Capital One Q3 2026 Earnings: $5.61 EPS ($6.69 Adjusted) as the Discover Deal Starts Paying

On October 20, 2026, Capital One reported third-quarter net income of $3.5 billion, or $5.61 per diluted share — up 16% from a year earlier — on revenue of $16.3 billion (+6%). On the adjusted basis management prefers, earnings were $6.69 per share, and the engines of that number all moved the right way: the net interest margin widened 40 basis points to 8.41%, net charge-offs fell 14% sequentially with a $225 million reserve release, and interchange income jumped 29% year over year as the acquired Discover payments network starts contributing fee income. The $1.08 gap between GAAP and adjusted EPS is the merger still flowing through the income statement — acquisition amortization plus Discover and Brex integration costs. Our new Capital One ledger models all of it — two annuals plus this quarter — in plain-text double entry.

The Headline Numbers​

Every figure below comes from the Q3 2026 financial supplement (8-K Exhibit 99.2) and the earnings release (Exhibit 99.1):

Three things stand out. First, the year-over-year comparisons finally mean something: Discover closed on May 18, 2025, so the year-ago quarter already reflected the combined company — this is the first clean combined-vs.-combined quarter, and revenue (+6%), net income (+11%) and adjusted EPS (+12%) all grew on it. (Nine-month comparisons are still garbage, polluted by the $8.8 billion Day-1 provision that crushed the first half of 2025; ignore them.) Second, the tax line did quiet work: pre-tax income grew just 4% year over year, but the effective rate fell to 22.6% from 27.1%, converting that into 11% net income growth. Third, EPS (+16%) outran net income (+11%) because the share count keeps shrinking — 601.5 million shares outstanding, down 5.4% year over year on sustained buybacks — while the dividend rose 33% to $0.80. Pre-provision earnings of $7.5 billion, up 10% sequentially, confirm the operating core accelerated regardless of credit or tax timing.

Revenue Deep Dive​

Net interest income rose 4% both sequentially and year over year to $12,886 million, and the mechanics are all on the asset side. Interest income on loans grew 4% sequentially to $15,515 million on a loan book that expanded 1% to $462.5 billion, while total interest expense was flat sequentially at $4,013 million — deposit costs actually fell 7% year over year even as deposits grew 3%. The rate paid on interest-bearing deposits ticked up just 3 basis points sequentially to 2.94%, so asset yields repriced faster than funding costs and the margin captured the spread. That is the structural advantage of a card-heavy book: $276.4 billion of credit card loans reprice with a speed and spread no regional bank's mortgage book can match.

Fee income is where the Discover acquisition shows up most directly. Discount and interchange fees jumped 29% year over year to $2,335 million — the Global Payment Network (the former Discover Network plus PULSE and Diners Club) earning its keep — while service charges slipped 5% and other income grew 21%. Total non-interest income of $3,458 million was flat sequentially but up 17% from a year ago, and net of the small $40 million securities loss it would have been higher still.

The segment split shows a card franchise carrying the company and a consumer bank inflecting:

Credit Card is 74% of revenue, and its mix deserves a careful read: revenue grew 5%, but profit slipped 2% as the segment's provision rose 5% and its expense grew 9%. Domestic Card — 95% of card revenue at $11,480 million, now including Brex from the second quarter — tells the same story at higher resolution. This is the mature, high-return engine being harvested, not the growth story. The growth story is Consumer Banking, up 17% in revenue and 33% in profit: net interest income grew 7% on 4% auto-loan growth, but the standout is non-interest income of $769 million, up 62% year over year — the Global Payment Network fees live in this segment, and they are compounding. Commercial Banking was flat on revenue with profit up 9% on lower expenses, a steady contributor. The Other segment's loss narrowed to $193 million from $433 million; it holds the unallocated integration expenses, so its improvement is partly the merger noise quieting.

On the management-signal scan, the release is thin — and that thinness is itself the finding. There is no demand, pricing, or credit-cycle commentary anywhere in it; the only forward-looking statement is Fairbank's: "The Discover integration continues to go well and we're making good progress on delivering the expected synergies," plus "We continue to invest to deliver sustained growth and value creation over the long-term." No call transcript was available at publication. For a bank, the behavioral signals substitute: marketing spend down 4% sequentially says management is not chasing growth at any price, while the $225 million reserve release says the chief risk officer sees delinquency data improving before the market does. Both read as quiet confidence — but neither quantifies the synergies, which remain a narrative rather than a number.

The Margin Story​

Four quarters of margin data show a bank whose earning power troughed with the merger and is now compounding out of it:

The mechanics are clean. The 40-basis-point sequential NIM expansion is asset repricing (loan interest income +4%) against frozen funding costs (interest expense flat at $4.0 billion) — price realizations beating volume on both sides of the balance sheet. The 286-basis-point efficiency improvement is positive operating leverage in its purest form: revenue up 3% while expenses fell 2%, with marketing down 4% ($1,661 million to $1,594 million) and operating expense down 2% ($7,382 million to $7,262 million). The $868 million of adjusting items sits inside that $8,856 million expense base, which is why the adjusted efficiency ratio of 48.76% — and the adjusted operating ratio of 38.99% — reads so much better than the GAAP print.

Two honest caveats. Year over year, operating leverage was negative — expenses grew 7.2% against revenue growth of 6.4%, so the GAAP efficiency ratio is 39 basis points worse than a year ago. The merger's cost base is still digesting. And the Q1 trough (NIM 7.87%, card net charge-offs seasonally peaking) is a reminder that this franchise has a seasonal saw-tooth: first quarters run hot on charge-offs, so the current 2.73% charge-off rate should be read as mid-cycle, not trough.

The One Big Question: Is the Discover Deal Earning Its Keep?​

Capital One paid $51.8 billion in stock for Discover — 1.0192 shares per Discover share — booking $168.6 billion of identifiable assets, $108.2 billion of loans, $106.9 billion of deposits, $13.4 billion of goodwill and a $15.6 billion intangible balance that now amortizes at roughly $500 million a quarter. The bull case for the deal was always twofold: scale economics on the card franchise, and ownership of a payments network. This quarter lets us grade both, starting with what the deal still costs:

The $868 million quarterly run-rate of deal costs is 16% of GAAP EPS — down from the $8.8 billion Day-1 provision shock but still material, and the amortization piece runs for years. Against that cost, the revenue evidence is accumulating on schedule. Interchange income up 29% year over year is the network thesis in one number. Consumer Banking non-interest income up 62% is the network fees compounding inside the segment that houses them. The 10-K notes the company has "substantially completed the reissuance of legacy Capital One customer debit cards onto the Global Payment Network" — the plumbing conversion that turns a cost synergy (routing volume off competitors' rails) into a revenue synergy (keeping the network fee). And Fairbank's "making good progress on delivering the expected synergies" is at least consistent with the adjusted operating efficiency ratio improving to 38.99%.

But intellectual honesty requires the other column of the ledger. Integration spending is still ramping, not tapering: nine-month Discover integration expenses of $1,051 million already exceed the $757 million of the year-ago period. The synergies themselves remain unquantified — no dollar figure, no timeline, no synergy-vs.-target bridge appears anywhere in the release or supplement. The Brex acquisition ($4.5 billion, closed April 7) layers a second integration on top of the first, with $47 million of Brex integration expense this quarter and $3.1 billion of fresh goodwill. And the capital math is tightening: the CET1 ratio slipped to 13.8% from 14.4% a year ago as buybacks consumed capital faster than earnings rebuilt it.

Our read: the revenue side of the deal is outperforming while the cost side is still an open question. The falsifiable markers for the next two quarters are specific: Discover integration expense must start declining sequentially from this quarter's $338 million; the adjusted operating efficiency ratio must hold below 40% while marketing normalizes upward from the seasonally soft $1,594 million; and management must put a number on realized synergies — "good progress" expires as guidance after six quarters of ownership. Hit those, and the $1.08 quarterly adjustment converges toward zero on schedule. Miss them, and the $6.69 is a non-GAAP construct atop a GAAP franchise earning $5.61 with rising deal costs.

Tracking a $673 Billion Bank in Plain Text​

A bank balance sheet is where double-entry accounting earns its keep: every dollar of the $672.9 billion in assets must reconcile to a dollar of liabilities or equity, and Beancount enforces that identity mechanically. Our ledger models Capital One exactly the way we model every company — read how we model every company for the shared conventions — with the fleet's money-center-bank chart: revenue is total net revenue, the provision for credit losses is the cost of revenue, and non-interest expense is operating expense, with the release's three adjusting items isolated as their own postings. Income is negative (credit) and expenses positive (debit), so the transaction below sums to zero:

; Check: -16344 + 2916 + 7988 + 483 + 338 + 47 + 1033 + 3539 = 0
 
2026-09-30 * "Capital One Financial Corporation" "FY2026Q3 Income Statement"
  Income:Revenue                             -16344 MUSD
  Expenses:CostOfRevenue                       2916 MUSD
  Expenses:OperatingExpenses                   7988 MUSD
  Expenses:OperatingExpenses                      483 MUSD
  Expenses:OperatingExpenses                      338 MUSD
  Expenses:OperatingExpenses                       47 MUSD
  Expenses:IncomeTax                           1033 MUSD
  Equity:Adjustments                           3539 MUSD

Contrast the cleanliness of that transaction with FY2025 in the same ledger, where the single $53,434 million revenue posting sits above an $8,800 million Day-1 allowance posting and a $1,100 million integration posting — the merger year preserved in amber, with the $365 million discontinued-operations gain from the Discover Home Loan sale on its own line. Two annuals — one clean standalone year and the merger year — plus this quarter are committed in the repo: that short history is deliberate, because anything before FY2024 describes a company whose scale, share count and business mix no longer exist.

The balance-sheet number that tells the narrative is accumulated other comprehensive loss: a ($10,008) million debit, wider by $3,708 million in a single quarter. That one line explains the quarter's paradox — Capital One earned $3,539 million yet total equity fell $4,137 million to $109,656 million. Walk it with filed numbers: additional paid-in capital −$708 million, retained earnings +$2,963 million (earnings less dividends), AOCI −$3,708 million, treasury stock −$2,684 million on continued buybacks — the buybacks, the dividend and the securities-loss swing consumed more than 100% of earnings. Tangible book value per share fell 2% sequentially to $102.72 even as the franchise earned a 21.12% return on tangible equity. The earnings engine is roaring; the equity account is hostage to the rate curve.

The second number worth your attention is the allowance for credit losses: $22,739 million at 4.92% coverage, down from 5.21% a year ago, with the quarter's $225 million release inside the $2,916 million provision. A bank releases reserves only when the delinquency data lets it — 30-day delinquencies fell to 3.41% from 3.50% — so this is management's quantified vote that the credit cycle is turning in its favor. Against it, keep one eye on the $15,569 million of other intangibles: every quarter amortizes roughly $500 million of it, the mechanical reason the GAAP-to-adjusted gap will close slowly.

The Multi-Year Arc​

Two annuals plus the new quarter, every figure from the ledger's filed statements:

*Q3 is a single quarter; balance-sheet columns are period-end snapshots.

The arc is a company that doubled in size and is now proving the earnings power follows. FY2024 is the standalone baseline: $39.1 billion of revenue, $4.75 billion of earnings, $11.59 per share on 381 million shares. FY2025 is the merger year in all its accounting violence — revenue up 37% to $53.4 billion on seven months of Discover, but the $8.8 billion Day-1 provision and $1.1 billion of integration expense crushed net income to $2.45 billion, or $4.03 per share on a share count that ballooned to 625 million. Q3 2026 is the first glimpse of normalized earnings power: $3.5 billion in a single quarter annualizes past the entire FY2024 profit — before seasonality and cycle effects — with the share count already retreating to 601.5 million on buybacks. The question the next four quarters answer is how much of the $6.69 adjusted run-rate converts to GAAP as the deal costs fade.

One comparability note: each annual in the ledger is modeled from its own 10-K as filed, so two retrospective presentation changes step the detail lines with totals unchanged — the FY2025 10-K renames FY2024's "Interchange fees, net" to "Discount and interchange fees, net" (same $4,882 million), and splits $233 million of other intangible assets out of FY2024's Other assets. The break that matters is the Discover acquisition itself: FY2024 is standalone, FY2025 includes Discover only from May 18, and FY2026Q3 adds Brex from April 7, 2026. See the Comparability note in the ledger's main.bean.

The Verdict: Bull vs. Bear​

Bull Case

• The margin expanded 40bps in one quarter to 8.41%. Asset yields repriced 4% while funding costs froze — the card-heavy book's pricing power, converting directly into a 21.12% ROTCE, up 230bps year over year.
• Credit is inflecting. Net charge-offs down 14% sequentially to a 2.73% rate, a $225 million reserve release, delinquencies improving — the provision is falling while loans grow, the combination every bank guides to and few print.
• The network thesis is printing. Interchange +29% year over year and Consumer Banking fee income +62% are Discover's rails contributing real revenue, with the debit-card reissuance substantially complete.
• Adjusted EPS of $6.69 grew 15% sequentially and 12% year over year. Pre-provision earnings of $7.5 billion (+10% QoQ) confirm the operating core, not just credit timing, drives the beat.
• Capital return compounds. Shares down 5.4% year over year, dividend up 33% to $0.80, CET1 at 13.8% — the buyback more than offsets the merger dilution trajectory.

Bear Case

• $1.08 of the $6.69 — 16% of EPS — is adjustments. The $483 million quarterly amortization runs for years against a $15.6 billion intangible balance, and GAAP efficiency (54.19%) is still worse than a year ago.
• Equity shrank $4.1 billion in an earnings quarter. The $3.7 billion AOCI swing plus buybacks and dividends consumed more than 100% of earnings; tangible book value fell 2% both sequentially and year over year.
• Integration spend is still ramping. Nine-month Discover integration of $1,051 million exceeds the year-ago $757 million, synergies remain an unquantified "good progress," and Brex adds a second concurrent integration.
• The core card franchise shows strain. Card profit fell 2% year over year despite 5% revenue growth, and Consumer Banking's provision jumped 23% as auto loans grow 4% a quarter — late-cycle exposure accumulating where growth is fastest.
• Funding is tightening at the margin. Deposits fell 1% sequentially while loans grew 1%, and the deposit rate ticked up 3bps — small moves, but the wrong direction for a bank priced on margin expansion.

Our Take

Capital One just printed the quarter the Discover deal was supposed to produce — margin expansion, network fee growth, improving credit, and a 21% return on tangible equity — while honestly labeling the $868 million of quarterly deal costs that separate GAAP from adjusted reality. The bears' best arguments are real but bounded: the AOCI swing is rate noise against a held-to-maturity-capable balance sheet, and the integration spend has a finite life. The number that decides the debate is next quarter's Discover integration line: declining sequentially from $338 million with synergies finally quantified, and the $6.69 run-rate starts converting to GAAP on schedule. Anything else, and the market is right to keep a conglomerate discount on the combined company until management proves the merger math bottom-up.