Deep Investment Research10 min read
Gemini Card Revenue Rose 231%. Credit-Loss Provisions Nearly Consumed It: A Financial Gateway Is an Underwriting Business First
A bitcoin-rewards card looks like a low-friction distribution channel for an exchange. But when one quarter’s card CECL provision reaches $16.1 million against $16.2 million of card revenue, investors must ask whether growth is acquiring valuable customers or warehousing mispriced risk.
One Growth Number, Two Financial Statements
The Information’s public homepage highlighted identity fraud in Gemini’s second quarter. The company release shows card revenue up 231% to $16.2 million while transaction revenue fell 15%. The card is becoming central to the growth story.
The other side is less flattering: card-portfolio CECL provision was $16.1 million, while rewards, promotions and referral incentives were $8.7 million. Together they equal about 153% of card revenue. This is a stress test, not complete segment economics; it excludes funding, processing and service costs as well as incremental trading and custody gross profit.
| Q2 2026 | Amount/change | Investor read-through |
|---|---|---|
| Card revenue | $16.2m; +231% YoY | Financial services now drive growth |
| Card CECL provision | $16.1m | Nearly all quarterly card revenue |
| Rewards and promotions | $8.7m; +33% | Growth is subsidized |
| Managed card receivables | $219.6m | Risk sits on the balance sheet |
| Company net loss | $107.7m | Cross-sell has not covered companywide cost |
Crypto Rewards Are an Underwriting Contract, Not a Feature
The bull case is a high-frequency loop: spend, earn crypto, trade and custody. But Gemini’s 2025 10-K says it must purchase receivables offered under its WebBank arrangement and generally certain identity-theft, fraud and over-90-day delinquent receivables. It owns the customer relationship and the cost of being wrong.
Management says the quarter’s provision was concentrated in one Q1 identity-fraud cohort rather than broad deterioration. The test is future net charge-offs, delinquency migration, cohort performance, payback and incremental cross-sell—not the phrase “one-time.”
The Real Formula Is Risk-Adjusted Lifetime Value
Risk-adjusted LTV equals interchange and interest plus cross-sell gross profit, minus rewards, fraud, credit loss, funding and service cost. Card count and revenue disclose only half the equation.
First-half operating cash flow was negative $105.9 million, versus $188.6 million of quarter-end cash. Receivables turn quickly, but a thin economic margin can disappear when a single fraud cohort is mis-scored.
Three Master-Investor Stress Tests
Buffett, Munger and Howard MarksStudy what can destroy the economics before pricing growth
- Buffett: underwriting growth creates owner earnings only when loss cost is priced correctly.
- Munger: invert—synthetic identities, subsidy arbitrage, weak retention and dormant post-reward accounts can make every new customer destroy value.
- Howard Marks: growth is not the opposite of risk. Reward repeatable selection skill, not immature receivable balances.
The Countercase: A Fraud Event Could Create a Better Risk Asset
Gemini can combine exchange identity, asset balances, trading behavior and card spending. If the failed cohort produces persistently lower later loss rates, the provision may prove tuition for a risk-data moat.
More data is not automatically a better model. Watch provision as a percentage of average receivables, engagement after rewards normalize and incremental card-user gross profit through a full credit cycle.
Conclusion: Turn the Super-App Story Into a Unit-Economics Table
Card acquisition and rewards appear today; fraud, delinquency and charge-offs emerge later. The valuation mistake is capitalizing front-end growth immediately while calling back-end loss one-time.
Until later cohorts prove loss normalization, Valuefocus would value the card as underwriting rather than software distribution. Use the reverse DCF calculator and moat framework to turn growth, reward rate, loss rate and cross-sell into falsifiable assumptions.
Recent posts
%22%2F%3E%3Cpath%20d%3D%22M0%20610%20370%20370l300%20160%20400-330%20530%20360%22%20fill%3D%22none%22%20stroke%3D%22%23c99a62%22%20stroke-width%3D%2218%22%20opacity%3D%22.65%22%2F%3E%3Ccircle%20cx%3D%221070%22%20cy%3D%22200%22%20r%3D%2246%22%20fill%3D%22%23c99a62%22%2F%3E%3C%2Fsvg%3E)
%20-1.jpg?width=1600)
Frequently asked questions
How much did Gemini card revenue grow?
It reached $16.2 million in Q2 2026, up 231% year over year. Card-portfolio CECL provision was $16.1 million.
Does $24.8 million prove the card business lost money?
No. It is the sum of provision and rewards/promotions used as a stress test. Full economics also require funding, processing, service cost and incremental cross-sell gross profit.
How did management explain the provision?
It said the elevated provision was concentrated in one Q1 identity-fraud cohort, not broad portfolio deterioration. Later loss rates must verify that claim.
Do not value the growth rate alone. Put rewards, losses, cross-sell and capital cost into one falsifiable Valuefocus memo.
Download Valuefocus