Article published Sep 15, 2026. Prices below use latest available snapshots.
META Price Structure
Loading timeframes…
Verdict: the re-rate is real, the re-earn is unproven, and the keynote is tomorrow. Medium conviction. A personal AI agent called Muse turned Meta from the weakest one-year performer in its peer group into a name trading within a few percent of its old high, in about four weeks. Not one line of the financial statements changed while that happened. The setup is now a gap to be tested, not a breakout to be chased.
Quick Snapshot (as of 2026-09-22)
| The call | 🟡 Wait for the gap to be tested, not the keynote. Entry $670–$690 · stop $665 · first target $765 · medium conviction. |
| What it is | The advertising business behind Facebook, Instagram, WhatsApp and Messenger — 3.58 billion daily active people — plus Reality Labs, a hardware and research arm that loses about $19 billion a year on purpose. |
| Why now | Muse, a personal AI agent that performs tasks rather than answering questions, launched September 8; the company reports a 60% jump in daily assistant users since it shipped. On September 21 Wells Fargo raised its target to $796 from $640 and the shares gapped 11.3% on 2.6 times normal volume. Meta Connect's keynote is September 23. |
| Proof it is working | Revenue growth of 26.2%, 23.8%, 33.1% and 28.0% over the last four reported quarters, with ad impressions up 12% and price per ad up 9% in 2025. Muse is the first engagement readout on the AI build-out that anyone outside the company can see. |
| The objection | The re-rate moved the multiple, not the estimates. Wells Fargo left its revenue forecast alone and raised the multiple on 2027 earnings from 20 times to 25 times. Operating margin is still 30.9% against 43.0% a year ago, and 2026 capital-expenditure guidance of $115–135 billion still rivals a full year of operating cash flow. |
| Stock setup | Up 34.0% in a month, 3.7% below the 52-week closing high it set exactly one year ago, and arriving at RSI 76.0. The September 21 gap is unfilled and sits on a volume shelf. The 50-day is still below the 200-day: the move outran its own moving averages. |
Action Matrix (as of 2026-09-22)
| Signal | If | Then |
|---|---|---|
| 🟢 Buy | A pullback holds $670–$690 on a daily close — the September 21 gap, the volume shelf and the value-area high all sit inside it | Entry zone. A gap that holds on a retest is what turns a sentiment re-rate into a trend-hold base. |
| 🟡 Wait | Here at $736.60, into the September 23 keynote | Up 34% in a month at RSI 76, on a multiple rather than a number, the day before a binary event. Nothing about this is an entry. |
| 🔴 Out | A daily close below $665 | The gap fills, the shelf is lost, and the Muse re-rate has been sold. The setup is void whatever the thesis says. |
| 🟠 Target | A daily close above $765 — the 52-week closing high from 2025-09-22 | Trim or trail. A full year of underperformance has been round-tripped; the prior $780 objective and the $785.73 intraday high sit just above. |
| 🔴 Thesis breaks | Operating margin under 30% for a second consecutive quarter at the Q3 print, or Connect and that print passing without a restated Muse engagement figure while capex guidance holds | The spend is then outrunning both the top line and the single engagement number the new multiple was built on. |
The Story Right Now (as of 2026-09-22)
Meta re-rated 34% in a month on a product launch and a higher multiple, and not one line of its financial statements moved while it happened. A week ago this was the weakest one-year performer in its mega-cap peer group, down 12.4% over twelve months and on its third attempt at reclaiming its own moving averages. It is now 3.7% below the 52-week closing high it set exactly one year ago today.
Data as of 2026-09-22 close: META $736.60, RSI 76.0, −0.6% on the day, +9.9% on the week, +34.0% on the month, −3.7% over one year, 6.3% below its 52-week high, and +17.8% against its 200-day.
The cause is a product. Meta launched Muse on September 8 — a personal AI agent that performs tasks rather than answering questions, able to send email, book travel, fill in forms and keep working after the user closes the app, running on Muse Spark, the company's most capable agentic model. The company reports daily assistant users up 60% since Muse Spark was integrated. On September 21 Wells Fargo's Ken Gawrelski raised his price target to $796 from $640, and the shares gapped 11.3% on 2.6 times normal volume. On September 22 they held the gap, ranging to $757.27 before closing at $736.60.
The discipline is in what that analyst note did not do. It left revenue estimates unchanged and raised the multiple applied to 2027 earnings from 20 times to 25 times. The entire move is a re-rating of the same forecast. That is a real event — price is the only currency that settles — but it belongs to a different evidence class than a revenue line, and a multiple granted on a narrative can be withdrawn on one.
It matters anyway, and for a specific reason. The September 15 view named the weakest link in this case precisely: nobody outside the company could price the return on $115–135 billion of annual capital spending, so the operating-margin line was the only readout available, and it was deteriorating on schedule. Muse is the first readout of a different kind. A 60% jump in daily assistant users is not revenue and does not pay for a data centre. It is the first outside-visible evidence that the spending produces something people actually use, which is the question the margin line cannot answer either way.
The setup is the awkward part. The name arrives at RSI 76.0, 17.8% above its 200-day, and still carrying a death cross — the 50-day sits at $610.42 against a 200-day of $625.11, because a 34% month outran both. The September 21 gap, from a $665.75 close to a $679.60 low, is unfilled, and the volume shelf at $669.59 and the value-area high at $675.12 sit inside it. That confluence is the level that decides whether this is a base or a spike, and it is roughly 9% below the current price.
Watch Wednesday: Meta Connect's keynote, September 23 at 4:00pm Pacific, with developer sessions the following morning. The stock has already paid for good news. The number that would convert the re-rate into something durable is a restated Muse engagement figure — users, retention, or any monetisation path at all. A keynote that ships hardware and leaves Muse as an anecdote puts the whole move back on the September 21 gap.
Extended analysis
Why This Moment Is Different
For two years the bear case on Meta was Reality Labs — a headset business that consumed roughly $19.19 billion of 2025 operating profit and, per the annual report, will lose about the same in 2026. That drag is now the stable part of the cost structure. The new one is larger, faster-growing, and sits inside the profitable segment.
Research and development spending rose 67.3% year over year in the June quarter, to $21.66 billion, against revenue growth of 28.0%. Operating income fell 8.2% over the same span. A company can grow revenue 28% and shrink operating profit only if it is deliberately spending the difference, and Meta says plainly what it is spending on: the annual report guides to capital expenditures of approximately $115 billion to $135 billion in 2026 to support its AI efforts and core business, against $69.69 billion of property-and-equipment purchases in 2025 and $115.80 billion of cash generated from operations that year. At the top of the guided range, one year of capital spending exceeds an entire year of operating cash flow.
That is the inflection. Not a demand problem — a financing-and-return problem, disclosed in advance, arriving in the margin line on schedule.
How the Business Makes Money, and What It Says It Will Do
Meta reports two segments. Family of Apps — Facebook, Instagram, WhatsApp, Messenger, Threads — generates substantially all revenue from advertising. Reality Labs holds the Quest devices, the Horizon store, and what the filing calls long-term research for products "not on the market today and may only be fully realized in the next decade."
The advertising machine is not decaying. Family daily active people averaged 3.58 billion in December 2025, up 7% year over year. Ad impressions across the family rose 12% in 2025; average price per ad rose 9%. Both legs of the ad equation grew, which is the distinction that matters — a network selling more impressions at a lower price is losing pricing power, and Meta is doing neither.
The stated direction is explicit in the filing's own language: the company is "innovating in artificial intelligence (AI) technologies to build transformative experiences and capabilities across our Family of Apps and new platforms, and to advance our vision to deliver personal superintelligence for everyone." The spending is the strategy, not an overrun.
Two commitments give the capex a shape. Management described a BlackRock data-center partnership on the second-quarter call, and reporting circulating in mid-September says Meta intends to deploy an in-house MTIA 450 accelerator, code-named Arke, across its data centers in the first half of 2027, targeting more than a gigawatt of custom silicon capacity. Custom silicon is the cost-curve answer to renting third-party accelerators; it is also a multi-year bet that has to land before the depreciation it is meant to relieve arrives.
Financing is already shifting. The company issued $29.91 billion of fixed-rate senior unsecured notes in November 2025 — a company that generated $115.80 billion from operations that year chose to borrow anyway, which preserves the $26.25 billion of buybacks and $5.32 billion of dividends it paid in 2025 while the build runs.
The filed risks that bear on this read, in the company's own framing: the ability to compete for user engagement against services like TikTok; the EU regulatory stack (GDPR, DMA, DSA, the AI Act) and the FTC consent order; and the dual-class structure that leaves voting control with the founder and chief executive. The second quarter carried $2.4 billion of legal charges, and the third quarter of 2025 carried a $15.93 billion tax charge on enactment of the One Big Beautiful Bill Act, $14.03 billion of it a valuation allowance against U.S. federal deferred tax assets. Neither is recurring; both show the tail is fat and lumpy.
What the Street and the Public Conversation Say
The loudest sell-side event since the last view is the September 21 Wells Fargo note, which raised its target to $796 from $640. Its construction is worth stating exactly: revenue estimates unchanged, and the multiple applied to the 2027 earnings estimate lifted from 20 times to 25 times, on AI product progress and the argument that heavy Muse users will eventually need a paid tier. Every dollar of that target increase is multiple. The shares gapped 11.3% on it.
The public conversation could not be read this session. The saved sweep returned zero results with two of its four sources — X and YouTube — unavailable because the capture browser was down. Hacker News and Reddit genuinely returned nothing. That is a gap in the evidence, not a quiet tape, and the Muse reception in particular is unmeasured here.
The second-quarter reaction was positive and narrow: coverage of the July 29 release led with 28% revenue growth to $60.8 billion and a 6.7% move in the shares, with the earnings-per-share decline attributed to the legal charges and the expense-guidance raise treated as the cost of the AI program. Third-quarter revenue guidance of $61–64 billion was read as confirmation the ad engine is intact.
The public conversation is thinner than the market's. The saved sweep returns one substantive market item — the Arke custom-silicon plan and its gigawatt-scale target — against general-interest material about scraping litigation and layoffs. This is low-quality discussion for a company this size, and it is worth naming as such rather than mining for confirmation.
The insider record is the loudest signal in the qualitative layer, and it points the other way. The leading-indicator panel counts zero insider buys against 100 sells, net −$198.7 million. The filed Form 4s show a coordinated pattern: restricted-stock vesting across the officer group on August 18, then sales by the chief financial officer on August 18 and again on August 20, the chief technology officer on August 20, the chief legal officer on August 20, the chief accounting officer on September 8, the chief operating officer on September 10, and the chief product officer on September 11 — 20,000 shares for about $13.0 million. Sales on this cadence are usually pre-arranged and are not evidence of a view. They are also, factually, the entire senior team reducing exposure into a 12.8% monthly rally, and a reader deserves both halves of that sentence.
The Conviction Chain
Claim 1 — the advertising business is compounding, not maturing. Revenue grew 26.2%, 23.8%, 33.1% and 28.0% year over year in the last four reported quarters. Impressions rose 12% and price per ad rose 9% in 2025 on a 3.58 billion daily-active base. Confirms it: a fifth quarter above 20% with price per ad still positive. Breaks it: impression growth holding while price per ad turns negative — that would mean volume is being bought rather than earned.
Claim 2 — the margin compression is capital allocation, not deterioration. Operating margin fell from 43.0% a year ago to 30.9% in the June quarter while revenue grew 28%; research and development went from 27.2% of revenue to 35.6%. The offsetting evidence is that the compression lives entirely below the gross line: gross margin has held between 81.4% and 82.1% for eight straight quarters, including the latest. The cost of delivering an ad has not moved; the cost of building what comes after it has. Confirms it: margin stabilizing in the 30s while revenue growth stays above 20%. Breaks it: margin below 30% for two consecutive quarters, which would mean the spend is outrunning even an accelerating top line.
Claim 3 — the capital cycle is financeable. Operating cash flow of $115.80 billion in 2025, $15.46 billion of cash and $74.80 billion of short-term investments at June 30, and a demonstrated willingness to borrow ($29.91 billion of notes). Confirms it: 2026 capex landing in the lower half of the $115–135 billion guide. Breaks it: a guide raise above that ceiling accompanied by a buyback cut — the buyback is the release valve, and cutting it is how the market would learn the build is no longer self-funding.
The weakest link is Claim 2, and it is weak for a structural reason rather than a factual one: no one outside the company can currently price the return on $115–135 billion of capital expenditure. The margin line is the only readout available, it is deteriorating, and it will keep deteriorating through the depreciation ramp whether or not the investment is working. That is what caps conviction here at medium. An investor is being asked to underwrite a return that will not be separately visible for several years, on the evidence of a top line that is genuinely accelerating.
Financial Proof
Quarterly series, from the filed statements. Revenue, operating income and research spending in billions; margin and growth computed from the same rows.
| Quarter | Period end | Revenue | Rev YoY | Op income | Op margin | Net income | Diluted EPS | R&D | R&D % rev |
|---|---|---|---|---|---|---|---|---|---|
| 2024 Q3 | 2024-09-30 | $40.59B | — | $17.35B | 42.7% | $15.69B | $6.03 | $11.18B | 27.5% |
| 2024 Q4 | 2024-12-31 | $48.38B | — | $23.36B | 48.3% | $20.84B | $7.98 | $12.18B | 25.2% |
| 2025 Q1 | 2025-03-31 | $42.31B | — | $17.55B | 41.5% | $16.64B | $6.43 | $12.15B | 28.7% |
| 2025 Q2 | 2025-06-30 | $47.52B | — | $20.44B | 43.0% | $18.34B | $7.14 | $12.94B | 27.2% |
| 2025 Q3 | 2025-09-30 | $51.24B | +26.2% | $20.54B | 40.1% | $2.71B | $1.05 | $15.14B | 29.6% |
| 2025 Q4 | 2025-12-31 | $59.89B | +23.8% | $24.75B | 41.3% | $22.77B | $8.87 | $17.14B | 28.6% |
| 2026 Q1 | 2026-03-31 | $56.31B | +33.1% | $22.87B | 40.6% | $26.77B | $10.44 | $17.70B | 31.4% |
| 2026 Q2 | 2026-06-30 | $60.80B | +28.0% | $18.77B | 30.9% | $15.85B | $6.18 | $21.66B | 35.6% |
The September 2025 quarter is not comparable to the quarters around it: its bottom two rows absorb a one-time tax charge of $15.93 billion recognised on enactment of the One Big Beautiful Bill Act, of which $14.03 billion was a valuation allowance against U.S. federal deferred tax assets. Operating income that quarter, which the charge does not touch, was $20.54B.
Trailing twelve months through the June quarter: revenue $228.25 billion, net income $68.10 billion, diluted earnings per share $26.54, net margin 29.8%. Excluding that one-time tax charge, trailing earnings per share is $32.73.
Annual, from the SEC filing: 2025 revenue $200.97 billion, operating income $83.28 billion, net income $60.46 billion, diluted earnings per share $23.49, operating cash flow $115.80 billion, capital expenditures $69.69 billion, and $72.22 billion including finance-lease principal payments. Cash and equivalents $15.46 billion and short-term investments $74.80 billion at June 30, 2026. Headcount 78,865 at year-end 2025, up 6%.
Catalysts
- Meta Connect, September 23–24, keynote 4:00pm Pacific on the 23rd, developer sessions the following morning. The first event since Muse shipped and the first chance to restate its engagement with a number. Reporting ahead of it points to new Ray-Ban variants running Muse Spark and to Muse Voice Transcribe, a real-time audio perception model. The stock has already paid for good news here.
- Third-quarter results, guided to $61–64 billion of revenue. The number that matters is not revenue but whether operating margin stabilizes in the 30s or takes a second step down.
- Muse monetisation. The bull case circulating with the September 21 re-rate assumes heavy users eventually convert to a paid subscription. No such product has been announced. Any pricing disclosure is the difference between an engagement story and a revenue line.
- 2026 capital-expenditure guidance updates. A move toward the $135 billion ceiling changes the financing question; a move toward $115 billion relieves it.
- The MTIA 450 deployment, targeted for the first half of 2027. Any slip pushes the cost-curve relief past the depreciation ramp it was meant to offset.
- The EU regulatory stack — DMA and AI Act enforcement, and the FTC consent order — which the filing names as an ongoing risk to the ad product itself.
Risks and Kill Tests
- The spend does not convert. Kill test: two consecutive quarters of operating margin under 30%.
- The ad engine slows while the build continues. Kill test: revenue growth under 20% year over year with capital-expenditure guidance unchanged.
- Pricing rather than volume turns. Kill test: average price per ad negative year over year on positive impression growth.
- The technical structure fails again. Kill test: a daily close under $563. This is not hypothetical — it is the level the desk's own July call named, and it broke on twelve sessions between late June and mid-August, low close $539.03 on July 30.
- Governance. Class B voting control rests with the founder and chief executive; there is no mechanism by which outside holders can slow the capital program.
Refresh History
One row per view. A refresh rewrites the top region at the current tape; the previous top's values land here.
| Date | Price | RSI | vs SMA20 | 30D | What changed |
|---|---|---|---|---|---|
| 2026-09-22 | $736.60 | 76.0 | +16.4% | +34.0% | Muse launched September 8 and Wells Fargo re-rated to $796 from $640 on an unchanged revenue forecast, gapping the shares 11.3% on September 21. Levels re-anchored from the reclaim trade to the September 21 gap: entry $670–$690, stop $665, first target $765. Connect keynote September 23. |
| 2026-09-15 | $670.24 | 71.6 | +7.5% vs 200-day | +13.6% | Third attempt at reclaiming the moving averages, first to hold two closes above $665. Entry $665–$680, stop $563, target $780. Operating margin 30.9% against 43.0% a year earlier; insiders zero buys against 100 sells, net −$198.7 million. |
Cross-references
- Desk call
2026-07-09-meta-second-try-at-the-turn— watch above $665, invalidation under $563, review date 2026-08-31 now passed. The flag record shows the full path: in-zone July 12, invalidation July 30, in-zone again September 14 with a recorded verdict. - Tracked on the
mag7,faang,tech,ai-scan,bargain-bin,insider-universeandtech-insider-buyswatchlists, and on three of the AI-build-out lane lists (data-center operators, custom silicon, nuclear energy for data centers). - Not a member of any monster tier — parabolic, secular, or steady climber — on the 2026-09-14 board, and not on the washout-turn board.
- Convergence alarm at 3 of 5 bells: research, tape and regime lit; the trigger bell is a price alert only and the advisory bell has no suggestion row.
Sources
Primary filings and company releases
- Meta Platforms Form 10-K for fiscal 2025, filed 2026-01-29 — segment structure, the 2026 capital-expenditure guidance of approximately $115–135 billion, Reality Labs' ~$19.19 billion 2025 operating-profit impact and the similar-in-2026 guide, the One Big Beautiful Bill Act charge, the November 2025 note issuance, Family of Apps metrics (3.58 billion daily active people, impressions +12%, price per ad +9%), headcount, and the filed risk factors.
- Form 10-Q for the quarter ended 2026-06-30, filed 2026-07-30
- Meta Reports Second Quarter 2026 Results, 2026-07-29 — quarterly revenue, the 31% operating margin, and the raised expense guidance.
- Meta Platforms Q2 2026 earnings call coverage, AlphaStreet, 2026-08-03 — the $61–64 billion third-quarter revenue guide, the $2.4 billion of legal charges, and the BlackRock data-center partnership.
- Form 4 filings for the officer transactions between 2026-08-18 and 2026-09-11, routed into the desk's filing captures from
https://www.sec.gov/Archives/edgar/data/1326801/.
The September 2026 Muse launch and re-rate
- Wells Fargo raises Meta price target to $796 on AI progress, Investing.com, 2026-09-21 — the target change, the unchanged revenue estimates, and the 20x-to-25x multiple change on the 2027 earnings estimate.
- Stock Market Today, Sept. 21: Meta Surges on Excitement Over Muse Personal AI Agent, The Motley Fool, 2026-09-21 — the September 8 Muse launch, what the agent does, Muse Spark, and the reported 60% increase in daily assistant users.
- Meta Connect 2026 agenda — the September 23–24 dates and keynote time.
- What to expect at Meta Connect 2026, Engadget — expected Muse updates, the Ray-Ban variants running Muse Spark, and Muse Voice Transcribe.
Attributed, not independently verified
- @wallstengine, 2026-09-14 — the MTIA 450 "Arke" accelerator plan and the stated gigawatt-plus custom-silicon target for the first half of 2027. This is a single social attribution with no filing behind it; it is written as reporting, not as fact.
Desk receipts
- mag7 — the 2026-09-22 settled close, all tape values and peer comparisons.
- 2026-09-22-social-search-meta-meta-platforms-muse-ai-agent — the September 22 social sweep, partial: X and YouTube unavailable.
- income-statements and
cash-flow-statements.json— the eight-quarter series. - company-facts and
key-financials.json— the SEC XBRL annual figures. - 10K-2026-01-29.html — the filing text quoted above.
- latest — the news pull, fetched 2026-09-16T01:58Z,
ok: true. - 2026-09-16-social-search-meta — the saved social sweep.
- 2026-08-18-meta-4-*,
2026-08-20-meta-4-*.md,2026-09-08-meta-4-*.md,2026-09-10-meta-4-*.md,2026-09-11-meta-4-*.md— the Form 4 captures. - CALL-FLAGS — the flag path on the July call.
- 2026-07-09-meta-second-try-at-the-turn — the prior desk call.