COF — Capital One Financial Corporation

The tape and the sell side agree: shares rose 3.05% on 2026-10-21 and 5.97% over the trailing week after Q3 2026 showed adjusted EPS of $6.69 versus ~$5.50 expected, net interest margin up 40bp sequentially to 8.41%, provisions down $73M to $2.9B, and a $225M reserve release. Management said Discover integration is running ahead of schedule with $1.8B of annualized synergies already captured, and analysts responded by nudging targets up (Piper Sandler to $269 from $260). The consensus rating is Buy with 36 buys, 17 holds and 4 sells, and a $250.70 average target. Note the sentiment is constructive but not euphoric: the stock is still 20% off its high and flat over six months, so the market is rewarding execution without yet re-rating the multiple.

The 3-year series has one giant discontinuity and one genuine inflection, and they must not be confused. The discontinuity is Q2 2025: the Discover close lifted stockholders' equity 75% ($63.5B to $111.0B) and total assets from $493.6B to $659.0B, and the quarter printed a GAAP EPS loss of -$8.58 on acquisition-related charges. That is a one-off accounting event, not an operating result — it distorts every trailing-twelve-month window ending through mid-2026 and is the single biggest reason the deterministic score looks mediocre. Strip it out and the picture is: revenue stepping from $13.4B (Q1 2025) to $16.4B to $19.8B as Discover consolidated, then three straight quarters of steady ~2.5% sequential growth in the combined entity — $19.32B (Q1 2026), $19.86B (Q2 2026), $20.36B (Q3 2026). Organic growth is modest (~2.7% YoY on the like-for-like Q3 comparison of $20.36B vs $19.82B), not the 48% headline you'd get from a naive YoY read. Earnings are the real story: diluted EPS of $3.26 (Q4 2025) → $3.34 → $4.73 → $5.61 is clear sequential acceleration, and Q3 2026's $5.61 is +16% versus Q3 2025's $4.83. Credit is normalizing rather than deteriorating (provision $2.9B, net charge-offs $3.1B, $225M released), and non-interest expense fell 2% sequentially. One caveat: the FMP operating-income and gross-profit lines are unreliable for a bank — Q2 2026 shows $7.34B operating income versus Q3 2026's $4.57B while net income simultaneously rose from $3.02B to $3.54B. Any model leaning on those lines is leaning on noise. Verdict: inflecting upward, not decelerating.

Cheap, with a real caveat about the benchmark. As of 2026-10-21 the stock traded at 12.25x trailing EPS of $17.85 and about 9.9x consensus FY2027 EPS of $21.05, against 1.15x book and 5.95x EV/EBITDA. The 19.4x five-year average P/E is a misleading anchor — it is inflated by 2020-2021, when massive reserve builds crushed the E in the denominator. Against a more honest pre-2020 normal of 10-12x forward, the current ~10x is at the low end but not screamingly so. The PEG of 0.02 is a mathematical artifact and should be ignored. What matters: a ~13% ROTCE business (Q3 2026 net income of $3.54B annualizes to ~$14.2B on $109.7B of equity) growing EPS at a low-to-mid-teens clip, now owning a proprietary payment network, at roughly 10x forward earnings and 1.15x book. That is cheap-to-fair, and the discount reflects genuine credit-cycle and integration risk rather than a broken business. The reported ROE of 0.022 and ROIC of 0.011 in the data feed are artifacts of the acquisition period and should be disregarded.

The market is paying roughly 10x forward for a company that just beat by $1.19 on adjusted EPS and guided nothing down. Two readings are possible and they point in different directions. The bear reading: the market is discounting the durability of these earnings because card charge-offs at $3.1B are still elevated in absolute terms and a 4.5%+ unemployment scenario would eat the reserve release and then some — so the low multiple is a rational risk premium, not a mispricing. The bull reading: estimates are demonstrably too low (a $1.19 beat is a 22% surprise), $1.8B of synergies are already banked with more to come through 2027, and the market has not yet given Capital One any credit for owning Discover's network, which structurally improves fee income and funding mix. My read leans toward the second: consensus of $21.05 for FY2027 already implies ~13% growth off an FY2026 base of roughly $18.60, and if Q3's run-rate holds that number has room to move up rather than down. The stock does not need multiple expansion to work — it needs estimates to hold.

Range-bound with a fresh post-earnings lift, not extended. As of 2026-10-21, RSI(14) stood at 57.7 — comfortably neutral, nowhere near overbought despite the 5.97% one-week gain. The price was 4.04% above its 20-day SMA (the earnings gap), 0.55% below its 50-day SMA, and 1.64% above its 200-day SMA — a flat, tangled moving-average stack that describes six months of going nowhere (+0.5% one month, +2.28% three months, +3.27% six months). Against a 52-week range of $174.24-$259.64, the stock is in the lower-middle third. Entry-timing implication: this is a neutral setup, not a chase and not a falling knife. There is no momentum signal to ride and no oversold washout to buy; a pullback toward the $200 area (the 50-day) would offer better risk/reward than buying the earnings gap, and a decisive reclaim of the 50-day with volume would confirm the Q3 inflection is being believed.

• Interest rate path: NIM expanded 40bp to 8.41% on higher card yields, but aggressive Fed cuts would compress asset yields faster than deposit costs fall — though the offset is relief on consumer credit and funding costs, so the net effect is not obviously negative.
• Consumer credit and unemployment: net charge-offs of $3.1B and a $225M reserve release assume continued normalization. A rise in unemployment, particularly among subprime and near-prime cohorts, would reverse the provision trend that drove the Q3 beat.
• Post-merger regulatory scrutiny: the Discover deal cleared, but data migration, compliance frameworks and consumer-protection oversight (late fees, APR caps, CFPB posture) remain live risks to the synergy math and to fee income.
• Competitive intensity in cards: marketing spend is set to re-accelerate into 2027, and premium/prime share is contested by Amex, JPMorgan and the neobanks — expense discipline may prove harder to sustain than the 2% sequential decline suggests.
• Consumer spending backdrop: card volumes and interchange revenue are levered to nominal spending; a slowdown hits both net interest income and fee income simultaneously.

• Discover system migrations through 2027 — each wave moves legacy accounts onto Capital One's platform and unlocks incremental cost synergies beyond the $1.8B already annualized.
• Marketing re-acceleration into 2027, which management framed as an offensive investment in prime and premium segments using the proprietary network.
• Further reserve releases if credit continues to stabilize, which would flow directly to EPS ahead of consensus.
• Q4 2026 earnings on 2027-01-28 (estimated), the first clean read on whether the Q3 run-rate is durable.
• Capital return: a 1.45% dividend yield plus buyback capacity against a $109.7B equity base and 1.15x book.

• Credit deterioration: the Q3 beat leaned on a lower provision and a reserve release. If charge-offs climb, that reverses and the low multiple turns out to be justified rather than an opportunity.
• Integration execution: the equity and asset base jumped 75% and 34% respectively in a single quarter. Data migration failures, customer attrition or regulatory findings on the Discover book would hit both the synergy case and the credibility of guidance.
• Multiple compression risk: at ~10x forward the stock is already cheap, but cheap banks get cheaper in a credit scare — the 52-week low of $174.24 is a 16% drawdown from the 2026-10-21 price and entirely reachable.
• Share-count and EPS optics: diluted EPS is being flattered by buybacks and by the absence of the one-off Q2 2025 charge from forward comparisons; a purely EPS-driven bull case can mask flat absolute earnings growth.

This is a better business than its 47.7 deterministic score implies, and a cheaper stock than its 19.4x five-year average P/E implies — but the cheapness is not free money, it is payment for credit risk. The core case: Capital One just earned $5.61 GAAP ($6.69 adjusted) in a quarter, has three straight quarters of sequential EPS acceleration, expanding margins, easing provisions, and $1.8B of Discover synergies already in the run-rate, all for ~10x forward earnings and 1.15x book. My independent fair value sits around $242, roughly 17% above the 2026-10-21 price, which is a decent but not extraordinary expected return for a stock with this much cyclical exposure. I would not call this a compelling buy at $207.71 — it is a reasonably priced, improving franchise where the next leg depends entirely on credit holding up. The honest lean is neutral: good business, fair-to-cheap price, real tail risk.

Applied 11.5x to consensus FY2027 EPS of $21.05 — a multiple between Capital One's crisis-era trough (~8x) and its pre-2020 normal (~11-12x), justified by a ~13% ROTCE, low-to-mid-teens EPS growth, and $1.8B of already-realized Discover synergies. That $242 figure cross-checks against a Gordon-growth justified P/B of ~1.35x on roughly $180 of book value per share, which lands at ~$243.