Capital One

What the headline numbers say

The credit metrics on COF’s bank-subsidiary book remained orderly through Q1. Net charge-off rate at 6.31% is a 3-basis-point uptick from Q4 2025, well within quarterly noise. Net charge-offs in dollar terms moved up just 0.4% sequentially to $3.80 billion. By the metrics that historically signal credit deterioration, Q1 2026 was unremarkable.

Two numbers contradict that read. End-of-period card receivables fell to $239.2 billion, a 3.7% sequential decline. For a recently-merged platform that absorbed Discover’s card book in May 2025, three quarters into consolidation the platform should be producing growth, not contraction. Card net revenue tracked receivables lower, slipping 1.9% to $11.4 billion.

A book that is shrinking is not, mechanically, the book of an issuer expecting credit improvement. Issuers expecting better credit underwrite more, not less.

What the underlying drivers say

The dominant exogenous variable on Capital One’s Q1 print is the May 18, 2025 Discover acquisition. Q1 2026 is the fourth quarter of consolidated reporting. The merged receivables base peaked in 2025 as Discover’s book joined the COF balance sheet; the sequential decline since suggests deliberate trimming — closing overlapping accounts in the combined portfolio, exiting sub-prime origination cohorts the combined risk team has flagged, or running off Discover’s installment-loan and student-loan exposures that don’t fit the post-merger card-first strategy. The interactive receivables chart on this page makes the trajectory legible.

The provision build supports the deliberate-trimming read. Three credible explanations compete:

1. 1
   CECL forecast pressure. Management’s CECL model is forecasting deterioration on the absorbed Discover book that has not yet appeared in 30/60/90-day delinquency data.
2. 2
   Merger-cover discretionary reserves. The integration is enabling a “kitchen sink” quarter where unrelated reserves are quietly trued up under the cover of merger noise.
3. 3
   Macro input shift. The macroeconomic forecast in the CECL inputs (unemployment, GDP, household leverage) shifted meaningfully between Q4 2025 and Q1 2026.

The first two explanations are operational. The third is testable. Capital One’s NCO rate has historically tracked unemployment with a multi-quarter lag — toggle Unemployment rate under the macro overlay above to see the relationship over the last five years. Unemployment drifted higher through 2025–2026, the kind of macro shift that CECL term-structures translate into elevated provision well before delinquencies move.

What we’ll be watching

The signal will resolve cleanly over the next two quarters. We are tracking four watchpoints:

1. 1 · Delinquency formation

   If the provision build is forward-looking, 30+ day delinquencies should rise in Q2 or Q3 2026. If they don’t, the build is either kitchen-sink or model-driven, not symptom-driven.

2. 2 · Receivables trajectory

   A continued sequential decline pushes the read toward deliberate de-risking. A return to growth would suggest Q1 was an integration-driven one-off.

3. 3 · Peer provision response

   Synchrony, Bread Financial, and the consolidated Discover book all have meaningful subprime exposure. If they take similar provision builds in their next prints, the macro thesis strengthens. If COF is alone, the merger-cover thesis strengthens.

4. 4 · Earnings-call framing

   Whether management ties the provision to specific subportfolios (Discover-acquired vs. legacy COF), to macro inputs (CECL forecast change), or to discretionary judgment will be informative on its own.