Apple's Profit Risk in the AI Agent Era
An Analyst-Grade Deep-Dive Report

A slow burn, not a collapse: agents erode the "take a cut of every transaction" channel power, burning 10–22% of Services gross profit over five years. What matters most isn't how big the loss is, but that the loss is a one-time deleveraging — once the tax base is bypassed, the revenue is lost every year.

0.1 Headline Numbers (2031 Annualized Run-Rate; Target-Year Scenario Model Output (Assumed))

Note: Net gross profit loss = commission loss + TAC loss − offsets (only the aggressive scenario deducts incremental revenue from WeChat mini-programs at 15% (Estimated), $0.4B). Revenue loss and gross profit loss are approximately 1:1 — both tax bases carry gross margins near 100%; this is the iron law of the model (Assumed; see §7 for rationale).

0.2 What the Risk Pool Looks Like

0.3 Structural Conclusions (Three "Largest"s)

1. Largest category exposure: gaming. 53.5% of the commission pool ($13.7B (Estimated)); ~49% of commission losses under the base scenario. What agents eliminate is discovery cost (users today don't even know Supercell Store exists) and redirect friction — not the price signal. The +10% bonus has always been there; agents simply multiply the number of people who act on it.
2. Largest geographic exposure: the United States. 39.4% of the commission pool ($10.1B (Estimated)); ~58% of commission losses under the base scenario. Since the 2025-04-30 Epic contempt ruling, in-app external-link purchases in the U.S. carry 0% commission (Disclosed; ruling text as restated by the Ninth Circuit). This is the model's most sensitive binary variable: the Supreme Court granted certiorari on 2026-06-30; if the zero-commission order is overturned, base-scenario 5-year losses drop by $2.1B (−15%) (Assumed).
3. Largest "black hole" variable: the true bypass rate is unobservable today. The 2025-04-30 ruling bars Apple from requiring developers to report off-app purchases — Apple itself has lost the means to observe it (Disclosed). The known Appfigures datapoint, "U.S. commission revenue down −18% since the start of 2026" (reported by FT on 2026-08-18 (Estimated)), is a commission-revenue figure, not a bypass rate — it mixes fee cuts with slowing billings and cannot be converted directly.

0.4 Offsets and Keep-an-Eye-On Variables (Don't Miss These)

• WeChat mini-programs at 15% (2025-11 Apple–Tencent agreement (Estimated)): a new commission source in China, conservatively modeled as an offset of "5% / 10% of the China commission pool" (3-year $0.2B / 5-year $0.4B, aggressive scenario) (Assumed).
• TAC keep branch: In 2026-01 Apple announced Siri/Apple Intelligence would switch to Google Gemini (Apple paying Google ~$1B/year in reverse (Estimated)). If agent backends still run through Google's search/ads pipeline, TAC losses could approach zero; the erosion scenario only holds if agents pull the search-and-transaction loop into their own hands (Assumed).
• Apple's defensive cards: commission cuts to retain developers (the EU has already cut IAP from 30% to 26%); on 2026-08-14 Apple proposed U.S. external-link rates of 15%/5%/10% to the Northern District of California (pending approval; not approved as of 2026-09-23 (Disclosed)).

Apple isn't a hardware company — it's a company that uses hardware to sell channel tax: Services is only 26% of revenue yet contributes 42% of gross profit. Those 16 points of leverage are what agents are coming to burn.

Source: Apple FY2025 Form 10-K, consolidated income statement (p. 29) and MD&A "Gross Margin" table (p. 23). Units: $ millions.

Three Foundation Facts

1. Services is only 26.2% of revenue, yet contributes 42.2% of total gross profit ($82.314B ÷ $195.201B).
2. Apple does not disclose Services operating profit (it only discloses reportable segments by geography). Estimated under two cost-allocation methods (Assumed): the high method (all R&D to Products, SG&A by revenue share) gives $75.1B; the low method (all operating expenses by revenue share) gives $66.0B — Services operating profit is roughly half of company operating profit (49.6%–56.4%).
3. Trend: Services gross margin FY2023 70.8% → FY2024 73.9% → FY2025 75.4% → first nine months of FY2026 76.3% (Disclosed); FY2026 Q3 single-quarter Services revenue $30.739B (+12.2% YoY), gross profit $23.2B. But note that FY2026 Q3 company gross margin of 50.1% includes ~2pp of one-time tariff-refund benefit (10-Q MD&A); ex-refund it's ~48.1%.

Implication: Any erosion of the Services take rate hits roughly 2× as hard on operating profit as on revenue — that's the profit leverage of agent risk.

Breaking the $109.2B apart line by line: only two lines can actually be touched by agents — App Store commissions and Google TAC. The rest (AppleCare, iCloud, Music…) is recurring revenue that agents can't reach.

Apple never discloses revenue for any individual Services line. Everything below is third-party estimates and back-calculations, summing to a midpoint of ~$96.8B (Estimated); the ~$12.4B gap to the reported $109.2B goes into "other/residual" — a gap of ~11%, within acceptable range.

Key inputs (risk side):

• Pure-profit pool directly threatened by agents = App Store commissions (~$27B–$30B) + Google TAC (~$20B–$25B) ≈ $47B–$55B of revenue, almost entirely gross profit, or 57–67% of the $82.3B Services gross profit (Estimated).
• Indirect threats: Apple Pay (agents routing through the cheapest channel could bypass it); advertising (agents comparison-shopping for users erodes search-ad value).
• Relatively safe: AppleCare (hardware-tied), iCloud (storage is a necessity), Apple Music (subscription habit) — low agent substitution elasticity.

The $26B isn't monolithic: gaming is over half, the U.S. is ~40%. Agent firepower naturally concentrates in those two cells — while China is ironclad (m=0) and the EU only erodes the spread.

3.1 Scale and Basis (Important: Two Bases Coexist)

3.2 Category × Region Matrix (Commission-Revenue Basis; Total ≈ $25.7B ≈ $26B)

Category mix (share of $115B billings): gaming IAP 45% ($52B, dual-sourced from BofA/Goldman) / non-gaming subscriptions 38% ($44B, estimated, range 35–42%) / other digital goods 17% ($19B, residual) (Estimated/Assumed). Region mix (Analysis Group 2025 regional data: U.S. $62B / China $21B / Europe $24B / Japan $15B; EU estimated at 70% of Europe (Estimated/Assumed)).

Blended effective commission rate ≈ 22.3% ($25.7B ÷ $115B), far below the nominal 30% — reflecting the blended reality of 15% for small businesses, 15% subscription renewals, 0% U.S. external links, and EU fee cuts (Assumed).

3.3 Region-by-Region Rules (As of 2026-09-23; Verified Item by Item)

• United States: On 2025-04-30 Judge Yvonne Gonzalez Rogers held Apple in contempt, barring Apple from collecting any commission on U.S. external-link purchases (currently 0%); on 2025-12-11 the Ninth Circuit upheld the contempt finding but vacated the "blanket collection ban" and remanded; on 2026-08-14 Apple proposed new rates of 15% / 5% for small businesses / 10% for subscription renewals and partners, pending district-court approval as of 2026-09-23 (Disclosed). The Supreme Court granted certiorari on 2026-06-30 (narrowed to a single legal question), with the parties' agreed briefing schedule pointing to oral argument in the first week of 2027-01 at the earliest, decision expected before 2027-06 (Disclosed).
• European Union: Old terms (17%/10% + €0.50/install CTF) superseded twice over. Since 2026-01-01 the CTF became a uniform 5% CTC; announced 2026-08, effective 2026-10-01, a new four-tier flat commission: IAP 26% (15% if eligible) / alternative payments 20% (10%) / external links 15% (10%) / third-party stores & web distribution 5% CTC (Disclosed). On 2025-04-23 the Commission fined Apple €500M for anti-steering (Disclosed).
• Mainland China: Still full 30%/15%, no external links, no third-party stores (Disclosed). SAMR antitrust probe rumored 2025-02; complaints in 2025-10 (55 consumers) / 2026-06 (48 developers); no case filed and no policy change as of 2026-09-23 (Disclosed). New variable: the 2025-11 Apple–Tencent agreement — WeChat mini-program/mini-game IAP processed by Apple at 15% (Apple previously earned nothing from this ecosystem; Bloomberg called the deal "worth billions of dollars") — a new commission source in China, not a fee cut (Estimated).
• Rest of world: Predominantly standard 30%/15%. After Korea's 2021 legislation Apple charges 26% on third-party payments, but total cost exceeds 30% so actual adoption is minimal (Disclosed); Japan's Mobile Software Competition Act (MSCA) took full effect 2025-12-18; rate details not independently verified (see unverified list) (Disclosed/Unverified).

The bypass channels are already paved; agents just turn "people who don't mind hassle" into everyone. Four layers of friction removal, each pushing the bypass rate toward the ceiling.

4.1 Channel Inventory (Hard Instances, Not Theory)

Gaming IAP: Supercell Store (store.supercell.com, covering Clash of Clans/Brawl Stars etc., +10% bonus gems on every purchase), Scopely/Warner Bros official web stores, DTC payments infrastructure like Appcharge (directly pluggable for small/mid developers) (Disclosed). In-app direct external links in the U.S. (in theory games could also place zero-commission external links after the 2025-04-30 ruling), but verified large-scale adoption instances concentrate in subscription/reader categories (Assumed).

Subscriptions: Spotify iOS v9.0.40 (approved 2025-05-02, full in-app pricing + external link to its website), Patreon iOS v125.5.0 (2025-05, U.S. users default to web checkout), Kindle "Get Book" orange button (2025-05), Netflix reader external links (global since 2022) (Disclosed).

4.2 Three Tables: Technically Bypassable / Current Actual / Agent-Era Ceiling

4.3 Four Layers of Friction Eliminated by Agents (This Study's Core Mechanism)

4.4 Appfigures −18%: Existence Verified (But Don't Misread It)

• The number exists: Appfigures data, reported by FT 2026-08-18 and relayed by MacRumors: "Apple's U.S. commission revenue has fallen 18 percent since the start of 2026" (Estimated).
• Cross-corroboration: Sensor Tower says U.S. App Store consumer spend fell −6% YoY in 2026-Q2; Morgan Stanley: App Store net revenue down −0.6% YoY through mid-August, the first decline in four years; Apple's 2026-06 quarter earnings call acknowledged regulation starting to drag on Services growth ($30.7B vs. $31.4B expected) (Estimated/Disclosed).

Apple collects ~$20B a year from Google — for doing nothing, simply because Safari's default search engine is Google. It's the purest rent in all of tech, and the pipe agents most want to bypass.

5.1 Scale: $18B–$20B/Year (Modeling Base: $20B)

Verification conclusion: All public estimates across 2022–2026 converge on $18B–$20B/year; the 2022 $20B is the trial-disclosed figure for that year, with no new disclosures since (Unverified: actual 2025/2026 payment amounts). At $20B: ~18% of FY2025 Services revenue and 24% of Services gross profit — the second-largest pure profit in Services after App Store commissions (Estimated).

5.2 Contract Status: Ongoing; the 2025-09 Remedies Ruling Preserved Payments

• 2024-08-05: Judge Mehta ruled Google an illegal search monopolist (remedies tried separately) (Disclosed).
• 2025-09-02/03 remedies ruling: rejected DOJ's "ban all payments" request — Google may keep paying for default placement, but on non-exclusive + annually renegotiated/switchable-default terms. Mehta's words: "Cutting off payments from Google almost certainly will impose substantial — in some cases, crippling — downstream harms to distribution partners." Market reaction: Alphabet +8% after hours, Apple +4% (Disclosed).
• The ISA (Information Services Agreement, fully confidential) runs at least through September 2026, renewable unilaterally by Apple for two more years (→ September 2028); no public reporting of any post-2024-08-ruling change to payment amounts/ratios (Disclosed/Unverified). DOJ appealed the remedies ruling in 2026-02 (Disclosed).
• Where it sits in the filings: buried in Services' licensing sub-line; Apple never breaks it out (Disclosed).

5.3 Agent Erosion Scenarios (Base $20B/Year; Pass-Through Coefficient 1.0)

Transmission chain: personal agents answer directly (no Google results page shown) → users bypass google.com / Safari address-bar search → Google search-query volume on Apple devices falls x% → search-ad impressions/clicks fall ≈ x% → the ISA revenue share (tied to Safari search-ads revenue) shrinks proportionally → Apple's annual TAC revenue falls ≈ $20B × x% × pass-through coefficient (Assumed).

Rationale for coefficient 1.0: Bernstein/Morgan Stanley back-calculations both treat TAC as a fixed share of Safari search-ads revenue (the revenue-share contract structure), and ad revenue is roughly proportional to monetizable query volume. Sensitivity: if agents mainly replace informational queries (low monetization), coefficient <1; if they replace high-commercial-intent queries like shopping/local, coefficient >1 (Assumed).

5.4 The Biggest Uncertainty: If Agent Backends Still Run Through Google's Pipes, TAC May Survive

Whether erosion happens depends on where agents' "answers" come from and where the money flows:

1. Keep branch: the agent retrieval layer calls the Google Search API / Gemini grounding on Google Search, and commercial queries still trigger Google ad auctions → TAC survives or even grows. Real-world support: Apple announced in 2026-01 that Siri/Apple Intelligence would switch to Google Gemini (Apple paying Google ~$1B/year in reverse — money flowing the other way) (Estimated). If Apple-ecosystem agents' backends are Google models + retrieval, TAC's foundation is reinforced, not weakened.
2. Erode branch: agents answer from their own indexes/model weights, or via Bing/other retrieval, or close the transaction loop inside the agent (ordering for users through the cheapest compliant channel) — Google gets neither queries nor ad impressions → TAC shrinks per the §5.3 scenarios.
3. Hybrid: informational queries get eaten by agents (low monetization) while commercial queries still flow back to Google → coefficient <1, actual losses smaller than the query-replacement share.

Target-year annualized run-rate, fully transparent formulas. Anyone can reproduce the same numbers from §6.3's formulas and assumptions — that's the fundamental difference between this report and a typical sell-side note.

6.1 Master Model Output Table (model.py Actual Output, Precise to 0.1B)

(model.py has been re-run; output matches this table. Detail table: parts/model_assumptions.csv.)

6.2 Base-Case 5-Year Loss Structure (Where the $13.5B Comes From) (Assumed)

By category: Gaming $5.1B (~49%) / Subscriptions $3.3B (~31%) / Other $2.0B (~20%).
By region: United States $6.1B (~58%) / Rest of world $3.7B (~36%) / China $0.0B / European Union $0.7B (~6%, computed on the 26%→15% spread).
(model.py detail: gaming = U.S. 2.97 + rest 1.83 + EU 0.32; subscriptions = U.S. 1.91 + rest 1.18 + EU 0.21; other = U.S. 1.18 + rest 0.73 + EU 0.13.)
Structural read: gaming is the largest category exposure, the U.S. the largest regional exposure, China is ironclad (m=0), and the EU only erodes the spread.

6.3 Formulas (Fully Transparent) (Assumed)

Parameter sources: base(c) gaming 15% / subscriptions 10% / other 10% (current actual; subscriptions/other use conservative assumptions); ceil(c,s) three-scenario three-category ceilings (§4.2); regional multipliers above. All (Assumed); rationale in §7.

6.4 The 3 Most Sensitive Assumptions (Anchor: Base 5-Year $13.5B; model.py Single-Factor Sensitivity Output)

1. Whether the U.S. 0%-commission order survives (the runaway leading binary event): If overturned, the U.S. m drops from 1.0 to ~0.66 → net loss −$2.1B (−15.3%). The single largest risk point in the entire model.
2. Gaming bypass ceiling ±10pp: ±$0.9B (±6.9%). Gaming is 53.5% of the commission pool; the 45–65% ceiling range directly determines nearly half the losses.
3. TAC pass-through coefficient 0.7/1.3: ±$0.9B (±6.7%); if agent backends run through Google's pipes (the §5.4 keep branch), TAC losses could approach zero.

(Subscription ceiling ±10pp moves only ±$0.5B (±3.5%) — relatively insensitive.)

6.5 One-Sentence Conclusion

What the model rests on, on one page. Disagree with any item — change the parameter and re-run; the conclusions will speak for themselves.

A summary. No figure on this list entered the model — this is the report's honesty boundary.

Profit & Services Layer

1. Services segment operating profit: Apple's 10-K does not disclose it (reportable segments are disclosed by geography only, and geographic segments don't disclose profit either). The report's $66.0B–$75.1B is an estimated range under two cost-allocation methods (Assumed).
2. Precise internal breakdown of Services (App Store / ads / AppleCare / iCloud / Music / TV+ / Pay revenue and gross profit each): 10-K Item 1 lists business names only, with no breakdown disclosed.
3. Apple Arcade / Fitness+ / News+ revenue each: no reliable third-party estimates (combined at $1B–$2B (Assumed)).
4. Apple Card standalone revenue: partnership structure undecided after Goldman Sachs' exit.
5. Within-bundle split of the Apple One subscription pack (Apple only reports 1.5 billion total subscriptions, not first-party vs. third-party).
6. Precise iCloud revenue: only the single ~$10.4B estimate, no second-source cross-check.
7. Actual Apple TV+ revenue: only subscriber count (~45M) and annual loss (>$1B); revenue is back-calculated.
8. Original named sell-side reports from Bernstein/Morgan Stanley on App Store commissions: public channels only surfaced Appfigures and Visible Alpha consensus-level figures.

Commission-Pool & Rules Layer

9. Precise share of subscriptions in App Store billings: non-gaming subscriptions at 38% is this study's estimate (range 35–42%).
10. Precise EU share of "European" digital goods: Analysis Group only gives Europe $24B; no public EU/UK/Switzerland/Norway split, estimated at 70% (range 65–75%).
11. Actual U.S. external-link migration share: no public precise data on how much billings already flow through the 0% external-link channel, estimated at ~10–15%.
12. Blended effective commission rates by category (no public precise data on big/small-developer revenue mix): gaming 28% / subscriptions 18% / other 27%, all estimated.
13. Coverage and revenue scale of the WeChat mini-program 15% agreement: Bloomberg called it "potentially billions of dollars," no base figure; treated only qualitatively as an offset.
14. Korea alternative-billing 26% rate: actual adoption qualitatively "minimal," no number.
15. Brazil CADE settlement rates (external links 15% / third-party payments 25%): single source, not independently verified.
16. Japan new rate details (IAP 10%/21%, CTC 5%): aggregator sources; no Apple official announcement found.

Bypass-Rate Layer (Core Black Hole)

18. True bypass rates by category (migrated billings ÷ total billings): no public data source; Apple is barred from tracking, vendors don't disclose, third parties can only measure inside the App Store.
19. Specific Spotify iOS external-link conversion figures: only "significant growth" qualitatively, no percentages.
20. Supercell Store's share of Supercell revenue: undisclosed.
21. Patreon web-checkout adoption share: undisclosed.
22. Kindle "Get Book" button click/conversion data: undisclosed.
23. U.S. gaming in-app direct-external-link adoption instances and rates: no large-scale reporting seen; unknown.
24. Precision of Appcharge's $17B/15%: vendor survey basis, median extrapolated from 281 valid responses; directional evidence only.
25. Appfigures' "−18%" primary report: only the FT 2026-08-18 report and MacRumors relay seen; the original report could not be opened to verify the primary figure.

TAC Layer

26. Actual Google TAC payments in 2025/2026: no new disclosures since the 2022 $20B; no analyst has published updated 2025/2026 annual estimates.
27. ISA's specific revenue-share rate and payment formula: confidential.
28. Renewal terms after the September 2026 expiry: no reporting yet.
29. TAC's exact amount within Apple's reported licensing sub-line: Apple does not disclose.

Translating the $13.5B into stock-price language: the base case is roughly equivalent to −5% of today's share price, plus a layer of multiple-compression risk not in the model. A slow-burn derating, not a halving.

9.1 Transmission Chain: Gross-Profit Loss → EPS Impact (2031 Annualized)

9.2 Discounted to Today: Share-Price Impact

9.3 Second Layer: Multiple Compression (Not in the Model, but Potentially Just as Damaging)

The calculation above only moves the numerator (profit). If the market re-rates Apple from "platform collecting channel tax" to "ordinary hardware + services company," the denominator moves too: each 1x of 2031 P/E compression ≈ $13.7/share; 35x→32x adds roughly −$41/share — the same order of magnitude as the profit loss itself. That's the "slow-burn derating": profit a little less each year, multiple killed a little at a time.

9.4 File Index and Reproduction Guide

Reproduce: python3 model.py (no third-party dependencies). Change parameters → re-run → cross-check model_assumptions.csv. Scenario definitions: Conservative/Base/Aggressive × 3-year (2029) / 5-year (2031).

Ordered by first appearance in the text. Every source that can be directly linked is linked; name-only citations (no direct link) are relayed sources.

1. Apple Inc. FY2025 10-K — SEC EDGAR. aapl-20250927.htm
2. Apple Inc. FY2026 Q3 10-Q — SEC EDGAR (accession 0000320193-26-000020). EDGAR filing index
3. Analysis Group, Apple's Global App Store Ecosystem and Its Growth 2025 (commissioned by Apple, 2026-06). MacRumors summary (PDF via Apple Newsroom)
4. TechCrunch 2026-06-04: Apple says 2025 App Store developer billings and sales reached $1.4T, 90% without a commission. Article
5. Appfigures: 2024 global App Store net commissions $27.39B — via FT 2026-08-18, relayed by MacRumors. MacRumors · FT report archive
6. Sensor Tower State of Mobile 2026: global in-app purchase + subscription spend $167B — PR Newswire release. Release
7. TechCrunch 2026-01-21: 2025 non-game app spend exceeded games for the first time (ChatGPT $3.4B). Article
8. Bank of America — via 9to5Mac 2025-07-03: gaming was 45% of the 2025 single-store App Store. 9to5Mac
9. Goldman Sachs: 2025 gaming share 44% — via AppleWorld.today. AppleWorld.today
10. Sensor Tower State of Gaming 2026: 2025 App Store gaming IAP $52.5B — via 9to5Mac 2026-02-25 (name citation, no direct link)
11. Business of Apps / MIDiA: Apple Music annual revenue ~$10.12B. Business of Apps
12. ainvest: AppleCare+ annual revenue ~$8.4B. ainvest
13. Search Engine Land: JPMorgan estimates Apple's ad business reaching $6B by 2025. Article
14. 9to5Mac 2025-03-20 / The Information: Apple TV+ losing >$1B/year, ~45M subscribers. 9to5Mac
15. electroiq: Apple Pay annual revenue ~$4.45B (forecast, methodology attribution unverifiable on the source page). electroiq
16. Computerworld / Trefis: iCloud annual revenue ~$10.4B (single source; CIRP only provides 70% paid-adoption rate). Computerworld
17. GDC × Appcharge survey: mobile-game DTC already a $17B market, ~15% of publisher revenue — PocketGamer.biz. PocketGamer.biz · Press release
18. Supercell Store web top-up +10% bonus — Fandom user blog. Fandom
19. TechCrunch 2025-05-02: Apple approves Spotify iOS v9.0.40 (with external subscription link). Article
20. bbntimes: Patreon iOS v125.5.0 adopts web checkout. bbntimes
21. MacRumors 2025-05-06: Kindle iOS "Get Book" button. MacRumors
22. Netflix reader external links (since 2022) — AppleInsider / beebom. beebom · GSMArena
23. N.D. Cal. 2025-04-30 Epic v. Apple contempt order (U.S. external links at 0% commission, tracking ban) — full text via idownloadblog. idownloadblog
24. 9th Circuit 2025-12-11: contempt finding upheld, "blanket collection ban" vacated and remanded — MacDailyNews (opinion via scribd). MacDailyNews · Opinion
25. Apple's 2026-08-13 (PT) proposed new rates 15%/5%/10% to the court, disclosed 08-18 — iClarified. iClarified
26. U.S. Supreme Court grants certiorari 2026-06-30 (Epic v. Apple contempt case) — ipwatchdog. ipwatchdog · Engadget
27. MacObserver: oral argument at the earliest 2027, not yet scheduled. MacObserver · WebProNews
28. 9to5Mac 2026-08-18: EU's new four-tier flat commissions (effective 2026-10-01). 9to5Mac · MacRumors
29. MacDailyNews 2026-08-19: EU fee simplification. MacDailyNews
30. European Commission's 2025-04 €500M DMA fine against Apple — MacRumors 2025-04-23 (name citation, no direct link)
31. MacRumors 2025-11-13: WeChat mini-programs join Apple's Mini Apps Partner Program at 15%. MacRumors · MacObserver
32. Bloomberg (via PYMNTS): WeChat deal "worth billions of dollars". PYMNTS
33. 3u.com: Korea App Store alternative billing at 26%. 3u.com
34. games.gg: Japan's Mobile Software Competition Act (MSCA) effective 2025-12-18. games.gg
35. Bernstein: Google pays Apple $18–20B/year — via The Register. gfmreview
36. 2022 trial disclosure of $20B; Morgan Stanley ~$20B — Reuters 2024-08-06. Reuters
37. Evercore (Amit Daryanani) $15–20B; J.P. Morgan (Samik Chatterjee) "low-to-mid single-digit earnings impact" — AppleInsider forum compilation. AppleInsider forum
38. indexbox.io: post-remedies Alphabet +8.7%, Apple +3–4.3% after hours; BTCC cites Bloomberg ~$20B/year. indexbox.io
39. AppleInsider 2025-09-02: Google remedies ruling (no exclusives, payments allowed). AppleInsider
40. AppleInsider 2026-02-04: DOJ cross-appeals the remedies ruling. AppleInsider
41. Agarwal & Sen (ISB/CMU, 2026-04 SSRN): AI Overviews cut outbound clicks 39.8%, zero-click +34.5% — ppc.land. ppc.land
42. Ahrefs 2026-02: top-position CTR on AI-Overview keywords down a cumulative 58%. Ahrefs
43. Engadget 2026-01-12: Apple's Siri / Apple Intelligence to run on Google Gemini (Gurman: Apple paying ~$1B/year in reverse). Engadget
44. MacRumors 2026-08-18: Appfigures U.S. commissions −18% (via FT); Sensor Tower −6%. MacRumors
45. Dataconomy 2026-08-21: Morgan Stanley measures U.S. −0.6%. Dataconomy
46. Apple 10-Q risk disclosure: "may not earn a commission at all" — AppleInsider 2026-08-18. AppleInsider
47. China antitrust complaints (no SAMR case filed) — PYMNTS. PYMNTS · ctol.digital
48. TechNode 2017-09-18: Apple's China tipping policy. TechNode
49. Apple FY2026 Q3 earnings call CFO: ~2pp tariff drag on gross margin — Alphastreet transcript, via NPR / Mac Observer (name citation, no direct link)
50. Supreme Court decision expected before 2027-06 (Estimated) — WebProNews. WebProNews
51. Apple FY2025 10-K excerpts — stocktitan. stocktitan
52. MacRumors 2026-08-18: EU App Store fee changes. MacRumors
53. 9th Circuit 2025-12-11: contempt upheld, blanket collection ban vacated, remanded for rate-setting under "true reasonable necessary cost" — ScaleJuris. ScaleJuris
54. 9th Circuit 2026-04-28: stay vacated, case remanded to Judge Gonzalez Rogers for rate-setting hearing — WebProNews. WebProNews
55. Supreme Court grants Apple's appeal of the contempt order (cert granted ~2026-06-30), but declines to stay lower-court proceedings — ipwatchdog · TechCrunch. ipwatchdog · TechCrunch
56. Apple's 2026-08-13 proposed external-link rates 15%/10%/5% (disclosed 08-18); filing concedes the rate should be 0% under the "necessary cost" definition (Epic Newsroom) — TechCrunch. TechCrunch