To The Moon

Last Thursday the Philadelphia Semiconductor Index fell 12.2% in one session, its worst day in more than six years, the S&P 500 technology sector dropped 7.3% and the Nasdaq lost 5.5% as investors marked down anything tied to datacenter construction (Cailianshe weekly wrap, 8 Nov 2026). By Friday's close the week's best large US stock was not a chipmaker:$ $AppLovin Corporation(APP)$ finished at $390.02, up 64.59%, while $SPDR S&P 500 ETF Trust(SPY)$ lost 3.07% and $Invesco QQQ(QQQ)$ fell 5.65% (ABCMoney, consolidated exchange data, week ending 8 Nov 2026).

AppLovin did not rally because the AI narrative improved. It rallied because its 4 November quarter made the capex argument irrelevant to its own income statement: revenue of $2,185M, up 55% year on year; GAAP net income of $1,439M, up 72%; an 84% adjusted EBITDA margin; and a $3.2B addition to the buyback (AppLovin Q3 2026 release, SEC 8-K, 4 Nov 2026). The question is not whether AppLovin is an AI stock. It is whether a business priced on ad budgets can remain the market's favourite AI proxy while the datacenter complex is repriced.

Figure 1 — Q3 FY26 headline numbers: revenue, profit, margin and buyback capacity all set records in the week chips had their worst session since 2020. Sources: AppLovin Q3 2026 press release, SEC 8-K, 4 Nov 2026.

The week AI split into two trades

The trigger was political, not industrial. After the Democratic sweep of both chambers, the market priced tighter scrutiny of datacenter power use and permitting, and the selling landed on the companies that supply that pipeline. Cailianshe put Thursday's decline at 7.3% for the S&P 500 technology sector and 12.2% for the Philadelphia Semiconductor Index, both the worst single sessions in more than six years.

Valuation made it violent. Truist's Keith Lerner notes the sector now trades near 17x forward earnings, below the broad index's 18x for the first time since 2017, against 32x at the October 2025 peak (Cailianshe, citing Truist). A sector priced for perfection was repriced in four sessions.

Friday brought a partial bounce, with $Advanced Micro Devices(AMD)$ up 11.04% and $ASML Holding NV(ASML)$ up 5.02% (BNN Bloomberg, 8 Nov 2026), but storage did not recover. For the week $TICKER(STX)$ fell 30.17% to $462.99, $TICKER(WDC)$ 29.96% to $264.18, $TICKER(SNDK)$ 29.24% to $835.65, $TICKER(MRVL)$ 29.68% to $141.52 while absorbing its $115bn all-stock deal for Lumentum, and $TICKER(LITE)$ 37.65% to $540.32 (ABCMoney).

Two discount rates now apply to one theme. Capex earnings are a derivative of a permitting decision nobody can forecast; ad earnings are a derivative of app-install budgets, set annually and indifferent to zoning hearings.

Figure 2 — Five-session performance across the AI complex. One ad platform tops the board; the storage and interconnect names that feed datacenters sit at the bottom. Sources: ABCMoney consolidated US exchange data, week ending 8 Nov 2026.

The quarter that made the capex debate irrelevant

Revenue rose to $2,185M from $1,405M; net income to $1,439M from $836M; adjusted EBITDA to $1,842M from $1,158M, a 59% increase. Over nine months, revenue reached $5,951M against $3,823M and net income $3,911M against $2,231M (AppLovin Q3 2026 release, 4 Nov 2026).

The margin line carries the argument. Net income margin was 66%, adjusted EBITDA margin 84%. Free cash flow was $1.4B; the company repurchased and withheld 4.1 million Class A shares for $1.4B, funded by the quarter's own cash rather than debt. Diluted EPS was $4.30.

The operating story, per sell-side research summarised by [CFI.cn](http://CFI.cn) on 8 November, is that AppLovin's bidding model lifted eCPM more than 20% while retention improved for a fifth straight quarter. That is an ecosystem claim, not an audited fact, but it fits a margin that is audited.

Figure 3 — Revenue, adjusted EBITDA and GAAP net income, Q3 FY25 versus Q3 FY26. All three grew 55-72%, so the margin came from operating leverage, not financial engineering. Sources: AppLovin Q3 2026 press release, 4 Nov 2026.

What the guidance actually promises

AppLovin guided fourth-quarter revenue to $2,365-2,395M and adjusted EBITDA to $1,995-2,025M at an 84-85% margin (same release). At midpoints that is $2,380M of revenue, roughly 9% above the September quarter, and $2,010M of EBITDA.

A 9% sequential guide at an unchanged margin says the ad market is not expected to crack within eight weeks, and each extra dollar at 84% margin is nearly pure profit — which is why 55% revenue growth produced 72% income growth.

One gap deserves flagging: the release provides no forward-looking GAAP equivalents for adjusted EBITDA, citing uncertainty around items such as stock-based compensation. Non-GAAP guidance to a 1% revenue range is precise; the missing GAAP bridge is not.

Figure 4 — Q3 FY26 actuals versus the midpoint of Q4 guidance. The implied sequential step is about 9%, with margin held at 84-85%. Sources: AppLovin Q3 2026 press release, 4 Nov 2026.

Bull Case

1. The buyback is funded by operations. $1.4B of quarterly free cash flow against a $3.6B authorisation means shares can be retired without borrowing — at this price, roughly 9% of the share count, arithmetic on disclosed figures rather than a forecast.

2. The margin is a structural toll, not a cycle peak. An 84% margin survived 55% revenue growth. That cost structure needs a normal ad market, not a strong one.

3. Guidance implies no deceleration. A 9% sequential guide, issued four days before an election result that then cratered the AI complex, is this month's closest thing to an all-clear.

4. The demand driver is software budgets, not capex. App-install spending sits in annual marketing plans, far less sensitive to permitting politics than a hyperscaler's build schedule.

5. The relative trade is now visible. A 64.59% week against a 3.07% index decline is the dispersion that forces generalist funds to look.

Bear Case

Figure 5 — Week close against the 52-week range. Even after a 64.59% week the stock sits roughly 47% below its high: a recovery inside a drawdown, not a breakout. Sources: ABCMoney week-ending data; MEXC 52-week range; percentage calculated from those levels.

1. The week was multiple expansion, not news. The quarter was published on 4 November and guidance has not changed since; only the price moved. MEXC puts the trailing P/E at 25.86 and market capitalisation at $129.62B, with the stock 47% below its 52-week high.

2. Ad budgets are cyclical. A December rate rise is the market's base case ahead of the 10 November CPI print (Cailianshe). The first cut at any app developer is user acquisition; an 84% margin protects profit, not revenue.

3. Concentration is under-discussed. MEXC's profile attributes about 80% of revenue to the AppDiscovery demand-side platform. One algorithm and one set of platform policies carry the thesis.

4. Buybacks get less efficient as the stock rises. The same $3.6B buys roughly 9.2 million shares at the week's close against more than 15 million at the 52-week low, and there is still no forward GAAP bridge: if stock compensation rises with the share price, the gap between the 84% headline and GAAP reality widens precisely when the stock is dearest.

Strategic Slam

• $TICKER(APP)$ — an advertising compounder classified as AI, not an infrastructure name. The thesis holds while fourth-quarter revenue stays at or above the $2,365M low end and margin at or above 84%. If either breaks, the multiple is the risk, not the business.

• $TICKER(SPY)$ / $TICKER(QQQ)$ — the index still carries the capex complex, so a broad-market position is a position in the permitting debate. That is a reason not to treat it as a hedge against AppLovin.

• $TICKER(STX)$ / $TICKER(WDC)$ / $TICKER(SNDK)$ — a separate trade on a separate clock, priced on contract pricing and inventory. Wait for pricing data before reading a 30% week as a valuation event. No entry, stop or target levels are given here; those need a live quote service and a stated horizon.

Catalyst Calendar

Figure 6 — Dated events that confirm or break the thesis. The fourth-quarter report date is expected, not announced. Sources: AppLovin investor relations and SEC 8-K, 4 Nov 2026; ABCMoney week-ending data.

The next macro input is US October CPI on Tuesday, 10 November, the print Cailianshe reports the market is using to price a roughly 80% probability of a December rate rise. For an ad-funded compounder the channel is indirect but real: higher rates compress long-duration multiples before they touch budgets. After that, the February report is the only event that can validate the 9% guide.

Two scenarios, one decision

Figure 7 — The 2x2 that matters: whether ad budgets hold, and whether datacenter fear fades. Neither favourable quadrant requires the capex debate to break AppLovin's way. Sources: AppLovin Q3 2026 release; ABCMoney.

Verdict

AppLovin is the cleanest way to own AI-adjacent earnings without owning permitting risk, and last week proved the market agrees — a toll booth on advertising traffic rather than a semiconductor. At 84% margins it prints cash faster than it can deploy it, hence the $3.6B buyback.

The stance: constructive on the business, neutral on the week's move. The 64.59% gain was a rotation into the one AI name priced on profit and loss, and such rotations overshoot before they consolidate. The test is falsifiable: revenue at or above $2,365M and margin at or above 84% in February, or the thesis needs rewriting.

Key Takeaways

1. AppLovin rose 64.59% in a week when the Philadelphia Semiconductor Index fell 12.2% in one session, a rotation rather than a sector rally.

2. Q3 revenue of $2,185M (+55%) and net income of $1,439M (+72%) came with an 84% margin and $1.4B of free cash flow; Q4 guidance implies roughly 9% sequential growth at the same margin.

3. The bear case is valuation and cyclicality, not solvency: the stock is ~47% below its 52-week high and about 80% of revenue sits in one platform.

4. February is the arbiter: revenue at or above $2,365M with margin at or above 84% keeps the re-rating intact.

The question back to you

If the market has started paying for AI profits rather than AI capex, which other name earns its multiple from a budget line that needs no permit — and would you rather own that, or the index still carrying the datacenter?

Tickers: $TICKER(APP)$ $TICKER(SPY)$ $TICKER(QQQ)$ $TICKER(AMD)$ $TICKER(ASML)$ $TICKER(STX)$ $TICKER(WDC)$ $TICKER(SNDK)$ $TICKER(MRVL)$ $TICKER(LITE)$

Sources: AppLovin Q3 2026 release and Form 8-K, 4 Nov 2026 (all financials, buyback, Q4 guidance); ABCMoney "Sunday Money", consolidated exchange data, 8 Nov 2026 (weekly moves and closes); Cailianshe weekly wrap and preview, 8 Nov 2026 (sector declines, Truist forward P/E, CPI timing, rate odds); BNN Bloomberg, 8 Nov 2026 (AMD, ASML); MEXC APP quote page, 6 Nov 2026 close (52-week range, P/E, market cap, revenue mix); [CFI.cn](http://CFI.cn) digest, 8 Nov 2026 (eCPM, retention).

Disclaimer: Research commentary, not investment advice, with no price targets or entry and exit levels. Figures are as reported above and may be revised. Positions can lose value.