Credit Powers a 24% EPS Beat and ROTCE Snaps Back to 18%, but the Operating-Leverage Question and Brex Stay Unanswered — Maintaining Hold

Results vs. Consensus

Q2 is the strongest headline print of the four quarters we have covered since initiation, and on the surface it is exactly the quarter the bull case needed: a large earnings beat, an 18% return, returning loan growth, and improving credit. The work of the analysis is separating what is durable from what is not, because the composition of the beat matters more than its size. The bulk of the upside came below the pre-provision line, from a provision that was about a billion dollars lighter than the Street modeled. Above that line, the business grew revenue 4% and grew expenses 7%, so the earnings engine was flat sequentially even as the reported number jumped.

Q2 2026 Scorecard

Year-Over-Year Comparison

Quarter-Over-Quarter Comparison

Quality of the Quarter

Revenue: Better than it has looked in two quarters, and better composed. Total net revenue rose 4% sequentially, split between net interest income up 2% and non-interest income up 13%, the latter helped by the full run-rate of debit network revenue synergies now flowing through Consumer Banking. The revenue margin recovered to 10.27% from 9.87%, and NIM rose 14bps to 8.01%. Underneath the reported figures, ex-Discover revenue grew about 9.5% year over year, so the organic franchise is not just holding but accelerating modestly. The top line has clearly stopped going the wrong way, which was the new worry we flagged in April.

Expense: This is the part the headline hides. Total non-interest expense rose 7% sequentially, with marketing up 11% to $1.66B and operating expense up 6%. That is management doing exactly what it told us it would do in April: leaning marketing back in after a seasonally light Q1, and layering Brex and the Hopper travel infrastructure into the base. The adjusted efficiency ratio rose 167bps to 51.38%, confirming that Q1's sub-50% reading was a trough, not a new run-rate. Only about a third of the announced operating-expense synergies have been realized, with the balance not landing until the second half of 2027. The operating-leverage story remains a 2027 story.

EPS: The $5.81 is a real number and a strong one, but its quality is credit, not operations. GAAP EPS of $4.73 carried $1.08 of adjusting items (acquisition amortization $0.60, Discover integration $0.36, Brex integration $0.12). Above that, the swing from Q1's $4.42 adjusted to $5.81 was driven overwhelmingly by the provision falling from $4.07B to $2.99B, including a reserve release rather than a build. Normalize the provision to a mid-cycle charge-off run-rate with no release, and underlying earnings power sits meaningfully below the reported figure. The reported ROTCE of 18.04% is genuine and reassuring about the earnings base; the sequential jump in EPS is not repeatable.

Segment Performance

Every consumer-facing segment posted a strong sequential earnings recovery, driven by the lighter provision and, in Consumer Banking, by a genuine revenue step-up as network synergies and auto growth compounded.

Credit Card — Earnings and Balances Both Grow

Credit Card net income rose 29% sequentially to $2,411M, and this quarter, unlike Q1, the balance sheet cooperated: segment loans grew, with domestic card period-end loans up 2% to $259.0B. Domestic card is the engine, and it did nearly everything well: revenue up 4% to $11,103M, the revenue margin up 55bps to 17.44%, purchase volume up 15% sequentially and 26% year over year to $249.2B, and charge-offs down 39bps to 4.71%. The one persistent drag is the acquired book. Legacy Discover card loans shrank about 1.5% year over year in the deliberate "brownout," and management now expects that contraction to deepen before it bottoms around Q4 2026.

"If I separate out the Discover brownout effect, Capital One continues to deliver very consistently solid loan growth… when we look at the originated upmarket part of Capital One… this thing is absolutely humming and is right up there at the top of the league tables in the key growth metrics." — Richard Fairbank, Founder, Chairman & CEO

Assessment: The segment did what the current thesis needs it to start doing: grow earnings and balances at the same time, with credit improving rather than deteriorating. The legacy Capital One book is accelerating and the acquired Discover book is a temporary, self-imposed drag with a defined bottom. The gap between "domestic card grew 2%" and "the growth inflection is a 2027 event" is the brownout, and it is closing on schedule but not yet closed.

Consumer Banking — The Quarter's Quiet Star

Consumer Banking was the standout, with net income up 63% sequentially to $486M on revenue up 10% to $3,209M. The revenue jump is the network synergy showing up in the numbers: with the debit-to-Discover conversion complete, the full quarterly run-rate of network revenue now flows here, and non-interest income rose 14% sequentially and 97% year over year. Auto is compounding underneath, with originations up 19% year over year to $12.9B and the auto charge-off rate improving 22bps sequentially. Deposits grew 2% on average with the average deposit rate falling 8bps to 2.76%, a clean funding-cost tailwind.

"Network transaction volume increased 156% compared to the partial quarter… in the second quarter of 2025 and the successful completion of… the conversion of Capital One debit customers to the Discover Network." — Richard Fairbank, Founder, Chairman & CEO

Assessment: This is the vertical-integration thesis paying its first visible dividend in the P&L. The network is no longer a slide about strategy; it is a double-digit revenue tailwind to the segment that houses it, layered on a genuinely growing deposit and auto franchise at improving funding costs. Consumer Banking is quietly doing the growth work while the card book waits out the brownout.

Commercial Banking — Small, Stable, Slightly Lower Revenue

Commercial Banking net income rose 16% sequentially to $239M as the prior quarter's idiosyncratic reserve build did not repeat and the segment released a small $59M of allowance tied to loans charged off in the quarter. Revenue fell about 7% to $850M and loans grew 1%. The net charge-off rate rose 24bps to 0.53%, consistent with the lumpiness of commercial credit rather than any trend, and the criticized-loan rate actually improved.

Assessment: Immaterial to the earnings thesis and behaving normally. The Q1 reserve concern proved idiosyncratic, as management said it would. Commercial remains a stable, deliberately unremarkable book.

Key KPIs — Combined Company

Key Topics & Management Commentary

Overall Management Tone: Management was the most confident it has sounded since the Q3 2025 clean-quarter call, leaning expansively into a long list of growth investments while returning again and again to a single anchor: that post-integration "earnings power" remains consistent with the deal-announcement expectation, now inclusive of Brex and the insourced travel technology. The posture on disclosure did not soften a bit. Pressed directly on the earnings-power level and on Brex's economics, management held the line, declining to quantify either and redirecting to ROTCE as the summary metric. Fairbank was notably philosophical, spending real airtime on the company's "horizontal accounting" and technology transformation, the register of a founder defending a decades-long compounding story rather than a manager defending a quarter.

1. The Anatomy of the Beat: Credit Did the Work

The single most important fact about the quarter is where the earnings came from. Revenue grew 4%, expenses grew 7%, and the provision fell 27%. The earnings beat lives almost entirely in that last number.

"Relative to the first quarter, revenue increased 4% and noninterest expense grew 7%, resulting in pre-provision earnings growth of 1%. On an adjusted basis, pre-provision earnings were flat quarter over quarter. Our provision for credit losses decreased $1.1 billion, or 27%, to $3 billion in the quarter." — Andrew Young, Chief Financial Officer

Assessment: Flat adjusted pre-provision earnings is the tell. The operating engine did not accelerate this quarter; the credit line got better. Both are good, but only one of them is the thing the bull case is waiting on, and it is the one that did not move. This is why we treat a 24% EPS beat as a Hold-confirming quarter rather than an upgrade trigger.

2. The Reserve Release and a Lifting Macro Cloud

After two consecutive precautionary builds, Capital One released reserves, and the reasoning matters: it was driven by observed credit strength and a reduction in the macro overlay that management itself added in April.

"The $662 million allowance release in the quarter brought the allowance balance to $23 billion. Our total portfolio coverage ratio decreased 26 basis points and now stands at 5.02%… The decline in the coverage ratio was driven by continued favorable observed credit in the quarter and a modest decrease in the consideration given to economic uncertainties." — Andrew Young, Chief Financial Officer

Assessment: This is a real positive and a clean reversal of the Q1 story. The Persian Gulf overhang that justified the second straight build has eased enough for management to give some of it back. Coverage at 5.02% is still conservative. The release flatters this quarter's EPS and is not repeatable, but the underlying signal, that management sees less downside risk than it did ninety days ago, is a genuine improvement in the credit bear.

3. Loan Growth Turns Positive; the Brownout Bottoms in Q4

Period-end loans grew 2% sequentially, reversing Q1's decline, and management gave the clearest brownout timeline yet.

"So that brownout will continue and, in fact, increase. The bottom of the brownout will be somewhere around the fourth quarter of this year. But then we look forward to leaning into that growth over time… the front book is a much easier thing." — Richard Fairbank, Founder, Chairman & CEO

Assessment: Two things are true at once. The reported balance sheet has stopped shrinking, helped by the legacy Capital One book accelerating (ex-Discover loans up about 5.3% year over year). But the acquired Discover book gets worse before it gets better, bottoming around Q4 2026, with the growth payoff still gated on the back-book conversion that completes in Q1 2027. The trajectory improved this quarter; the inflection is still not here.

4. Marketing Steps Up and the Efficiency Ratio Rises, As Promised

The seasonal trough we flagged in April reversed on cue. Marketing rose 11% sequentially and the efficiency ratio moved up with it.

"Total company marketing expense in the quarter was about $1.7 billion, up 23% year over year… As we've discussed, these investments will continue to be reflected in the efficiency ratio. They are also the engine that powers long-term growth and returns." — Richard Fairbank, Founder, Chairman & CEO

Assessment: This is the confirmation, not the resolution, of the operating-leverage debate. Q1's sub-50% adjusted efficiency ratio was indeed a trough; at 51.38% the ratio has stepped back up, and with two-thirds of the opex synergies still to come and investment "reflected in the efficiency ratio," the path to demonstrated operating leverage runs through 2027. Nothing here refutes the reinvestment bear; it plays out exactly as management guided.

5. NIM Recovers to 8.01%, With More to Come

Net interest margin rose 14bps, and the CFO laid out a mechanical path higher.

"There should be a bit of a NIM catch-up that happens in the third quarter as the average cash catches up to the ending cash. And then also… in the back half of the year we have one more day in each of the quarters, so that adds a nine basis point tailwind to NIM." — Andrew Young, Chief Financial Officer

Assessment: Straightforwardly positive and largely mechanical. The elevated cash that dragged Q1 came down, funding costs improved, and the day-count math adds to margin in Q3 and Q4. The post-Discover structural NIM level looks intact, which supports the "stabilizing returns" read and is one of the cleaner marks in the quarter.

6. Brex: 100 Days In, Still a Black Box

Q2 was billed as the quarter Brex's financials would come into view. They did not. Management provided purchase-accounting detail in the appendix but declined any P&L breakout.

"We didn't provide the breakdown of the specific amount of Brex… given the relatively small percentage of Brex within the context of Capital One, the P&L and balance sheet on a run-rate basis just aren't that material. And that said, we're incredibly excited about the long-term prospects… think that the growth that this platform provides will drive significant accretion." — Andrew Young, Chief Financial Officer

Assessment: The one disclosure we and the Street were told to expect this quarter did not arrive. "Not that material" is a fair characterization of Brex's current scale, but it is also a convenient reason not to quantify a deal that management concedes is dilutive now with the payoff deferred. The accretion-legibility and serial-acquirer bears stay alive precisely because the information event keeps sliding.

7. Earnings Power Reaffirmed, Level Still Withheld

The recurring anchor of the call is that combined-company earnings power on the other side of the integration matches the deal-announcement expectation, now inclusive of Brex and Hopper. What that number is remains undisclosed.

"We still expect our earnings power on the other side of the Discover integration to be consistent with what we expected at the time we announced the deal… So implicit in that, there needs to be an efficiency ratio that makes the numbers work. We're not specifically guiding on that." — Richard Fairbank, Founder, Chairman & CEO

Assessment: Reassuring in direction, unquantified in level, for the fifth straight quarter. The reported 18% ROTCE is a strong supporting data point, and the reaffirmation is credible. But management has now been asked in four different ways across two calls to put a number on earnings power, and it will only offer a metric (ROTCE at a 12.5% CET1) and a claim of consistency, never the level itself. The market is asked to take the payoff on faith.

8. The Network: Debit Done, Now Testing Credit-Volume Migration

The strategic centerpiece advanced on two fronts. The debit conversion is complete and fully in the run-rate, and management is now testing something new: routing legacy Capital One credit volume onto the owned Discover network.

"Earlier this year, we completed the conversion of our debit card business to the Discover Network… a smashing success… So now as we think about building credit card volume on the Discover Network, there's two ways to do that, with the front book and the back book." — Richard Fairbank, Founder, Chairman & CEO

Assessment: This is a genuinely new leg to the vertical-integration story that did not exist as a concrete program a quarter ago. Beyond migrating Discover cards onto Capital One's tech, the company is now testing moving its own credit volume onto the network it owns, which captures interchange economics it currently pays away. It is early, gated on international acceptance, and years from full scale, but it widens the long-run optionality even as it adds to the near-term investment bill.

Guidance & Outlook

Capital One again declined formal guidance, but the call sharpened the near-term shape: recovering NIM, a deepening-then-bottoming loan brownout, marketing that stays elevated into the back half, back-loaded opex synergies, and an earnings-power claim that stays qualitative. Notably, this quarter graded well against the checklist we set in April on almost every item except the one we most wanted: Brex legibility.

Implied trajectory: The near-term path is one of firming returns on a still-flat operating engine. NIM rises mechanically into the back half, credit is a tailwind rather than a drag, and the buyback keeps shrinking the count, but marketing stays high, Brex and Hopper sit in the base, and loan balances only truly inflect after the Q1 2027 back-book conversion. The result is a stronger, more legible earnings base (an ~18% ROTCE, a reaffirmed earnings-power anchor) without the revenue-growth-plus-operating-leverage combination that an upgrade requires. We expect the Street to lift 2026 adjusted-EPS estimates on the credit beat while holding the 2027 re-acceleration thesis intact.

Analyst Q&A Highlights

When Brex's Economics Become Visible

The dominant opening topic was Brex: how much of the promised investment is already in the run-rate, and when the benefits show. Management framed the payoff as gradual and unlockable over time, most of it not yet in the numbers.

Q: "You had previously indicated that you could accelerate Brex's growth almost from day one through stepped-up marketing and tech spend. So I'm just curious, to what extent have those investments already been absorbed into the current expense run rate? And then when should investors see more tangible benefits become more visible?"
— Terry Ma, Barclays

A: "It's been over 100 days since we closed the deal, and we are as excited as ever about Brex… Brex is already experiencing some of the early tailwinds that will come with our brand… Over the coming months, as we test and learn, we will start leaning in with marketing dollars. Fully leveraging the marketing machine of Capital One requires a little more technical integration… so that will come a little further down the road."
— Richard Fairbank, Founder, Chairman & CEO

Assessment: The answer confirms the shape of the deal: investment now, payoff later, with the marketing lean-in and the full integration both "down the road." Combined with the CFO's refusal to break out the P&L, Brex remains exactly as unmodelable as it was in April, three months after close. The "next real information event" slipped again.

Backing Into the Earnings-Power Number

The sharpest exchange of the call was an attempt to triangulate the withheld earnings-power figure from the deal-announcement accretion math. Management rejected the inputs rather than confirm the output.

Q: "You mentioned over 15% accretion to 2027 EPS at the announcement. That rounds up to, let's call it, like $24.50 if we use 16%–17% accretion… So if you use 12.5% on the current share count, you can get to a mid-20s ROTC pro forma. What is wrong with that line of logic?"
— Erica Najarian, UBS

A: "Let me just unpack a couple of the assumptions that you made that don't actually tie to things that we've said… at the time the weighted average consensus for CET1 was 12.5%… we just wanted to remain consistent with that denominator of 12.5 for the sake of doing the math. It is not saying that that is our target."
— Andrew Young, Chief Financial Officer

Assessment: A firm, precise deflection. Management corrected the analyst's assumptions without supplying the correct ones, and reiterated that the 12.5% CET1 is a math convention, not a target. Nearly two and a half years after the deal was struck, the market still cannot check the earnings-power claim against a number the company will confirm. This is the core of the legibility bear.

The Efficiency Ratio as Synergies and Brex Layer In

An analyst pressed on where the efficiency ratio settles now that Brex and Hopper are in the run-rate and two-thirds of the opex synergies are still to come. Management again anchored on earnings power rather than a target.

Q: "Should we think about the impact to adjusted operating efficiency ratio as more marginal on a go-forward basis versus what we've seen with Brex and Hopper now in the run rate? … you do also have the remaining two-thirds of the opex synergies coming as we move into next year as well."
— Sanjay Sakrani, KBW

A: "The debit revenue synergies are essentially in the numbers. The operating expense synergies are more backloaded… we've realized about a third of the operating expense synergies to date, and we're on track to achieve the remaining… by the second half of 2027… With all the moving pieces, we've chosen to focus our conversation on earnings power."
— Richard Fairbank, Founder, Chairman & CEO

Assessment: The clearest statement yet that operating leverage is a 2027 event. Revenue synergies are in; the cost synergies that would bend the efficiency ratio are back-half-2027. Management continues to refuse an efficiency target, substituting the qualitative earnings-power anchor. The reinvestment bear is not resolved; it is scheduled.

Whether This Should Be a 20%-Plus Return Business

A follow-up tried to pin down the through-cycle return the combined company should generate. Management confirmed the direction while declining to brand a precise number.

Q: "If you look back to when the deal was announced… you layer on synergies, it got to a return that was 20% plus or minus… is the right way to think about it this should be at least a 20% return business?"
— Ryan Nash, Goldman Sachs

A: "Out the other side of this we… expect an earnings power very consistent to what we talked about at the outset. We're not branding a precise number because there are a lot of things about Capital One performance that don't lend themselves to… precise settling out with precise numbers. But when we look at the earnings power as reflected in ROTCE, we feel… we're headed for a performance very consistent with what we expected."
— Richard Fairbank, Founder, Chairman & CEO

Assessment: The reported 18% ROTCE this quarter, and higher on an adjusted basis, makes a low-20s through-cycle return look plausible, which is why management's confidence reads as credible rather than promotional. But "very consistent" is again a claim, not a figure, and the analyst did not get the confirmation the question sought.

Capital: 11% as a "Need," and the Buyback Pace

With the deal closed, an analyst asked whether the buyback steps up and how the path to the stated capital level looks. The CFO reframed 11% as a floor of need, not a target to sprint toward.

Q: "Now that the deal has closed, do you think we could see a further step up in the buyback from here? And how do you think about a path towards the stated… capital targets?"
— Ryan Nash, Goldman Sachs

A: "The 11% we define as a long-term capital need as opposed to a target… every year when [CCAR] comes out, we've just seen quite a bit of volatility… our need is derived by our internal modeling… we view capital as having asymmetric value, particularly in times of stress… So we are not in a race to drive it down as quickly as possible to any specific number."
— Andrew Young, Chief Financial Officer

Assessment: A deliberate hold on pace despite ample capital. The $2.7B repurchase is a slight step-up from the recent $2.5B, but CET1 still fell 70bps as Brex and RWA growth consumed capital alongside the buyback. Management is keeping dry powder and signaling no rush to 11%. Prudent, and it caps the near-term per-share tailwind slightly below what the balance sheet could support.

Credit and the Health of the Consumer

An analyst asked about the striking June loss-rate improvement and, more importantly, how the consumer is holding up. Management gave an unambiguously constructive read and explicitly noted the macro backdrop had improved since the spring.

Q: "On the June loss rate, it was down quite a bit month over month… Anything to call out there? … And then just… how the consumer is faring, but more importantly how the Capital One customer is faring."
— Mihir Bhatia, Bank of America

A: "The unemployment rate in June was lower than in February before the Iran conflict began… payment rates… are meaningfully above pre-pandemic levels across all of our customer segments… revolve rates have stabilized… None of these observations are conclusive on their own, but I think collectively they paint a picture of strength to the consumer."
— Richard Fairbank, Founder, Chairman & CEO

Assessment: The direct counterpoint to April's Persian Gulf overhang. Management now cites an unemployment rate below the pre-conflict level and healthy customer-level indicators, and the reserve release put money behind the words. Credit has moved from a precautionary worry back to a clear tailwind, the single biggest positive delta in the quarter.

Extending the Owned Network to Capital One's Own Cards

An analyst asked where the credit-card migration onto the Discover network stands and whether more volume could move than originally planned. The answer revealed the new front-book/back-book testing program.

Q: "Can you talk a little bit about the credit card migration? I know there's been some testing, moving it over to Discover Network. Where are you on that? Is it encouraging? And then do you see a scenario where maybe you could move a little more volume over than you initially thought?"
— Don Fandetti, Wells Fargo

A: "We completed the conversion of our debit card business to the Discover Network… a smashing success… now as we think about building credit card volume on the Discover Network, there's two ways… the front book and the back book… we are leaning hard into… testing originating legacy Capital One branded accounts on the Discover Network, as well as testing the conversion of existing Capital One accounts."
— Richard Fairbank, Founder, Chairman & CEO

Assessment: A concrete new growth and economics lever. Moving Capital One's own credit volume onto the network it owns captures interchange the company currently pays to Visa and Mastercard. It is early and gated on closing international-acceptance gaps, but it is the first time this specific optionality has been framed as an active program rather than a someday possibility.

What They're NOT Saying

1. Still no Brex P&L, one quarter after it was promised. April's call pointed investors to Q2 for Brex's economics. This quarter delivered purchase-accounting marks in an appendix but no revenue, loss rate, deposit base, or EPS dilution, with management calling it "not that material" to break out. The information event keeps sliding.
2. Still no earnings-power level, and no efficiency target. Pressed multiple times and offered a specific line of math to confirm, management reaffirmed only the metric (ROTCE at 12.5% CET1) and the claim of consistency, never the number. Five quarters in, the payoff is still asserted, not shown.
3. The provision-normalized earnings run-rate. With a reserve release flattering the print, management did not frame what "normal" earnings look like without releases, leaving the market to separate the durable ~18% ROTCE base from the non-repeatable sequential EPS jump on its own.
4. How far CET1 falls. Capital dropped 70bps to 13.7% this quarter on buyback plus Brex plus RWA, and management reframed 11% as a "need" it is "not in a race" toward, without a glide path or a quarter-by-quarter capital plan.
5. The near-term cost of the new network-migration testing. Testing legacy Capital One volume onto the Discover network, plus the international-acceptance build, adds to an already long investment list, but management would not size the incremental spend, folding it into the same qualitative "reflected in the efficiency ratio" framing.

Market Reaction

• Pre-print setup: COF closed at $206.22 on July 21, down 14.9% year-to-date against the S&P 500 up 9.7%, and down 4.5% over the trailing twelve months. The stock sat about 2.3% above its trailing-30-day level and well within a 52-week closing range of $176.10 to $257.94. Options implied a move of roughly 5% into the print.
• After-hours move: Muted despite the size of the beat. The immediate reaction ranged from roughly unchanged near $206 to up about 2%, with nothing approaching the 5% the options priced. Volume color was unremarkable in the after-hours session.

A 24% EPS beat that moves the stock 2% or less is the market telling you what kind of beat it was. Investors did the same decomposition we did: the surprise was a lighter provision and a reserve release, not an acceleration in the operating engine, and the disclosure the Street most wanted, Brex's economics and a quantified earnings-power figure, did not come. A recurring framing in the coverage was that the quarter was good enough but did not answer the big question hanging over the stock, which is precisely the operating-leverage-and-legibility question we have kept it at Hold on.

The tape is consistent with our rating. A cheap, high-quality franchise printing an 18% ROTCE and improving credit deserves a firmer bid, and it may well get one as the reaction session plays out. But the absence of a violent move up tells you the market is not yet willing to re-rate on a credit-driven beat that left the structural questions open, which is the same reason we raise conviction without changing the rating.

Street Perspective

Debate: Is a Credit-Driven Beat "High Quality" or a Head-Fake?

Bull view: Credit is the core earnings lever for a card lender, so a reserve release on genuinely improving delinquencies and charge-offs, with the macro overlay lifting, is the highest-quality kind of good news. The 18% ROTCE proves the earnings base is real, and lower losses are a durable feature of Capital One's superior underwriting, not a one-off.

Bear view: Reserve releases are not revenue. Pre-provision earnings were flat, expenses grew faster than revenue, and strip out the release and underlying EPS is well below $5.81. A beat that depends on giving back reserves cannot repeat, and it tells you nothing about whether the heavy investment is finally producing operating leverage.

Our take: Both are right, which is why it is Hold-confirming rather than upgrade-triggering. The credit improvement is real and durable enough to strengthen the base and resolve the macro bear, so we raise conviction. But the bears are correct that this quarter did not answer the operating-leverage question, and a non-repeatable release is not the catalyst that re-rates the stock.

Debate: Does the Reaffirmed Earnings Power Deserve the Benefit of the Doubt?

Bull view: Management has now delivered an 18% ROTCE while carrying peak investment and an unfinished integration, and it keeps saying earnings power is on track. A founder-CEO with a three-decade record of value creation reaffirming the deal thesis, backed by a real return this quarter, should be believed; the number will reveal itself as synergies land in 2027.

Bear view: Trust-me is not a disclosure. Five quarters after the deal and one quarter after Brex was promised to become legible, the company still will not confirm an earnings-power level, an efficiency target, or a Brex P&L, and it actively deflected an analyst's attempt to back into the math. Opacity at this stage is a choice, and it caps the multiple.

Our take: The strong ROTCE tilts us toward giving management the benefit of the doubt on direction, but the persistent refusal to quantify anything is exactly why the stock trades at 8x. The market will pay for legibility; until the payoff is shown rather than asserted, the discount is rational and the re-rating is deferred.

Debate: At ~8x With Credit Inflecting, Is the Risk/Reward Finally Favorable?

Bull view: You are buying the largest domestic card book and a unique owned network at 8x an ~18% ROTCE earnings base, with credit inflecting positively, loan growth returning, a new network-migration lever, and a steady buyback. The catalysts are stacking up and the 2027 growth re-acceleration plus Brex and network optionality are close to free at this price.

Bear view: Cheap has a reason. The growth inflection is still a 2027 event, the operating-leverage proof is 2027, Brex is unquantified, and capital is being consumed. A credit-driven beat that the market barely rewarded is not a catalyst, and estimate risk cuts both ways once the reserve tailwind is spent.

Our take: Closer to favorable than it was in April, but not yet there. The credit inflection and the reaffirmed base genuinely improve the risk/reward and are why we lift conviction. But with the two things we said we needed for an upgrade still absent, and with the stock unlikely to offer a cheaper entry after a beat, the balance is improved-but-symmetric rather than clearly favorable. We want one more quarter of the efficiency ratio holding and, ideally, some Brex legibility before we pay up.

Model & Valuation Framework

Valuation framework: At roughly $206 against a normalized combined-company adjusted-EPS base, COF trades near 8x, still an undemanding multiple that reflects the deferred operating leverage and the withheld earnings-power level rather than any doubt about the franchise or the credit. We lift our fair-value range to $205–245, anchored on roughly 8.5–10x normalized adjusted EPS. The low end is supported by the buyback, an 18% ROTCE base, improving credit, and tangible book near $105; the high end still requires the two things this quarter did not deliver, demonstrated operating leverage as synergies land and a legible Brex/earnings-power figure. The reserve release lifts our 2026 EPS view but we are careful not to extrapolate a non-repeatable credit tailwind into the out-years. The network and Brex optionality is real and, with the new credit-migration testing, arguably larger than a quarter ago, but it remains back-end-loaded and, for Brex, still unmodelable, so we keep it out of the base case.

Thesis Scorecard Post-Earnings

We grade the standing thesis, the same eight pillars carried since initiation, against this quarter's print and call.

Overall: The thesis strengthens, on balance. Two bear points moved in our favor: the consumer-credit/macro bear eased materially (reserve release, overlay reduced), and the earnings base was reaffirmed at an 18% ROTCE. Bull-2 (card scale) improved as loan growth turned positive, and Bull-1 (network) added a concrete new leg. But the two pillars that gate an upgrade did not move: the reinvestment/operating-leverage bear played out exactly as guided (efficiency stepped up, PPNR flat), and the accretion-legibility bear is intact because Brex stayed a black box and the earnings-power level stayed withheld.

Action: Maintain Hold; raise conviction to 6 from 5. The credit inflection and the reaffirmed, higher-quality earnings base move us closer to an upgrade and improve the downside. But with operating leverage unproven and Brex still opaque, and with the stock unlikely to offer a cheaper entry after a beat, the risk/reward is improved-but-balanced rather than clearly favorable. We would upgrade to Outperform on the adjusted efficiency ratio holding or improving as marketing stays elevated (leverage proving real), a quantified earnings-power or Brex accretion figure, or a visible loan-growth inflection ahead of the 2027 timeline. We would revisit the downside if credit turns or investment resumes outrunning revenue.

Bottom Line: A Higher-Quality Base, the Same Two Open Questions

Rating decision: We maintain Hold on Capital One and raise our conviction to 6 from 5. This was the strongest headline quarter we have covered: adjusted EPS of $5.81 beat consensus by roughly 24%, return on tangible common equity snapped back to 18%, loan growth turned positive, credit improved, and the macro cloud that darkened the April call has lifted. It confirms that the franchise's earnings base is high quality and that the credit engine, the heart of a card lender, is running well.

It was not, however, an upgrade quarter, for two connected reasons. First, the beat was credit, not leverage: a $1.1B lighter provision and a $662M reserve release did the work, while pre-provision earnings were flat, expenses rose 7%, and the efficiency ratio stepped up exactly as management told us in April it would. The operating-leverage question that drove our January downgrade is confirmed as a 2027 event, not resolved. Second, the one disclosure the Street was told to expect this quarter, Brex's economics, did not arrive, and management again declined to put a level on "earnings power," even deflecting an analyst's explicit attempt to back into it. The two catalysts we named for an upgrade both failed to fire.

Why conviction rises but the rating does not: the credit and macro bear resolved in our favor and the earnings base is now demonstrably an ~18% ROTCE, which meaningfully improves the downside and moves us toward Outperform. But an upgrade needs the offense, not just a firmer defense, and the offense (operating leverage proving real, a legible Brex and earnings-power figure) is still absent. At roughly 8x with a stronger base, the stock is more attractive than it was in April, yet a credit-driven beat that the market rewarded with a shrug does not re-rate it, and the deferred payoffs still argue for patience.

What would change our mind:

• Upgrade to Outperform: the adjusted efficiency ratio holding or improving through the back half as marketing stays elevated (operating leverage proving real rather than seasonal); a quantified earnings-power figure or a credible Brex accretion path; a visible loan-growth inflection ahead of the 2027 timeline; or a lower entry that widens the margin of safety.
• Downgrade to Underperform: a genuine reversal in credit that outruns the reserves just released; investment spend re-accelerating while revenue growth stalls (the reinvestment bear fully materializing); or Brex and the network build proving to be persistent, unquantified drags with no accretion in sight.

Signposts for Q3 2026 (October): whether the adjusted efficiency ratio holds as the second-half marketing lean-in continues (the operating-leverage tell); any actual Brex financial disclosure now that the marks are booked; the NIM catch-up management guided (average cash converging to ending, plus day count); the Discover brownout approaching its Q4 bottom and the full transition of new originations onto Capital One's tech by end-Q3; whether credit stays a tailwind or the reserve release proves early; and the buyback pace against a CET1 that fell to 13.7%.