility before paying up, and we listed a quantified earnings-power figure as an upgrade trigger. Q3 delivered the efficiency ratio and the earnings power, but not Brex: the adjusted efficiency ratio fell 262bps as pre-provision earnings grew 10% on lower expenses (though marketing dipped 4% in a seasonally strong quarter), and management laid out a recipe for earnings power that we reconstruct at about $21 a share, or a 19–20% return on tangible equity. ROTCE printed 21%.

The upgrade is moderate in conviction because the same call that showed the earnings power also promised more spending against it. Q4 marketing will rise about 38%, 2027 marketing will rise further, network investment is "just kind of getting underway," and management would not say whether any of that is already inside the earnings-power figure. The settlement will take interchange that the remaining synergies are meant to replace. We are therefore not paying for growth on top of the ~$21. We are paying for the level, at a price that has not moved since before management showed it.

Why now: Capital One closed at $201.34 on October 23, below where it traded before the print and below its July level, despite the best quarter of the combined company and the first real disclosure on earnings power. At 9.6x that earnings power and 2.0x tangible book for a franchise earning ~20% on equity, with the largest U.S. card book, an owned network and a buyback that retires ~5% of shares a year, the risk/reward now favors owning it against the S&P 500 over the next twelve months.

What would change our mind:

• Back to Hold: Q4 2026's adjusted efficiency ratio worse than Q4 2025's 53.73%; a 2027 marketing step-up large enough to push normalized earnings power clearly below $20 a share; a settlement whose interchange cost materially exceeds the remaining synergies; or a slowdown in the buyback without a clear use for the capital.
• Higher conviction: Q4 holding near or better than last year's efficiency ratio despite the marketing step-up; a sized 2027 investment plan that leaves earnings power intact; the Discover book returning to growth ahead of the "several quarters" timeline; or any Brex disclosure that shows it adding to, rather than diluting, the base.

Signposts for Q4 2026 (January): the Q4 adjusted efficiency ratio against Q4 2025's 53.73% and whether marketing lands near the ~$2.2B implied by the guide; any sizing of the 2027 marketing and network investment; the November final-approval ruling on the Visa/Mastercard settlement and any update to management's estimate; the next back-book conversion wave and Discover originations growth; whether delinquencies stay seasonal or start rising faster; and the buyback pace with CET1 at 13.8%.