Meta Platforms, Inc.

Revenue beat in Q2; costs, cash flow and capital return all moved the wrong way. Our P/E basis is FY2026E and FY2027E consensus EPS as of October 7.

Since our initiation, when the stock closed at $656.73 on July 13, Meta has made a round trip and then rallied. It slid 14.0% in the two weeks before the Q2 report, then fell 8.0% on July 30, the day after, to $539.03, as costs rose 55% and EPS missed. It fell back to $543.67 on August 18 as the multistate youth trial opened, then steadied after the August 26 settlement. The leg up came with Muse: the stock rose 6.6% on September 9, the day after the launch, and 11.3% on September 21, a day of broad gains in technology stocks on which Amazon moved to block Muse from its store. It closed at a 52-week high of $777.59 on September 24 ($779.82 intraday) and has given back 7.2% since.

At $721.31 the stock is 13.9% above its 50-day average ($633.43) and 14.6% above its 200-day ($629.54). Since the high, closes have held between $715.62 (September 28) and $751.66 (September 25); a close below that range would be the first lower low of the move. Meta has beaten the Nasdaq-100 ETF since our initiation (9.8% against 6.5%) but still trails it this year (9.3% against 23.3%).

The market has since paid for the AI options we flagged in July; the debate has moved to the bill. Revenue rose 33% in Q1 and 28% in Q2, and Q3 is guided to $61B–$64B, growth of 19% to 25%. Costs grew faster: 24% in 2025, 35% in Q1 and 55% in Q2 (42% before the legal and severance charges), and the $167B midpoint of the 2026 guide is 42% above 2025.

The consensus does not publish a cost line, but its EPS implies one. Holding revenue at consensus and using Meta's guided tax rate:

We think 2027 costs grow faster. The 2026 guide implies $91.5B of costs in the second half, $183B a year at that pace. This year costs are on track to finish 27% above the annualized pace of the second half of 2025; the consensus has 2027 finishing only 10% above the second half of 2026, about $18B of room. Depreciation rose 46% to $6.36B in Q2 after $50.9B of first-half capex, and the guide implies $79B–$94B more in the second half; servers bought this year are depreciated next year. And management keeps leaning in: Meta has "continued to underestimate our compute needs" (Q1 call) and has geared its plans to "maximizing 2026 and 2027 capacity" (Q2 call). Our base case uses 22%, two points above the consensus math; our bear case uses 26%.

The spending behind those costs is on paper. At June 30 Meta had $279B of leases not yet started, mostly data centers, and it signed $68B more in July. Separately it had $349B of non-cancelable purchase commitments, with $53.52B due in 2026 and $81.65B in 2027. Together that is about $696B, 38% of the market value, against $84B of debt. None of it threatens solvency; it shows how little of next year's cost base is still a choice.

The other change is per share. Meta repurchased $22.9B of stock in the first half of 2025 and none in the first half of 2026, leaving $25.03B of authorization unused, while stock comp rose 58% to $7.66B in Q2 and unvested awards rose to 146.5 million from 115.6 million. The share count rose by about 18 million in six months. The 10-Q lists equity among possible sources of capital, and the stock fell 5.5% on June 5 on a press report that Meta was weighing a large share sale; none has been announced. Our model assumes no buyback and a 1% higher share count in 2027.

Family of Apps is still the whole economic story. Its Q2 revenue was $60.37B, up 28%, and its operating income was $23.39B, down from $24.97B as costs, including the legal and severance charges, rose faster than revenue. Ad impressions rose 14% and the average price per ad 12%; daily active people reached 3.60 billion, up 3%. Other revenue passed $1B for the first time, up 73% on WhatsApp paid messaging and subscriptions, and Advantage+, the automated campaign suite, runs at more than $75B a year.

Reality Labs revenue rose 16% to $431M on AI glasses, and the segment lost $4.62B in Q2 and $8.65B in the first half. At its Connect conference in September Meta set a spring 2027 launch for Meta VR Glasses at $1,299.99.

What changed after July is the product list. Muse, launched September 8, sends email, books travel, fills in forms and buys things for its user; it is free with a usage limit, with paid tiers at $20 and $100 a month. Shopify, PayPal and Instacart signed on as partners, and Amazon blocked it from its store on September 21. Meta One, a subscription across the apps from $2.99 a month, had 15 million subscriptions and trials by its wide launch on September 15. More than 1 million businesses use Meta's business agents each week, and on September 28 Meta formed Meta Enterprise Platform under CJ Desai, MongoDB's former chief executive. Zuckerberg says Meta gets offers for its compute "at a significant premium over what we paid for it."

The new lines are real but small next to a $300B revenue base. Wedbush puts Muse subscriptions at about $5B in 2027, roughly 1.5% of revenue, and Oppenheimer calculates that even 500,000 U.S. downloads a day for more than a year would lift 2027 revenue only about 4%. Next year's earnings still ride on advertising and costs.

Our P/E basis is FY2027E. Trailing GAAP EPS is distorted in both directions: Q1 2026 included an $8.03B tax benefit worth $3.13 a share, and Q3 will carry the settlement charge. We do not use a DCF: with free cash flow near zero at the peak of the build, the answer would rest almost entirely on terminal assumptions. We hold revenue at consensus and change only the cost line, because that is where we disagree.

Our base case pays 23x our FY2027E EPS of $33.03, which gives $760: about 22.3x consensus FY2027E EPS and close to Alphabet's 23.1x. We do not add the $6.6B of net cash, because the $12.7B of settlement payments due over ten years more than uses it up. With the 10-year Treasury yield at 5.27%, the FY2027E consensus earnings yield of 4.7% already sits below the risk-free rate, so we do not underwrite a higher multiple. Each two points of cost growth moves our value by about $25 a share at 23x, about three-quarters of a turn of multiple.

Probability-weighted value: $756 (+5% vs the reference price).

Mark Zuckerberg controls the vote through Class B shares, and the build is his bet. In the first half Meta granted 20 million stock options with a weighted-average exercise price of $2,788, 3.9x the last close, which pay off only if the stock compounds for years. The May reduction affected about 8,000 employees and cost $1.18B in severance; headcount was 75,472 at June 30.

Meta is leaning on its balance sheet. It sold $25B of notes on May 4 in six series due 2031 to 2066 at coupons of 4.55% to 6.45%, taking long-term debt to $83.66B from $58.74B. Beyond its $90.26B of cash and securities, another $10.80B of money-market funds sits in escrow for multi-year infrastructure purchases until 2028–2030. Meta also builds through partners: it owns 20% of a Louisiana data-center venture (about $27B of costs; maximum exposure $46.03B) and agreed in July to a similar El Paso venture 80% owned by BlackRock funds (about $14B). Capital return has narrowed to the dividend, $0.525 a quarter (0.3% yield).

Insiders sold about $108M of stock over the past six months and bought none, almost all under Rule 10b5-1 plans. Chief Product Officer Chris Cox sold 80,000 shares for $55.0M between September 9 and 21 at an average $687.75, under a plan adopted May 19. Zuckerberg's $21.4M sale on September 24, the day of the closing high, was made by his philanthropies under a plan adopted January 31. We read the selling as routine. The table lists the six largest sellers at Form 4 transaction prices.

The legal tail is smaller and closer. On August 26 Meta agreed to pay about $18B over ten years to settle with 52 attorneys general, $12.7B of it to the states and $5.3B only if YouTube and TikTok adopt daily time limits, night mode and age checks and pay matching amounts; Meta expects a legal expense of about $10B in Q3. The headline risk had been far larger: across the youth cases, plaintiffs had said they could seek more than a trillion dollars.

What remains is concentrated. A New Mexico jury found on September 25 that Meta misled residents over the Cambridge Analytica affair, counting more than 43 million violations; the state asks $35B–$40B, Meta proposes a $3.45B cap, and the judge expects to rule this month. The top of the request is about $15.59 a share, 2.2% of the market value. New Mexico's youth case has cost $942M so far ($375M civil penalty, $567M abatement fund), which Meta says it will appeal, and two more personal-injury bellwethers start October 28.

Options price a move of about ±7.7% ($55.79) through the October 30 expiry. Implied volatility steps up from 34.8% for the October 23 expiry to 45.5% for October 30, the week the market expects results; at-the-money volatility of 44.1% has an IV rank of 62, elevated but not extreme.

Positioning leans long: call open interest of 2.22 million contracts is about twice put open interest of 1.03 million (put/call 0.47). Calls cluster at $750 (310K), $800 (205K) and $700 (184K); the largest put strikes sit at $500 and $600. Dealer gamma is net positive and concentrated at $750 and $700, which tends to damp moves between them; around $720 it turns slightly negative, so a break lower could run faster. Short interest is small: 31.0 million shares, 1.4% of Class A stock and 1.57 days of trading at the September 15 settlement, down from 37.8 million in mid-July.

The Street has chased the stock. The consensus mean target is $781 (median $775, range $595–$1,000) as of October 7, with 52 Buy, 11 Hold and 2 Sell ratings, and the six targets published in the past month average $887.67. Most raises came after the Muse launch and Connect, with the stock at or near its high: Wells Fargo went to $1,000 from $796 on October 6, two weeks after raising its target to $796 from $640. FY2027E consensus EPS has slipped to $34.01 from about $35 when we initiated, so targets have risen on the multiple, not on earnings.

Our $760 target is 2.7% below the consensus mean.

What would move us to BUY. A 2027 outlook that points to cost growth near 20% or below, which would lift our value to about $785; a pullback toward $650, about 20x our FY2027E EPS, with the outlook unchanged; or a return to buybacks.

What would move us to SELL. A 2027 outlook with cost growth of 26% or more while revenue growth slows below 20%; a share sale; or a New Mexico penalty near the state's request that survives post-trial motions. Two of the three would take our value toward the $549 bear case.

HOLD, 12-month target $760 (+5% versus the $721.31 close on October 7), down from BUY and $805. Meta is executing, and the largest legal tail is settled at a cost it can carry. But the stock has moved from 15.9x to 21.2x FY2027E consensus EPS since July 30, and that consensus assumes costs grow about 20% next year, half this year's pace, with much of the spending already contracted. The October report will show which side is right. For a momentum investor the trend is intact: hold through the print, and buy a pullback toward $650 or a 2027 cost outlook near the consensus math.