Moving the Apple Card to a new issuer
On January 7, 2026, Apple named JPMorgan Chase as the replacing issuer for roughly 12 million cardholders and more than $20 billion in balances. That is the only announced fact. No terms, no economics, and no migration plan have been published, so whether the deal pays for itself is still an open question.
This chapter answers it, and it is the business case I would put in front of both companies: what each side earns per account, how 12 million cardholders move in waves, what the deposit relationship is worth, and what holds up when losses run higher than planned. Every figure comes from a 100,000-account re-underwriting simulation over 36 months, scaled up to 12 million to show what the model implies at full size.
The business case
Every figure is an output of my own simulation model, not a reported result. I state these per account and as a sample total, then scale that sample up by 120 to reach the 12 million cardholders Apple announced. That x120 version is shown in Section 5 as an illustrative projection, not an estimate of the real portfolio: these are synthetic accounts I constructed rather than a random draw from the real portfolio, so scaling them enlarges the construction and its assumptions together rather than estimating the portfolio. 100,000 is large enough for the Central Limit Theorem to hold, the rule that a sample's averages settle into a stable, predictable shape as it grows, so the three-way routing split and the four behavioral segments settle into stable percentages; below it the band edges wobble, and above it nothing changes except the arithmetic gets harder to follow. The announcement also gives more than $20 billion in balances, implying roughly $1,667 per account, and reconciling the model against that figure is the check I would run before quoting any portfolio total.
The 12M-scaled totals are presented in Section 5: Profitability at scale.
Each point is worked through in the sections that follow, with definitions in the glossary.
What is known, and what I am proposing
One fact is public. Everything else on this page is a design I am putting forward, and the line between the two matters more than any single number.
On January 7, 2026, Apple named JPMorgan Chase as the replacing issuer, covering roughly 12 million cardholders and more than $20 billion in balances. Goldman Sachs exits. That is the extent of what has been made public.
The three-outcome structure, the wave schedule, the risk score, the reward split, and every figure here. None of it is an Apple or Chase plan. It is what I would recommend if I were advising both sides.
Three outcomes, ten waves
Every account is re-reviewed and lands in one of three places. Two are products. The third is not a product at all.
Passes the new credit review and moves onto the relaunched card. Premium benefits, no annual fee, and no sign-up bonus, since welcome offers have always been for new applicants and everyone here already holds the card. The rate each cardmember pays is set by their credit grade.
Does not clear the bar for the Apple Card (2028), so routes here instead of losing access entirely. Builds credit on it, with a documented path to graduate onto a new Apple Card (2028) account later. Declined applicants who reapply arrive here too, not at the Apple Card, unless they clear the top range outright.
Clears neither bar, so no card is issued. Rather than an abrupt cutoff, the account follows a managed pathway to closure, with a written notice naming the reason and what would have to change to qualify. This is the outcome that has to be defensible in writing, and it is the subject of its own section. These are still future customers, and consumer finance is the one place Apple cannot afford a bad story, so a migration that reads as mass account closures writes that story for it.
How the rollout is stagedThe rollout runs in 10 waves across 18 months rather than all at once, which keeps the operational load and the support volume survivable, and gives the credit policy nine chances to be corrected before the last wave moves. Blended across all three outcomes, 54.0% of the original accounts are still active at month 36.
Why three outcomes and not twoChase Freedom Rise as the middle path is what I would argue hardest for. The easy version of this migration has two outcomes, approve or decline, and it is worse for everyone: Chase loses accounts it could have earned later, Apple loses customers from the ecosystem entirely, and the customer loses access at the exact moment they are least able to replace it. A middle path costs something to stand up and it is the difference between a re-underwriting and a purge.
The cardholder impactIt also matters for the customer in a way the economics do not capture. Closing an account ends the credit history attached to it, and length of history is one of the things a score rewards. Routing someone to Chase Freedom Rise keeps that relationship open and reporting. The smaller credit line will cost them something, but a lower limit is a scratch next to losing the account outright.
Who the portfolio actually is
What the groups are worthFour behavioral groups, and they are worth very different amounts to the issuer. The credit grade says how likely someone is to repay. The segment says how they actually use the card, and that is what decides whether the account earns anything.
The inversion is the finding. The largest segment by headcount is not the largest by value, and the volume block earns modest interchange spread thinly. Retention spend, cross-sell, and wave order all follow the second chart rather than the first.
Both charts are outputs of my own simulation across 100,000 synthetic accounts, not reported results.
The segment worth keepingThe Installment Revolver is the group a business case usually deletes, and I would keep it. On the card by itself it loses money, and that is the point. These are people who used the card to pay for an iPhone over time and then never take it out again. Chase earns almost nothing from them, and Apple sold the phone because of them. Look only at what the card earns and you cut the thing that moved the hardware, which is why the two sides have to be read together. It is also the group with the most room to improve. They already carry the card. They just have no reason to use it anywhere else, so a rate worth reaching for turns a financing account into a spending one, and that is the cheapest new spending the issuer can get.
An inherited book earns from month 1
A new card portfolio loses money before it earns any, a shape lenders call a J-curve. It has to buy customers, wait for them to activate, and set money aside against defaults that have not happened yet. This portfolio does none of the first two, because every account in it already exists and is already spending.
Cumulative contribution is positive in month 1 and never turns negative. The lowest point is +$0.39M, in the first month.
A payback period, how long a portfolio takes to earn back what it cost to stand up, is a useful number when a business buys its customers. Here it mostly reports that the customers arrived with the portfolio, so it says more about the structure of the deal than about this design.
The same distortion runs through the ratio of what a customer earns to what it costs to win one, which reaches 36.5:1 on the best segment against 3:1 as the healthy line. That is not efficiency. It is a $15 comms cost sitting in the bottom of a ratio built for businesses that pay to acquire people, and it would move the moment this program spent anything real on acquisition. It is the best-sounding figure this chapter produces and I would not put it in a headline.
I would rather name that than let the numbers flatter the case. The figures worth arguing about are the ones re-underwriting actually moves: who re-qualifies, who routes to Chase Freedom Rise, who is turned down, and what the surviving book earns.
The finding underneath it is simpler. A migrated portfolio and a launched one are different financial objects. The migrated one starts with revenue and no acquisition cost, and its risk is entirely inherited. Reading it against launch benchmarks, payback windows and acquisition efficiency, gets all three wrong.
The portfolio total is what Chase earns, and nothing else. What Apple spends on rewards and what it makes back on hardware and services are counted separately, and neither one changes this number. With two companies splitting the work, the easiest mistake is reading one company's profit as the value of the whole deal.
Profitability at scale
The 100,000-account model earns from month 1. Multiply those numbers by 120 to reach the 12 million cardholders Apple announced, and the full-scale picture comes into view.
These totals are the sample times 120, not a forecast. The model is built from 100,000 synthetic accounts, not a random slice of the real portfolio, so the numbers show what the construction implies at full size, not what Chase will earn. Reconcile per-account balances against the announced $20 billion before quoting any portfolio total as a result.
Re-underwriting to risk, on two dials
Goldman’s losses came from credit that did not match risk. There are only two dials an issuer controls here: the price of credit, and how much of it someone gets. The first chart is price, the second is exposure.
Riskier grades pay a higher rate, and every grade still earns money on its own. Bars are scaled to the top rate.
Chase hands out more credit overall and still loses less, because the limits move toward the people who can carry them. The two bars share one dollar scale so they can be compared directly. The percentages are not on that scale, because each one measures a different group against a different starting point, so they are written out rather than drawn. Anyone whose limit drops more than 20% gets a written notice explaining why.
Both charts are outputs of my own simulation across 100,000 synthetic accounts, not reported results. Bars are scaled within each chart, and the two charts are not on a shared scale.
The declines, and why behavior beats the score
This is the outcome that has to survive scrutiny, so it is the one I built most carefully. The finding is counterintuitive, because most declined accounts have a perfectly acceptable credit score.
The 58% : the number I would lead withThat 58% is the number I would lead with in a room, and not because it flatters the model. It is the number that proves the policy is not a score cutoff with extra steps. If a majority of declines clear the old bar, then the old bar was measuring the wrong thing, which is a plausible account of how the incumbent ended up with a portfolio it did not want. A decline is also the one decision here that a customer experiences as a judgment about them, so it should be the decision with the most reasoning behind it.
Who pays for the cash back, and why it changes nothing else
Every dollar of Daily Cash a cardholder earns has to come from somewhere, and Apple and Chase split the bill. There is no convention to fall back on here. As I understand the common airline or hotel co-brand, the issuer funds rewards by buying the partner’s miles or points, which makes the reward cost revenue on the partner’s side. Daily Cash is cash, so Apple has no currency to sell and that mechanism does not exist. Someone writes a real check, and who writes it is negotiated rather than inherited. I ran two versions of that split, and the interesting part is what stays exactly the same in both.
The customer agreement sets the reward schedule: 3% Daily Cash on Apple purchases, 2% on Apple Pay, and 1% on everything else. It does not say which party economically funds each tier. Public reporting indicates Apple paid for Daily Cash under the Goldman arrangement, but the precise allocation of reward cost between Apple and Goldman was never publicly disclosed, so this bar shows no split. The 75% and 55.8% splits in my proposal are modeling assumptions for the Chase case, not claims about what Goldman paid. Traditional co-brand cards work differently: the bank buys the brand's loyalty currency (airline miles, hotel points) at a negotiated wholesale price, so the bank plainly funds the reward and the brand gets paid up front. Daily Cash is cash, not a currency the issuer buys, so there is no wholesale purchase to point to. Funding is whatever the program agreement says, which is why the split was never disclosed and why reporting that Apple paid matters.
Who gets approved, who gets routed to Chase Freedom Rise, who gets declined, how many stay, what Apple earns on hardware: all identical in both versions. Changing the split only moves money between Apple and Chase, and it moves a lot of it, a $26.6M swing on identical credit decisions.
Asking Apple to cover 75% is an aggressive opening, and I would say so in the room rather than pretend otherwise. What makes it arguable is that Apple is not trying to earn on the card in the first place, so the funding it takes on is acquisition spend against hardware and services rather than a loss. What makes it negotiable is that nothing in the credit policy depends on the answer, so the two companies can settle the underwriting on its merits and then move the money between them without reopening any of it.
Worth being clear about which one the page runs on. Every figure in this chapter is calculated at 75% Apple funded. That 75% is the share of the Daily Cash earned on the Apple Card itself. Across the whole book it works out to 73.3%, because Chase funds Chase Freedom Rise rewards in full, including the months a graduate spends there before moving up. The tougher split is there to test the argument, not as an alternative set of results. Under it Chase pays more of the cash back, so its own numbers come down while nothing about the customer decisions moves.
Daily Cash is one place Apple pays into a card. There is another place it collects. When an iPhone user taps any third-party credit card in Apple Wallet, such as a Chase Sapphire Preferred or a Capital One Venture X, the issuing bank (the bank whose name is on the card) remits 15 basis points of the purchase to Apple. That is 0.15%, or 15 cents on every $100 spent, and $1.50 on a $1,000 airline ticket, plus about half a cent on each debit transaction. Google Wallet charges issuing banks 0% for the same tap on Android. On the Apple Card the relationship runs the other way, with Apple funding the majority of Daily Cash rather than charging the issuer for the tap.
That wallet fee is estimated to produce around $1 billion a year for Apple, and in September 2026 a federal judge certified a nationwide class of U.S. banks and credit unions to challenge it together, seeking both an end to the fee and repayment of what has been paid since 2022. Under regulatory pressure Apple has begun opening the iPhone's tap-to-pay antenna (the NFC chip) to rival wallet apps, and the plaintiffs argue that whichever wallet opens by default still decides where card spend goes.
The antitrust case hands Chase leverage. If the wallet fee is reduced or unwound, Apple's passive payments revenue shrinks and a scaled first-party card becomes more valuable to Apple, which strengthens Chase at the table on the Daily Cash split and on loss sharing. I would price that into the negotiation rather than into the model, since no figure on this page depends on it.
It also raises a fair question: why does Apple need a bank at all? Apple tried the other path. In 2022 it set up a wholly owned subsidiary, Apple Financing LLC, under an internal effort reported as Project Breakout, and used it to underwrite and fund Apple Pay Later, its short-term installment loans, off Apple's own balance sheet. Apple shut Apple Pay Later down in June 2024 and went back to partners, routing installment lending through banks and through Affirm. Even on small loans, underwriting, collections, disputes, and state lending licenses turned out to be a business Apple did not want to run.
Buying its way in is harder still. Under the Bank Holding Company Act, any company that controls an FDIC-insured bank becomes a bank holding company supervised by the Federal Reserve, with capital requirements, examinations, and consolidated reporting reaching the parent. Credit unions are member-owned nonprofits and cannot be bought by a public company at all. Spinning a chartered bank back out would mean giving up control of it, which leaves Apple negotiating with an outside bank again, exactly where it stands today. In the current antitrust climate, with the Department of Justice already suing Apple over smartphone competition, a bank acquisition would invite the fight rather than settle it.
So the partnership is structural, and it is the standard fintech pattern. X Money runs the same way, pairing X's wallet and money transmitter licenses with chartered partner banks that hold the funds and carry the regulatory weight. Block, Chime, and most neobanks do the same. The platform keeps the interface and the engagement, and the bank keeps the balance sheet, the credit losses, and the examiners. That is the seat Chase is bidding for here, and it is why the terms of the split matter as much as the product itself.
Can Chase earn its cost of capital on this book?
Goldman’s failure does not transfer automatically. The same portfolio can miss one issuer’s return bar and clear another’s, because the bar is set by what the issuer already owns. Chase brings card scale, an underwriting and collections operation that exists today, servicing it does not have to build, and a banking relationship it can sell across, so the fixed costs land differently and the customer is worth more than the card by itself. My model prices this migration at a positive return. This section is what has to stay true for that to hold.
Apple Card Monthly Installments puts a 0% balance on the issuer’s books. That balance costs funding, capital, servicing, and credit risk while earning no interest, so whether it makes sense depends entirely on what Apple pays the issuer for supplying it. None of that arrangement is public. I name it as the sharpest open term rather than a number, because a figure here would be invented.
The other levers a renegotiation would reach for are the same ones: how much of the reward cost Apple funds, who controls credit limits and line increases, whether the two companies share downside if charge-offs run above plan, and how far Chase can monetize the wider relationship.
Nothing changes on the customer-facing terms, and the issuer tightens quietly underneath: lower limits, fewer line increases, marginal accounts closed. A cardholder whose limit drops from $15,000 to $5,000 does not read that as a return-on-capital decision by a bank. They read it as Apple doing it to them, because Apple’s name is on the card and the notification lands in Wallet.
That asymmetry is the argument for the structure this chapter already proposes. A routed third outcome instead of a silent decline, and a written reason on every decision, so tightening reads as a policy a cardholder can see and answer rather than as something Apple did to them without saying so.
The money that never shows up on the card
What the card earns is not the whole story. What the card earns already counts the two things a card normally earns, the fee on every purchase and the interest on balances people carry. The savings deposits the card brings in are tracked separately, so they sit outside that figure rather than inside it. And it counts nothing at all of what Apple makes on the iPhones and subscriptions the card helps sell.
Why the loss is the pointApple’s per-account loss is the most misreadable number in this model, and the one I would put in front of a skeptic first. It looks like a failing product. However, it is a loss leader, which is one of the oldest plays in retail. Sell one thing below cost to move the things that carry the margin. Apple is doing that with a credit card instead of a grocery item, and services margins run well above hardware, so subscription revenue from a retained customer covers it fast.
The same trick in reverseThe deposit line is the same trick in reverse. As a spread it is noise next to what the card earns. As a wedge it is the only line here that makes Chase someone’s bank rather than one of their cards, and everything Chase could sell that customer next, Chase Total Checking, a mortgage, a Chase Sapphire card, Chase investment products, sits outside this model entirely.
What I cannot verifyWhat I cannot do is check either one. The Apple and Goldman revenue split was never disclosed, so the Apple side of this model is estimated rather than reported, and that is the honest limit on it. What I would defend is the shape. Both numbers look like failures in the place people look and pay for themselves in the place they do not.
What breaks it
A business case that only shows the good case is a pitch. So I broke the model on purpose to see what survives.
What a stress test asksThe thing most likely to break a card portfolio is people not paying it back. A stress test asks what happens if that gets much worse than expected, and the standard way to run one is to pick a number a regulator would recognize rather than one that flatters the result. I raised charge-offs, the share of balances written off as unrecoverable, by three percentage points across every credit grade. That is the order of magnitude I would treat as a downturn case, and it lands well above the 3.2% Chase currently guides for its own card business.
The reason to run it is that every figure on this page rests on losses behaving as modeled. If a modest worsening wiped out the return, the case would only hold in good weather, and nobody underwrites for good weather.
There is a second reason, and it follows from the case for re-underwriting rather than assuming the portfolio as it stands. The test is really asking whether the migrated portfolio holds up under pressure the legacy one did not, and if it does not, the routing thresholds are wrong rather than the stress being unfair.
Drawn to scale against the $100.34M base case. The portfolio stays comfortably positive, which is the result the underwriting is supposed to produce. The point of declining accounts on behavior rather than score is that the surviving portfolio is not fragile. Losses are the risk I can test, though, and not the one I would worry about most. If people leave faster than the model expects, each percentage point of retention costs about $1M, so being five points off would cost roughly as much as this entire loss shock. Loan losses are the risk with real history behind them. Retention is the one I am guessing at.
The recommendation
Three things decide whether this migration works, and they are the three the headline figures are grouped under: who carries risk, who stays, and who spends.
Risk, decide it once and decide it earlyReview every account again rather than moving them across as they are, and give people three possible outcomes instead of two. Move them in ten waves over 18 months, so the rules can be corrected nine times before the last wave goes. Chase does not want to inherit the cardholders who ran their limits up and stopped making even the minimum payment, because that cohort is why the Apple Card wrote off a larger share of its balances than comparable issuers.
Base the decision on how someone actually pays rather than their credit score alone, and be able to explain every rejection in one sentence the person would recognize as fair.
Loyalty, keep the people you would otherwise loseChase Freedom Rise is what keeps people who would otherwise walk away for good, and it costs very little to offer. 59.26% of the people sent there end up holding the Apple Card anyway.
Why retention compoundsThe customers worth keeping are already tied to Apple through their hardware and subscriptions, and that tether is what makes them reachable for Chase at all. Every account retained is an account Chase can sell its other products to, which is what the deposit argument rests on. Chase Freedom Rise makes that concrete. As the product stands today it requires a Chase checking account, so anyone routed there who does not already bank with Chase opens one as a condition of receiving the card. The product chapter proposes widening that, letting the Apple Card Savings account these cardholders already hold satisfy it instead, which trades a new checking relationship for a lower barrier to accepting the offer.
The institutional tension: product ladder or private creditRouting borderline applicants to Chase Freedom Rise protects Chase's balance sheet, and it exposes the central friction in any co-brand: Apple wants every approved customer holding an Apple Card, while Chase refuses to warehouse near-prime risk after watching what that cost the outgoing issuer. There is a second answer to that friction now being explored in the market, and it is worth putting next to mine.
JPMorgan Chase has been reported to be weighing second-look programs for its co-branded cards, funded by private credit firms such as Blackstone, KKR, Blue Owl, and Sixth Street. Chase would keep prime accounts on its own book, route declines into an outside vehicle funded by those firms, and collect origination and servicing fees without the Tier 1 capital charge or the lifetime loss reserve that comes with carrying the loans itself. Reported as an exploration, not an announced program.
Why a fee-free card breaks the private credit mathNear-prime lending normally pays for its losses with a stack of fees: an annual fee, a monthly maintenance fee, late fees, over-limit fees, returned-payment fees. Those charges arrive whether or not the customer revolves a balance, and they are what lets a lender absorb double-digit charge-offs and still clear its return. The Apple Card bans every one of them. "No fees. Really." has been the card's loudest promise since launch, and it is the part of the brand Apple would defend before it defended the reward rate.
Where it turns predatoryStrip out the fees and a fund underwriting to a low to mid teens net yield has one lever left, which is the purchase rate. On a near-prime book that means pushing toward the top of the legal range, around 30%, on a card the customer believes was designed to help them lead a healthier financial life. The brand says financial wellness and the credit economics say high-rate revolving debt, and the customer cannot see the difference from inside Wallet. That is the definition of a product whose surface and whose incentives point in opposite directions.
It also runs straight into the failure this migration exists to avoid. A near-prime borrower priced at 30% with no fee cushion defaults faster, which pushes the fund's required yield higher, which raises the rate again. That is the charge-off spiral the outgoing issuer lived through, relocated off the balance sheet and left running. Freedom Rise is slower and less elegant, and it caps the line, prices closer to prime, and gives the customer a documented way out after 12 on-time payments. I would rather defend that in front of a regulator.
Growth, the part that does not happen at launchKeep marketing the card after the migration is done. Better approvals and better rewards decide who holds it and what they earn, but neither gets it out of the drawer. Chase only collects the merchant fee if people actually spend, and the group who only finance hardware are the easiest ones to win over, because they already carry the card.
What the migration leaves behindThe migration also leaves something behind that is worth more than the portfolio itself. Re-underwriting 12 million existing accounts produces a record of which behaviors actually preceded losses on this specific portfolio, and that record is what should set the bar for everyone who applies afterward. That is the one chance to calibrate before opening the doors again, which is what the acquisition offer in chapter 1 is built on top of.
How I would negotiate itGo into the negotiation asking Apple to cover 75% of the cash back, but treat that as a price rather than a requirement. None of the credit rules depend on it, so it can be traded away without reopening anything that matters.
The limits of this chapter
The accounts are synthetic. Nobody outside Apple and Chase has the real portfolio, so the 100,000 accounts are simulated against published behavior and rate data. That means the mechanics here are testable and the specific dollar outputs are not verifiable against reality. Treat the structure as the contribution and the figures as what that structure implies under my assumptions.
The 12 million figures are arithmetic, not a forecast. Multiplying a synthetic sample by 120 scales the construction and its assumptions together, so the full-scale totals inherit every assumption without adding evidence. Read them as what the model implies at size, not as what Chase will earn.
One announced fact, one modeled everything. The issuer change is public. The product lineup, the routing rules, and the reward split are all mine. A reader who takes this as reporting on Apple's plans has misread it, which is why the distinction sits in the first section rather than a footnote.
Response to an incentive is unmeasured. Nothing here prices what an offer buys, in retention or in acquisition, because I could not source a figure for it and the record does not contain one. That is why the acquisition offer in chapter 1 is specified as a structure with its match rate flagged for testing rather than costed here. Sizing either would need a live holdout, which is the first thing I would instrument.
The hardware baselines are stale, in the conservative direction. Apple raised prices on several lines after the model was built. I left the pre-increase baselines in and reported the sensitivity rather than quietly restating the headline, because a figure that moves when you refresh an input should be labeled, not improved. Raising the baseline 10% is the sensitivity I ran.
Retention is modeled, not observed. The 63.8% and 47.1% figures come from my own churn assumptions by product, so if those hazards are wrong the retention economics move with them.
The calibration argument cannot be tested on this data. Re-underwriting a real portfolio would teach an issuer which behaviors preceded real losses, and that is the case I make for it. This model cannot demonstrate it, because simulated accounts default according to the hazards I assigned them, so scoring them against their own performance recovers my assumptions rather than anything about credit. The argument is a reason to run the exercise, not evidence from having run it.
What Apple earns is estimated, not disclosed. The Apple and Goldman revenue split was never published in either company’s filings, so the Apple side of this model is built from what the card plausibly moves in hardware and services rather than from anything reported. That is the weakest-sourced half of the two-ledger argument, and the half I would most want a practitioner to challenge.
Partner acceptance is assumed, and it is the softest assumption here. The whole three-outcome design rests on Apple agreeing that some approved applicants can walk away holding a Chase-branded starter card. A co-brand partner has every reason to refuse that, because it hands the issuer a direct retail relationship with a customer Apple recruited. If Apple holds the line and insists every approval carries the Apple nameplate, the routing tier collapses into two outcomes, and Chase is left choosing between carrying near-prime risk itself or funding it through an outside private credit vehicle priced near the top of the legal rate range. Neither of those is what this model recommends, and the model does not price either one.
How the numbers were produced
Glossary
Independent professional-development project. The migration design is a proposed concept built on synthetic account data, not an announced plan, and it is an Apple-inspired concept rather than an official Apple product. Not affiliated with, authorized by, or endorsed by Apple, JPMorgan Chase, Goldman Sachs, Visa, or Mastercard. The public facts relied on are listed in the sources. Every modeled figure is an output of my own simulation, which is not published, and none of it is legal, financial, or investment advice. © 2026 Darnell Wilson. All rights reserved. The writing, analysis, and financial models in this case study are my own original work. Company names, card products, and trademarks referenced belong to their respective owners and are used for identification and commentary only. Please do not reproduce, republish, or reuse this material without written permission.