The Variable That Moved
Code does not lie, but it does hide. Last month a rewards ledger changed a single variable โ the earn rate on a consumer credit card โ and the change propagated further than the press release admitted. Square, now Block, began paying 3% back in bitcoin on purchases made through its card product. On the surface it is a marketing line, the kind of sentence a payments company issues to manufacture a news cycle. Underneath it is a settlement problem, and settlement problems are where I spend my working hours.
Here is the anomaly that made me open the editor. A credit card rewards program does not create value; it redistributes it. The pool that funds a reward is drawn from interchange โ the merchant discount rate a card network charges a merchant, typically between 1.5% and 2.5% for a mainstream consumer card in the United States. A 3% reward therefore exceeds, by design, the revenue that nominally funds it. The gap must be covered by something other than interchange: either a direct subsidy from Block's balance sheet, or an underwriting assumption that the customer's lifetime value โ spread across Cash App, lending, and bitcoin trading spread โ will exceed the present cost of the reward. I have audited enough token emission schedules to recognize the geometry. This is an incentive curve with a finite emission budget, wearing a card perk's clothing. The only question that matters is how long the emission runs.
Where Reward Money Actually Comes From
To read the announcement properly you have to understand the machine it plugs into. Block, formerly Square, is two companies fused at the balance sheet. The first is the merchant acquiring business: Square terminals, point-of-sale software, small-business lending. The second is Cash App, a consumer wallet that has become one of the largest retail distribution channels for bitcoin in the United States. The two share a treasury, a brand, and increasingly a single thesis โ that the future of money is a bearer asset settling on a public ledger, with Block positioned as the on-ramp.
The credit card is the bridge between those halves. A card is not a product; it is a two-sided market. On one side sits the cardholder, who is paid in points or cash for loyalty. On the other side sits the merchant, who pays interchange for the privilege of accepting the card. The network โ Visa or Mastercard โ rents its rails to both and takes a slice of the spread. The issuer, here Block's banking partner, carries the credit risk. What most people call "rewards" is a rebate of interchange, returned to the cardholder to induce more spending, which generates more interchange. It is a closed loop with friction, and the friction is where margins live.
The industry baseline is 2% cash back, a number stable for a decade because it approximates the sustainable ceiling of interchange-funded rebates on a no-annual-fee card. Premium cards exceed 2% only by charging annual fees or by imposing higher merchant rates on business and travel categories. A no-fee, 3%-in-bitcoin card therefore sits above the ceiling, and any engineer who has watched a system run above its thermal limit knows what happens next: the deficit is financed by an external energy source, and the question is not whether that source is sufficient, but for how long.
This is not the first time the industry has tried this. Between 2018 and 2021 a wave of crypto rewards cards launched โ BlockFi's, Gemini's, Crypto.com's, Coinbase's โ each promising a percentage back in a volatile asset, each funded by a mix of interchange and token issuance. Most of them failed for the same two reasons: the funding source was circular, and the bear market converted a generous reward into an unaffordable liability. The cards that survived did so by cutting the rate, capping categories, or migrating behind a subscription. This is the base rate, and it is the most honest prior you can bring to Block's announcement.
Block's own balance sheet changes the calculation at the margin. The company has held bitcoin on its books since 2020 and has repeatedly added to the position. Cash App's bitcoin revenue is reported gross, which flatters the top line but not the margin, and bitcoin gross profit is a meaningful share of Cash App's economics. In other words, Block is not a neutral payments company that happens to accept bitcoin. It is a bitcoin-exposed enterprise with a payments business attached. A rewards card that converts consumer spending into bitcoin accumulation is, for Block, a mechanism to deepen that exposure on the customer side while hedging it on the corporate side. The incentive alignment is real, and so is the conflict embedded in it: the company profits from the asset it is paying you in.
Now the structural observation that matters most. The "3%" is not a market-clearing price. It is an administered number, chosen for narrative effect and adjustable at management's discretion. This is precisely the species of arbitrary parameter I have spent years flagging in DeFi lending markets. Aave and Compound publish interest rate models with slopes and kinks that look empirical and are, in practice, governance-set constants with no more grounding in real supply and demand than a policy rate set by committee. A credit card rewards rate is the same artifact. It is a knob. Knobs get turned. When a customer reads "3% forever," they are reading a default value in a config file that a future product manager can overwrite, and history says the overwrite is coming.
The Settlement Stack, Dissected
Now the part the press release skipped: the mechanics of paying a bearer asset as a rebate on a fiat liability.
A card purchase settles in dollars. The cardholder swipes, the merchant receives dollars minus interchange, the issuer records a receivable, and the cardholder accumulates a claim on the issuer. When that claim is denominated in bitcoin, the issuer has introduced a second asset into a single-asset settlement pipeline, and every stage of that pipeline now needs a price.
Consider the accrual path as pseudo-code:
on purchase(p):
usd_reward = p.amount * 0.03
btc_reward = usd_reward / oracle.price(BTC/USD, t_purchase)
ledger.credit(user, btc_reward, lot_id = new_lot(t_purchase))
treasury.liability += usd_reward
emit RewardsAccrued(user, btc_reward)Read the third line carefully. The reward is denominated in bitcoin, but its value is fixed in dollars at the moment of purchase. The user is not earning bitcoin; they are earning a dollar amount that is immediately converted into bitcoin at the prevailing oracle price. The dollar figure is the invariant. Bitcoin is the delivery vehicle. The user bears price risk from the instant of conversion forward, and Block bears the conversion and custody cost.
Two consequences follow. First, there is an oracle dependency. The conversion uses a reference price โ presumably a composite of major spot venues โ and that price has latency. Between the timestamp of the purchase and the timestamp of the conversion, the market moves. On a portfolio processing millions of transactions a day, even seconds of oracle staleness aggregate into a measurable basis. I built a testnet during DeFi Summer 2020 specifically to demonstrate how price-oracle manipulation under liquidity stress could be arbitraged against invariant math, and the lesson generalizes: any system that converts one asset into another at a quoted rate is exposed to whoever can move the quote. The saving grace here is that Block controls both the oracle and the conversion, so the exposure is internal โ a hedging and treasury problem, not a flash-loan attack surface. But internal is not absent. It is moved onto the issuer's book, where it surfaces as slippage rather than exploit.
Second, there is a custody question, and here I will be blunt. "Root keys are merely trust in hexadecimal form." The user never holds a key. The reward accrues inside Cash App's custodial ledger, which means the user holds an IOU denominated in a bearer asset. This is not unique to Block โ it is the default architecture of every exchange and every fintech wallet โ but it matters for the narrative. A program that pays in bitcoin while holding the bitcoin for you is not adoption of self-sovereignty. It is adoption of bitcoin exposure, mediated by a custodian that can freeze, rehypothecate, or lose the underlying. The ethos and the product are different objects, and conflating them is a category error the industry commits every cycle.
Then there is the tax ledger, where consumer-facing crypto products quietly become compliance products. In the United States, a bitcoin reward is ordinary income at fair market value on the date of receipt, and each reward creates a new cost basis. When the user later sells, swaps, or spends that bitcoin, a capital gain or loss is realized per lot. A daily-spending cardholder could generate hundreds of micro-lots a year, each with its own basis and holding period. The issuer must maintain lot-level accounting, and the user inherits a reporting burden that scales with their coffee habit. Most will not notice until January. Some will notice and stop using the card. The retention curve of this product is, in part, a function of how gracefully Block handles that accounting. It is a software problem, and software problems are solvable, but they are not free.
Now the delivery layer. When a user withdraws bitcoin from Cash App to a self-custodied wallet, the transaction settles on Bitcoin's base layer and pays a fee that is a function of block-space demand, not of the reward's size. A three-dollar reward withdrawn on-chain during a fee spike can be worth less than the cost to move it. This is the practical argument for keeping balances custodial, and it is also the honest limit of "bitcoin payments" as a consumer story. Off-chain solutions โ Lightning, custodial internal transfers โ exist precisely because the base layer does not scale to coffee-sized transfers at consumer frequency. And here I will state a position I have held for years: the overwhelming majority of what markets call "Bitcoin Layer 2" is Ethereum infrastructure wearing a bitcoin costume, built by teams whose technical lineage, tooling, and incentive models are Ethereum-native, rebranded to capture bitcoin's liquidity and narrative. The real bitcoin community does not acknowledge most of them, and it is right not to. When Block pays a reward "in bitcoin," the honest description of the rail is a custodial ledger entry, occasionally reconciled to base-layer UTXOs when the treasury rebalances. The bearer asset is real. The immediacy is a database.
The scaling layer deserves one more note, because it is easy to overstate and just as easy to dismiss. Rollup systems that post data to Ethereum blobs are on a trajectory toward saturation โ my own estimate is that blob space will be effectively full within roughly two years of Dencun, after which rollup fee floors rise, and the second-order effect is that every application pricing itself against cheap data will reprice. Bitcoin's scaling story does not share that mechanism, but it shares the conclusion: settlement capacity is a scarce resource, and any product that assumes abundant cheap settlement is borrowing against a future that has not arrived. A rewards program that quietly assumes its withdrawals will be cheap is making exactly that loan.
Consider a worked example. Assume a cardholder spends $2,000 a month, a plausible figure for an affluent urban user. At 3% that is $60 of bitcoin rewards per month, or $720 a year. If interchange nets Block roughly 1.8%, the cardholder's spending generates $36 a month in gross interchange, leaving a $24 monthly deficit before operating costs. Scale that across a million active cardholders and the annual subsidy runs into the hundreds of millions of dollars, before a single default, fraud charge, or hedging cost is accounted for. This is the arithmetic that governs the product's life, and no marketing line changes it.
Block's mitigation is treasury management. The company already holds bitcoin, so rewards can be funded from existing inventory rather than open-market purchases, which removes the buy-side price impact but not the mark-to-market risk. A sophisticated treasury would hedge the accrual with derivatives โ options or futures โ to lock the dollar cost of the reward liability. That hedging is not free; it is a spread paid to a counterparty, and it belongs in the unit economics of the card. The industry rarely discloses it, which is why so many crypto rewards programs looked profitable until the market moved.
Let me put a probability on the sustainability question, because vague warnings are useless. Given Block's disclosed bitcoin holdings and the historical contribution of Cash App's bitcoin gross profit, my model assigns a roughly 55% probability that the 3% headline rate is reduced or category-restricted within 24 months, a 30% probability it holds at 3% but migrates behind an annual fee or subscription tier, and a 15% probability it expands, which would require either a material rise in interchange economics or a decision to run the card as a customer-acquisition loss leader against Cash App's broader P&L. Those are not forecasts of failure. They are a distribution over parameter paths, and the single most likely path is the one every rewards program in history has taken: an introductory rate that drifts down as the acquisition budget is exhausted.
Architectural Autopsy: The Blind Spot Nobody Prices
Now the counter-intuitive part, and the section I run on every hyped protocol I examine.
The consensus reading of Block's announcement is that the risk is bitcoin's volatility. That reading is wrong, or at least mis-prioritized. Volatility is a treasury and hedging problem, and Block is one of the few companies on earth with the balance sheet and the market-making relationships to hedge it. The real exposure is architectural, and it has three layers.
First, the funding model is an unamortized liability with a marketing label. A rewards program that pays above its interchange ceiling is, in accounting terms, a deferred customer-acquisition cost. It is capitalized optimism. As long as Cash App's engagement metrics justify the spend, the subsidy is rational. The moment engagement plateaus โ and consumer fintech engagement has plateaued across the sector โ the subsidy becomes a drag, and the first lever management reaches for is the rate. This is not a prediction of malice; it is the mechanical consequence of a system running above its funding limit. Every token project that promised a fixed APY eventually met the same constraint. The difference is that Block cannot pass a governance vote to change the terms retroactively without a customer-relations cost, so the adjustment will be quiet: category exclusions, monthly caps, a "program update" email. Watch the terms, not the headline.
Second, the compliance layer is the actual product, and it is surveillance. A custodial bitcoin rewards card is, structurally, one of the most effective financial-surveillance instruments ever deployed. It links a verified identity to a spending graph and to a bitcoin position, in one system, with full transaction-level visibility. From a regulator's perspective that is a feature. From a privacy perspective it is a panopticon with a cash-back incentive. I am not moralizing; I am describing the architecture. "Security is a process, not a product," and so is compliance. The card is a sensor. The reward is the payment for installing it. Users who understand this will still use the card, because 3% is real and the surveillance is diffuse. But they should understand what they are trading, and the industry's habit of describing that trade as "adoption" flatters everyone involved.
Third, the integration risk is concentrated in a single entity. Block controls the oracle, the custody, the ledger, the card, and the banking relationship. That vertical integration is efficient and fragile. A failure in any one layer โ an oracle outage, a custody incident, a regulatory action against the card-issuing bank โ propagates to all of them, because there is no seam at which the system can shed load. Distributed-systems engineers call this a single point of failure. Marketers call it a seamless experience. Both are correct. "Velocity exposes what static analysis cannot see" โ and the velocity that matters here is not the price of bitcoin. It is the rate at which rewards accrue against a fixed subsidy budget, and the rate at which the terms can be rewritten beneath the user.
The blind spot, in one sentence: the market is pricing this as a bitcoin-adoption event when it is a customer-acquisition event, and acquisition events have a lifecycle that ends in either retention or repricing. The bitcoin is the bait. The card is the hook. The P&L is the fisherman.
The Forward Curve
So what do I actually expect, and what should you watch?
The program will launch, generate a genuine uptick in card issuance, and produce a wave of "bitcoin goes mainstream" commentary that conflates a custodial rewards ledger with monetary adoption. Over the following quarters, the signal to track is not the rate but the fine print: category exclusions, monthly caps, minimum redemption thresholds, and the ratio of custodial balances to on-chain withdrawals. If withdrawals are negligible, the product is a deposit-gathering machine with a bitcoin veneer. If withdrawals are material and on-chain, the treasury's UTXO management becomes a genuine operational discipline worth auditing.
My probability-weighted expectation is slow normalization: the headline stays, the effective yield drifts toward 2%, and the product becomes indistinguishable from a cash-back card with a crypto skin โ which, economically, it always was. "Infinite loops are the only honest voids." Every rewards curve converges to the funding source that sustains it. The question is not whether Block's 3% is generous. It is whether the company can afford to be generous for longer than its customers can afford to be patient โ and which of the two runs out first.