Skip to content
Some content is members-only. Sign in to access.

Meta Bull Case Holds on Ads, Bear Case on Spending

26% revenue growth supports valuation as capex and leverage reshape risk

By KAPUALabs

The headline numbers for Meta's third quarter are a measurement failure. GAAP earnings and the underlying business point in opposite directions. Advertising revenue grew 26% to $63.06 billion 76,124, yet GAAP diluted EPS was $3.83 88,96. A $9.89 billion legal charge explains most of that gap 47,56,79,134. The charge is the noisy number. The quieter and more consequential one is free cash flow, which turned negative. Meta is now paying for its AI and data-center build with outside capital. The question is not whether the ad engine works, but how you know the spending it funds will earn its keep.

What the ad engine actually delivered

Revenue rose 26% to about $64.6 billion 52,92,117,121. This is the most heavily corroborated fact in the set. It was above guidance 121, which was $61–64 billion 53,79, and above a $63.34 billion Street expectation 60. Advertising was about 98% of the total 79, and Meta has now reported year-over-year revenue growth for 15 consecutive quarters 57.

The quality of the growth is what a skeptic wants to see. Both price and volume contributed. Average price per ad rose 14% 52,117 and impressions rose 10% 63,66,71,76,95. Daily active people reached 3.64 billion 76,88. One source has the average ad price up 12% 1,2,5,8,9,18,21,24,26,29,30,32,33,34,36,38,41,63,65,76,77,95,119, but it reaches back to 2025, so the 14% figure is the better match for this quarter. A department store that raises prices and sells more units has a healthier floor than one that does only the first. Price and volume together are the stronger result.

Several social posts cite 28% growth 49,128. Other items tie that number to Q2 ad revenue 103. Those claims are single-source and unverified as to period 49,102,127. I weight the 26% Q3 figure above them. The trend is a deceleration from 33% in Q1 16,53. Q4 guidance of 24–29% growth implies a continued but slower pace 100.

Relative position matters too. One source says Meta's growth gap over Google Search widened to a five-year high 98. Meta's $63.1 billion of ad revenue sat almost level with Google Search's $63.4 billion 85. Snap's ad revenue rose 4% 77 and Pinterest grew 12% to $1.17 billion 87.

One caution is in order. All of this is reported revenue. It tells us what advertisers paid. It does not tell us what they got back. Nothing in this material measures advertiser incrementality, so the durability of the 14% price increase is an inference, not a finding.

The charge relates to an August 2026 agreement with US state attorneys general over teen mental-health and social-media litigation 76,79. Reported operating income was $12.99 billion 47,56,79,134. Excluding the charge, it would have been $22.88 billion 94. Reported operating margin was 20.1% against 40.1% a year earlier 60,79,131. Excluding the charge it was about 35.4% 131. Management said operating income would have risen 11% without the charge 65 and attributed expense growth mainly to legal costs 65.

On EPS, diluted earnings were $3.83 against $1.05 a year earlier 94,131. Excluding the charge they would have been $7.40 61,96,97. Whether that is a beat or a miss depends on the benchmark:

The disagreement is about whether the charge is included. It is not a factual dispute about the quarter. Careful readers should state which basis they are using.

The "one-off" label needs care. Settlement payments of up to $18.6 billion are cited 117, and management warned that legal and regulatory matters could significantly affect results 65. A potential EU fine of 6% of annual turnover is also noted 6,7,19,68. An Italian fine of €805 million is about 1.1% of cash and securities of $82.45 billion 88,89. One isolated source cites a "$25 billion" fine 90. It is inconsistent with the rest of the evidence and I do not carry it forward.

The ex-charge margin is the number to watch, and it is lower. It fell from roughly 40% to about 35% 79, which management described as a 500 basis point contraction 52. The cost lines explain why. Total costs were up 68% 88, and depreciation grew 44% 123. Q3 server depreciation was cited at $5.34 billion 116. Headcount fell 15% to 66,775 79. Meta is cutting people costs while infrastructure costs climb. Reality Labs adds to the drag. It lost $4.13 billion on $443 million of revenue in Q3 62,79, and the 9M2026 operating result was –$12.8 billion 71.

Cash flow: where the measurement really breaks

Earnings can be adjusted. Cash is harder to adjust. Operating cash flow was $32.79 billion 60,73,94,133,134. Capital investment including finance-lease principal was $34.91 billion 94,134, up 80% year over year 131. Free cash flow was therefore about –$2.1 billion 65,76,88,91,111,121,129, or –$2.12 billion on the more precise figure 60,79,94,131,134. A year earlier the figure was +$10.63 billion 118. Capital expenditure ran at 106% of operating cash flow 115.

The figure is non-GAAP 125, and sources disagree on its definition and history. Other readings of –$1.16 billion or about –$1.2 billion 64,82 and $784 million 92 appear. The –$2.1 billion figure is stated after finance-lease principal 126. The gap is therefore probably definitional, although the evidence does not settle it. On history, one source calls this the first negative quarter in 14 years 111 and another says since 2012 113. A third says only that it is the first in the ten years shown on its page 110. The direction is consistent across all of them. The precise record is not.

Meta is not alone in this. Microsoft generated $58.0 billion of operating cash flow against $39.4 billion of capex 107, and Alphabet's quarterly free cash flow was reported as –$9.07 billion 79. Even so, the credit side is paying attention. S&P revised its outlook to negative 48 and forecasts a $39 billion free operating cash flow deficit in 2026 86. It would need free operating cash flow to turn positive in 2028 to return to a stable outlook 108. A cited risk case reaches –$71 billion in 2027 114. That is a scenario, not guidance.

Funding the gap with outside capital

The financing response is large and spans every instrument. Meta reportedly plans to raise about $95–97.5 billion through stock sales, and the sources differ on the exact figure 75,86. The pieces already visible fall into four groups.

Debt. A $15 billion five-part dollar offering 106 included $3.25 billion of 5.850% notes due 2031 51, with maturities running from 2031 to 2056 83. It followed a $25 billion sale in April 106. A €9 billion euro deal came next 50,84. It included €1.75 billion at 4.30% due 2031 130, with pricing of up to 6.4% 67. Moody's assigned Aa3 and expects a positive net cash position to be maintained 106.

Preferred and equity. Meta issued a $12.5 billion mandatory convertible preferred with a 6.5% dividend 60,94. A $60 billion at-the-market program is maintained 92. New Class A shares were priced at $625 65, and forward-sale coverage now totals 46,000,000 shares 51.

Buybacks. Trailing-twelve-month repurchases were zero, against about $26.2 billion in FY2025 74. Trailing financing cash flow was an inflow of about $27.8 billion, versus outflows in every year from FY2022 to FY2025 74.

Share classes. A one-for-one Class C stock dividend, effective January 19, 2027, roughly doubles the share count and halves the price 78,79,121. Several sources describe it as similar to a two-for-one split 91,92. Meta states that voting power does not change 79.

This is not a solvency story. Cash and securities were $82.45 billion 88 and total debt was $83.67 billion 106. Net debt to EBITDA is cited at 1.0x 120, while Pershing Square describes no net leverage 112. These readings differ because they use different measures, and I cannot reconcile them from this material. The debt-to-equity ratio is 0.24 12,13,25,27,28,37,42,80,93,99. The strain is in the direction of cash flow and the rising cost of funding. An investment narrative says both the bond deals and equity issuance lift Meta's funding costs 70. Cost of capital is rising while the stock is sold at around $625 65. That is the economics to scrutinise.

One interpretation, flagged as such. The share-class change adds flexibility without altering control. Consenting shareholders held about 56.1% of voting power 78. The sources do not state intent, so this is my inference.

Shareholder returns have shrunk in the meantime. The quarterly dividend is $0.525 1,22,25,27,31,41,53,56,80,93,96,99, a yield of about 0.3% 3,4,10,11,12,14,22,40,43,44,45,53,59,69,80,88,93,104 and a payout ratio of roughly 8% 10,15,80. With buybacks stopped 122,126, capital return is small and has slipped.

Guidance and the market's verdict

Q4 revenue guidance of $73–76 billion 46,58,79,88 compares with consensus of about $73.6 billion 88. Some authors describe it as above the Street 109,117, while one data summary says it missed midpoint estimates 92. These are single-source assertions with different benchmarks. I read the guidance as roughly in line to modestly ahead.

The cost guidance is the bigger signal. Full-year 2026 expense guidance rose to $175–179 billion including the legal charge, from $165–169 billion 79,88. Reality Labs losses are expected at $18–19 billion 79,88. Management still expects 2026 operating income excluding the legal charge to exceed 2025 79. It gave no 2027 revenue range 79, but it expects to sustain strong growth through 2027 79. It said AI businesses can begin contributing incremental revenue in 2027 72 and will add detail on 2027 capacity on the Q4 call 58.

The stock sold off. The reported magnitudes conflict. One source has a 7.0% close-to-close fall on the reaction day 54, another has 7% the following day 121, and a third says 21% 100. The 21% figure is single-source and out of line with the price evidence. The price evidence points to an 8.0% decline from the October 21 level of $697.52 88, with the stock around $623.61 on November 12 101 and down 2.8% year to date 88. This is the second straight post-earnings decline, after a 10.4% intraday drop on the July report 39,54,55. Short interest was 3.41% of float 81, and the October 30 total of 75,483,549 shares was up 165.6% from October 15 80.

One supplier note. Meta is described as a very large Nvidia customer 132. The material shows a broader selloff in Meta and Nvidia weighing on a Big Tech ETF on November 13 80, and the board inquiry at Nvidia was still open as of November 14 132. The evidence does not show any change in Meta's own demand for chips. Any read-across is indirect and is my interpretation.

Valuation: one tape, two stories

The bull case is simple. Trailing P/E is 21.9 80,93,99,135, forward P/E is about 18 105 and the PEG ratio is 0.83 10,17,20,27,35,40,80,93,99, against Meta's own recent median P/E of 25.4 135. One tally counts 54 Buy, 6 Hold and 1 Sell among 61 analysts 57. Average targets are $850.10 57 or $833.12 80, and Morningstar's $850 fair value implies a 26x 2026 adjusted P/E 23,52.

The dissent is also in the evidence. A reference intrinsic value is $409.55 71. One fair-value model puts its industry-comparison component at $466.47 against $978.77 for its cash-flow component 135. Overall it calls the shares fairly valued, about 11% below its fair value 135. A reverse DCF requires EPS to compound at 17.8% a year for five years 120. The spread between $466 and $979 is the argument. It turns on how much of the cash drain is a temporary investment cost and how much is a permanent change in capital intensity.

What this implies

The first conclusion is firm. The advertising business is delivering growth on both price and volume at very large scale, and the legal charge distorts reported earnings more than it changes the underlying margin. The second conclusion is also firm. That margin is lower than last year, and the cash required to support the AI build now exceeds what operations produce.

The third conclusion is an interpretation. The investment case has moved from earnings power to capital intensity. The 2027 payoff is the swing factor. Analyst projections of $27–29 billion from the Muse product by 2030 92 are projections, not company guidance. Management has said AI revenue can begin in 2027 72 and promised capacity detail on the Q4 call 58. Sources place that report in late January, with 27 January 135 and 26 January 72,88 both cited, and the company has not confirmed it 54.

Where this leaves a reader:

The history of advertising is a history of unmeasured waste. Meta's ad revenue is a measured number. The returns on the capital it is now raising are not. When the Q4 call arrives, what evidence will management offer that the spending converts to revenue?

More from KAPUALabs

See all
| Free

Meta’s AI Build-Out: Ad Gains Versus Funding Strain

By KAPUALabs
/
| Free

Italy’s Meta Ruling: Consent, Remedy and Regulatory Reach

By KAPUALabs
/
| Free

Meta’s Ad Growth Faces an Unproven AI and Subscription Upside

By KAPUALabs
/