Capital One Financial Corporation COF

Key data

The business

Capital One is one of the largest US card issuers and, since May 2025, the owner of Discover, the fourth American card network alongside Visa, Mastercard and American Express. It lends to consumers and commercial customers, funds most of that with retail deposits, and earns the spread between the two. On top of the spread sits non-interest income: the discount and interchange it collects when its cards are swiped, net of the rewards paid back. What Discover adds is the network itself. A Visa-branded issuer pays a network to move every transaction. Capital One now owns a set of rails, Discover for credit and PULSE for debit, and each cardholder it migrates onto them turns a cost into revenue. The catch is who holds the cards. Capital One's book skews to the middle and lower end of US credit, the customer who carries a balance and pays interest, which is where profit is fattest and where losses appear first when the economy turns. The moat is the network license paired with a deposit base cheap enough to fund the lending.

The numbers

Two things happened at once in 2025: revenue jumped to $53.4 billion as Discover consolidated, and reported profit collapsed to $2.453 billion, $4.03 a share, its worst in over a decade. The quarter-to-quarter line, measured in the adjusted earnings the market grades against its own estimates, shows the shape of the recovery.

The pattern is a business absorbing a large acquisition, not one falling apart. The $4.3 billion loss in the June 2025 quarter was the day-two accounting for Discover: the rules force a lender to reserve up front for the lifetime losses of loans it acquires, a one-time charge against a book that keeps earning. Two quarters later the run rate is back above $5. The credit-normalization worry a careful reader would have flagged two days ago has not broken here, because a book heading for trouble does not print a 21% upside quarter.

Across five fiscal years the earnings look worse than the franchise, because two different forces move through them.

The 2021 peak, $26.94, was a mirage: it was the year Capital One released the reserves it had piled up in the pandemic, and released reserves flow straight to profit. Everything since is the reverse, first that cushion rebuilding, then the Discover charge on top. Strip both and the underlying earning power is roughly the $20 in adjusted earnings the last twelve months produced, which is what the market is paying about nine times forward for. The variant here is narrow: at 1.9 times tangible book, the top of its range, the market already credits Capital One for owning a network. What it does not yet pay for is proof the migration works. The print that settles it is network purchase volume and interchange income climbing in the next two filings faster than card losses climb.

Management

Richard Fairbank founded Capital One and has run it since 1994, which is the most important fact about how it is managed: a founder betting his legacy on the biggest deal of his career. The insider ledger over the past year runs one direction, about $18.6 million sold across 35 transactions and not a single open-market purchase, the largest a pair of February sales by one executive worth about $4.4 million together. Plan status is not disclosed in the filings pulled, so read those as routine compensation-driven selling rather than a signal, though the absence of any buying near a 12-year-high valuation is its own quiet comment. The company resumed heavy buybacks in 2025, repurchasing $4.1 billion of stock after two years near silence at about $0.7 billion a year, now buying at a richer 1.9 times tangible book than those quiet years paid.

How it fails or surprises you

Network migration pays off (right tail). If Capital One moves even a modest share of its own card volume onto Discover and PULSE, each transaction converts a fee it paid Visa into fee income it keeps. The market pays nothing for this today. The tell is network revenue and purchase volume in the next two 10-Qs rising faster than the card book itself.

Consumer credit turns. The book leans on the middle-and-lower-end borrower who carries a balance, and Discover added more of the same. A real rise in unemployment would push charge-offs up on exactly the loans that earn the most. The print is the net charge-off rate over the next two quarterly filings drifting higher quarter over quarter, direction more than level.

The adjusted-to-GAAP gap. The nine-times multiple rests on about $20 of adjusted earnings, but GAAP has run well below that through the integration. If integration and amortization charges prove less temporary than billed, the true normalized number is lower and the stock is not as cheap as it screens. The tell is that gap failing to close through 2026.

Closing thoughts

The payoff here is a specific print resolving uncertainty: network purchase volume and interchange revenue climbing faster than credit losses, visible in the next two 10-Qs. If that lands, the market re-rates the stock for earning both sides of the transaction. If losses climb faster than network income, the migration thesis breaks. Underneath sits a cyclical exposure no print resolves, a lending book tilted to the borrower who defaults first when unemployment rises, so survivability matters and Capital One qualifies: it absorbed a $4.3 billion reserve charge and a quarter of loss without cutting the dividend or stopping buybacks.

The bet is still that owning Discover and migrating Capital One's cardholders onto those Discover and PULSE rails lets it earn the merchant discount fee and the interest on the same card, instead of paying Visa or Mastercard for every transaction. It breaks if losses on those cards climb faster than the fees the network throws off, which is why the one pair to watch is the net charge-off rate against network and interchange revenue, quarter by quarter. If by the end of 2026 charge-offs are drifting up while network income is flat, the read was wrong.