Meta: the ads are fine, but the price needs the AI spending to pay back

Meta reported diluted earnings of $3.83 a share for the third quarter, up 265% from $1.05. In the same release, free cash flow was negative $2.12 billion. The first comparison is bent by a $9.89 billion legal charge this year and by a tax rate of 87% in the third quarter of 2025. The second is what is left after $34.9 billion of capital spending in a single quarter.

At $697.52, the October 21, 2026 regular session close, the shares trade 28.9% above my triangulated Base value of $541.26, a downside of 22.4%. The Bear value is $177.18, 74.6% below the price, and the Bull value of $1,023.37 is 46.7% above it. Confidence is grade C (13/25), the required margin of safety (MOS) is 40%, and the maximum buy price is $324.75. That close came minutes before the results were released. After hours the shares traded at $712.61, which does not change the verdict (see the stress test below).

Valuation route: an advertising platform that is spending heavily on AI infrastructure, valued on owner cash: net income plus depreciation, less working capital, all capital spending including finance lease principal, and stock compensation, after interest. Primary method (70% weight): that cash for the rest of 2026 and for 2027 to 2029, discounted at 10%, plus an exit value of 22 times 2029 cash. Secondary method (30%): 16 times 2029 operating profit, with the same interim cash converted to a before-financing basis, less debt. Classification: quality growth company, 7 of 12 on the scorecard.

What Meta is

Meta is an advertising business with a very large AI and hardware bill attached. Advertising was 97.6% of third quarter revenue ($63.06 billion of $64.60 billion, up 26%). The Family of Apps segment earned $17.12 billion of operating profit, while Reality Labs lost $4.13 billion on $443 million of revenue.

Growth is coming from the auction, not from new users. Ad impressions rose 10% and the average price per ad rose 14%, while daily active people rose 3% to 3.64 billion. A 14% rise in price is what a better ad stack can deliver for a while, not every year, which is why my Base slows revenue growth from about 27.5% this year (my estimate, using the middle of the fourth quarter guidance) to 19.0%, 13.1% and 11.6% over the next three years.

The investment side is where the story changed. Capital spending including finance leases was $34.9 billion in the quarter against $19.4 billion a year earlier, up 80% (my arithmetic). Meta now guides 2026 capital spending at $135 billion to $145 billion, against $115 billion to $135 billion in January. Depreciation was $7.2 billion, up 44%, and long-term debt rose from $58.7 billion at the end of 2025 to $80.9 billion.

The latest results

The release quotes Mark Zuckerberg saying the investments are paying off. The cash flow statement says they are still being paid for: $34.9 billion of capital spending against $32.8 billion of operating cash flow in one quarter.

Two details change how the headline reads. Meta's outlook says 2026 operating profit should exceed 2025 only excluding the legal charge, a qualifier that the quick summaries of the release leave out. And I do not use the headline earnings miss against analyst estimates as an input, because I cannot verify the accounting basis of those estimates.

Which cash to value

Meta's free cash flow subtracts capital spending and finance lease principal, but it leaves stock compensation in as a free add-back. That is only fair if the shares issued cost owners nothing, and at $26.9 billion over the last twelve months it is not a small item. I value the cash that is left after all of it, and I call it owner cash.

In 2025 owner cash was 53.1% of Meta's reported free cash flow. The three-year average is $29.18 billion, or $11.20 a share, and I show it because a recovery thesis can quietly pick its best year. Over the last twelve months owner cash was negative $1.82 billion, so the forecast starts from a loss and nothing here treats the recovery as if it had already happened.

The $1.04 billion distribution from a data center venture sits in investing cash flow, and I do not count it as recurring. Meta does not split its capital spending into renewal and growth. In my 2029 Base the all-in figure is $120 billion; if $70 billion to $95 billion of that merely replaces equipment (my range), owner cash before growth spending would be $92.5 billion to $117.5 billion instead of $67.5 billion. I still deduct all of it in every forecast year, because the growth it buys is what the exit multiple pays for.

Cash before and after financing

Meta is not a heavily indebted company, but it is a borrower now: long-term debt is $80.9 billion against $82.4 billion of cash and securities. The table shows the Base cash at each level and where the balance sheet ends up if owner cash stays in the company.

The rest of 2026 is a drain. I assume fourth quarter operating cash flow of $42 billion and capital spending of $54.2 billion (the $140 billion guidance midpoint less the $85.8 billion already spent), which is negative $12.2 billion on Meta's definition and negative $20.2 billion after stock compensation of $8 billion. All three are my estimates. Cash would end 2026 at $62.3 billion.

The refinancing assumption is explicit: debt stays at $80.9 billion and is replaced as it matures, with net interest and other cost of $3.4 billion a year before tax (my estimate, tested from $2 billion to $5 billion). An extra $1 billion a year of net financing cost lowers 2029 owner cash by $0.82 billion (my arithmetic). The cash path ignores the dividend of roughly $5.4 billion a year (my arithmetic, $0.525 four times on about 2.55 billion shares), which the model treats as part of owner cash and does not add a second time.

The Bear case is tighter. Its owner cash is negative $35.1 billion in 2027 and negative $9.2 billion in 2028, so cash falls to $17.9 billion at the end of 2028 before dividends, and to about $6 billion after them, below the $20 billion I keep as operating liquidity. In that case Meta would have to cut spending, cut the dividend or borrow more. The falling capital spending in my Bear assumes the first of those.

Shares, claims and the equity bridge

I use 2.573 billion diluted shares, the third quarter weighted average, because the release gives no period-end count. At $697.52 that is a model market value of $1,794.7 billion. Basic shares rose 1.4% in a year (2.517 billion to 2.552 billion) while the diluted count stayed flat, so a flat diluted count does not prove that no new shares were issued. Stock compensation is charged as cash, so I hold the share count flat and do not dilute it a second time. A 10% higher count is a separate stress below.

One corporate event matters for per-share numbers. Meta has approved a new non-voting Class C stock, effective January 19, 2027, and plans to pay one Class C share for each Class A and Class B share. Meta says it expects the Class A price to reflect a two-for-one split after the dividend. All my values are per current share: after the distribution, the Base of $541.26 would be about $270.63 on each of two shares, and the buy ceiling of $324.75 about $162.38. A pro-rata distribution creates no value. What it does is give Meta stock to issue without diluting the founder's votes, and a later non-pro-rata issue is a risk that I treat separately, not in the Base.

Debt is deducted only in the secondary method, because the primary method already pays the interest inside owner cash. The legal reserve is $2.1 billion above the $9.89 billion charge, a difference worth $0.82 a share, and the release does not show the liability balance or the payment timing. Restricted cash of $13.13 billion is left out of excess cash; if all of it were recoverable it would add $5.10 a share (my arithmetic), but I cannot confirm what it secures. Operating lease liabilities of $31.17 billion are not deducted because the lease cost is already in operating profit and cash flow. Finance lease liabilities are not in the release, though their principal payments are inside capital spending.

The primary valuation: owner cash and an exit value

The primary method discounts the fourth quarter of 2026 and the three years to 2029 at 10% (a time-value assumption, tested at 8% and 12%) and adds an exit value of 22 times 2029 owner cash. The exit value is a future figure, so it is discounted too.

The primary Base is $465.59. The exit value is 91.4% of equity value (98.4% in the secondary method), over a horizon of 3.2 years and with a dividend of roughly $5.4 billion a year, above the 85% level where I start treating the answer as mostly an assumption about 2029 and beyond. The Bear is at 97.2% and the Bull at 92.4%. I price this dependence once, in the valuation robustness component of Confidence, and not again in the margin of safety. The explicit horizon is short and Meta pays only a small dividend, so none of the owner cash is distributed in the period; the model counts it as value available to owners.

A table default is not a justification for 22 times, so here is what it assumes. It is my build: 18 times as a mature cash multiple, plus 5 for continued growth and franchise strength, minus 1 for capital and governance uncertainty. As a check, if owner cash grew for five years after 2029 and then 3% forever, 22 times at a 10% discount rate needs about 12.7% annual cash growth in those five years (my arithmetic). That is demanding. The $50 billion gap between 2029 capital spending and depreciation would have to earn its keep: at a 20% to 30% after-tax return, it would add $10 billion to $15 billion of annual profit after a ramp of a year or two. That return is my plausibility bridge, not a measured project return. I challenge the multiple from 18 to 26 times and do not present it as a historical median.

The primary Bear is $122.18 (2029 owner cash of $23.01 billion, 18 times) and the primary Bull is $945.82 (owner cash of $117.26 billion, 26 times).

The secondary valuation: 16 times 2029 operating profit

The second method asks a different question. Instead of what cash is left, it asks what the operating profit after depreciation and stock compensation is worth. It uses the same interim cash flows on a before-financing basis, an exit of 16 times 2029 operating profit of $154 billion, and the same excess cash, debt and legal reserve.

The secondary Base is $717.80. I use 12 times for the Bear ($305.51 on 2029 operating profit of $96.1 billion) and 20 times for the Bull ($1,204.33 on $202.5 billion). These multiples are my assumptions, and I could not assemble a same-date peer set, so there is no peer premium or discount here.

What 16 times assumes is easy to see. It is 19.5 times after-tax operating profit of $126.3 billion in 2029 (my arithmetic). Owner cash before financing that year is $70.3 billion, only 55.7% of that profit. The earnings method therefore pays for profit that the cash flow does not show, and it is right only if a good part of the $50 billion gap between capital spending and depreciation is growth that pays back, or if equipment needs replacing less often than depreciation suggests. That is the dispute between the two methods.

The two methods on one basis

Because the two methods use different bases, I convert both to the same enterprise value, using the same net claims of $30.5 billion, and set them next to the market price.

The market sits almost on top of the earnings method: it pays 20.5 times this year's operating profit, against 21.1 times for the earnings method and 13.8 times for the cash method. The cash method's lower number is not a view that Meta's profit is poor. It is the cash that capital spending absorbs in 2027 and 2028 and the lower multiple on cash that has not yet appeared. My 2026 operating profit before the charge is $89.1 billion, a 34.8% margin: guided total expenses of $177 billion at the midpoint, less the $9.89 billion charge, against revenue of $256.2 billion.

A price check against history

I could not build a reconciled series of price to owner cash over time, so history carries no weight. What I can do is set today's price against past cash. The price is 63.3 times, 51.5 times and 77.5 times owner cash for 2023, 2024 and 2025, using each year's share count. Those are today's price over past cash, not past trading multiples. On a three-year average of $29.18 billion the owner cash yield is 1.6%, and on the last twelve months it is -0.1%.

The growth-adjusted check. On my normalized 2026 earnings of $27.31 a share (before the legal charge, 18% tax, after stock compensation) the shares trade at 25.5 times. On my 2027 earnings of $33.91 they trade at 20.6 times, for growth of 24.2% and a PEG ratio of 0.85. That is my own model and not a consensus estimate for the next twelve months, and I do not weight it: reported earnings growth of 265% is a base effect from last year's 87% tax rate and this year's charge, not a durable rate. This check sets no value.

How good is the cash growth?

Owner cash grows from $23.16 billion in 2025 to $67.49 billion in my 2029 Base. That is 30.7% a year, and it needs context before anyone extrapolates it.

The starting point is a low one. Owner cash in 2025 was 11.5% of revenue and the Base takes it to 17.5% in 2029, so a good part of the growth is recovery from heavy spending, not a new level of earning power. The estimates are my scenarios and I could not find a synchronized analyst series for cash flow, so reliability is low. The history is weaker than the forecast: from 2020 to 2025 owner cash grew 1.5% a year and 3.9% a year per share, after adding back the $5.0 billion FTC settlement paid in April 2020 (2020 free cash flow of $23.03 billion plus $5.0 billion, less stock compensation of $6.54 billion).

Two further stresses use the Base. With 10% more shares, 2029 owner cash per share falls from $26.23 to $23.85. With revenue unchanged at $385 billion and the 2029 margin at 37% instead of 40%, owner cash falls from $67.49 billion to $58.02 billion.

Triangulating the values

The Base methods differ by 42.6% (difference over midpoint), far above the 30% level at which I treat the gap as a finding in its own right. The cause is the one described above: capital spending well above depreciation, capitalized very differently by cash and by earnings. I gave the cash method 70% because financing the build-out is the binding constraint on what shareholders can receive, and the earnings method 30% because the spending may buy returns that near-term cash understates. The weights are my judgment. Both methods rest on the same recovery in advertising profit, so their agreement would earn no confidence bonus even if it existed.

Before weighting I split the $648.9 billion gap in equity value between the two Base cases into its causes. The investment treatment explains all of it, and the multiples work against it.

So the dispute is not about the multiple. It is a single question: does the spending above depreciation buy profit that lasts, or does it only keep the business where it is? I keep both values visible. The cash method could be right if replacement is the larger part of that $50 billion, and the earnings method could be right if growth is. The choice does not flip the decision, because the entry price at a 40% margin of safety is $279.35 on the first and $430.68 on the second, and both are far below $697.52. A point buy price is therefore shown, but the range of $279 to $431 matters more than the weighted $324.75.

The decision does not hinge on the weights. Equal weights give $591.70, the cash method alone gives $465.59 and the earnings method alone gives $717.80, which is 2.9% above the price. Even then the maximum buy price would be $430.68, 38.3% below $697.52.

What the price already assumes

The price is not impossible. Each extra dollar of operating profit adds about $0.82 of owner cash at an 18% tax rate, so revenue, margin, replacement spending and time can all make up the difference. But for scale: the best quarterly margin in my comparison was 40.1% a year ago, and the third quarter of 2026 was 35.4% before the charge. These equations apply to one unknown at a time, so they decompose the price and say nothing about what other investors believe.

Bear, Base and Bull

The Bear is an advertising slowdown with poor cost control and a delayed cash recovery, where spending eventually falls with demand. The Bull needs both stronger monetization and wider margins as the new capacity becomes productive, with spending still high in absolute terms. Terminal multiples differ because the long-run growth of the franchise differs; the discount rate, tax rate and share count are the same in all three, so the same operating risk is not charged twice. The returns are modeled outcomes under an economic realization convention, not forecasts of dividends or prices, and they are not probabilities.

One thing the Base requires is worth spelling out. Depreciation is 10.7% of 2026 revenue in my estimate and 18.2% of 2029 revenue, yet the operating margin still rises from 34.8% to 40%. So everything other than depreciation, meaning people, data center running costs, content and the rest, has to fall from 54.5% of revenue to 41.8%. In dollars that is $139.6 billion to $161.0 billion, growth of 4.9% a year while revenue grows 14.5% a year (my arithmetic, using a $27.5 billion estimate of 2026 depreciation). A 15% cut in headcount shows the intent. It does not show the result.

Sensitivity and the case that flips the verdict

No single change creates a Buy at $697.52. The entry price with a 40% margin of safety runs from about $260 to about $389 across the capital spending cases, and from $285 to $364 across the margin cases. The three estimates that decide the answer are 2029 capital spending (plausibly $100 billion to $140 billion), the sustainable margin (37% to 43%) and the exit multiple (18 to 26 times).

The generous combination gives $820.57, above the price, but a 40% margin of safety on it is $492.34, still far below $697.52. A better operating outcome can lift the value above the price, as the Bull case shows, so the conclusion is a conditional Hold and not a statement that the shares must fall. The assumption-sensitive band for the buy price, from the single-estimate tests above, is $260 to $389; a refreshed capital bridge replaces it.

How the buy price is set

Confidence C calls for a margin of safety of 35% to 50%, and I use 40% with no add-ons. The reasons you might want one are already in the confidence score: the exit value is 91% of the primary value and the two methods differ by 42.6%, and both are charged there once (valuation robustness scores 1 of 5). Capital intensity sits in the cash flows, interest in owner cash, the legal settlement in the equity bridge, stock compensation in the cash cost, and concentration and governance in the scenarios and the exit multiple. A 45% margin would put the ceiling at $297.69. The maximum buy price is $541.26 x (1 - 40%) = $324.75. It is a decision reference, not an order to place.

Three numbers answer different questions: what the shares are worth ($541.26), the annual return from buying at $697.52 and following the Base cash path (1.5%), and the price that would give a chosen return on that path. A 12% return is my illustration, not a requirement I know you have: the model gives $511.31 for 12%, $541.26 for 10% and $573.57 for 8%. The return is a modeled owner cash return, not a forecast of dividends, buybacks or share price. At $697.52 the shares are 28.9% above Base, so they do not qualify.

The dividend

Meta declared a quarterly dividend of $0.525 a share, paid on September 28, 2026 to holders of record on September 21. Annualized, that is $2.10 a share and a 0.30% yield at $697.52, assuming the board keeps declaring it. This is a growth position, not an income holding, and the dividend competes with infrastructure spending for the same cash. I do not add it to the value because the owner cash method already values the cash once. The Class C stock dividend is not a cash payment and not a free increase in wealth.

Reasons to own

• The ad engine works. Revenue grew 26% in the quarter on 10% more impressions and 14% higher prices, and the Family of Apps segment earned $17.12 billion of operating profit.
• AI feeds an auction that already pays. Better ranking and ad tools earn cash through a monetization system Meta already runs, so they do not need a new business model.
• Operating leverage is possible. Headcount is down 15% to 66,775 and, if the new capacity becomes productive and spending stops climbing, owner cash could recover quickly. The Bull case of $1,023.37 shows how quickly.
• The distribution is large. Messaging, AI assistants and wearables can reach the existing user base. I put no stand-alone value on them in the Base.
• Meta can fund itself. With $82.4 billion of cash and securities and operating profit of $22.9 billion in a quarter before the charge, it can keep investing, provided commitments stay in proportion.

Main risks

• Replacement cost. AI hardware may need renewing faster than depreciation suggests, leaving owner cash weak for years.
• Return on the spending. Frontier model spending may earn less than its cost of capital even if advertising stays strong. Reality Labs lost $4.13 billion in the quarter.
• Price-led growth. A 14% rise in price per ad, with user growth of 3%, cannot repeat every year.
• Tax and adjusted-earnings optics. Unusual tax rates (87% last year, 17% to 24% this year) can make earnings growth look better or worse than the underlying business.
• Legal and regulatory cash. The $9.89 billion settlement is not the end of regulatory exposure, and the reserve in my bridge is not a cap on future cases or product restrictions.
• Attention and access. Users and advertisers can move, and privacy, platform or content rules can lower ad efficiency.
• Governance and financing. Meta is controlled by its founder and will have a new non-voting stock to issue. Non-pro-rata issuance could dilute owners.
• Off-balance-sheet commitments. The data center venture and other infrastructure obligations need a full filing review that I could not complete from the release.
• Valuation construction. The exit multiple and 2029 cash decide about 91% of the value, and the Bear value is not a floor.
• Currency. For an investor who does not hold dollars, the return also depends on the exchange rate.

Management execution

Execution on cost and execution on capital look different. Total expenses, excluding the legal charge, are guided at $165.1 billion to $169.1 billion for 2026 (my arithmetic, $175 billion to $179 billion less $9.89 billion), against $162 billion to $169 billion guided in January, so the operating cost line is where it was promised. Capital spending is the one that moved: $115 billion to $135 billion in January, $130 billion to $145 billion before this release, and $135 billion to $145 billion now.

The outlook is also precise about what Meta is committing to: operating profit above 2025 only before the legal charge. Fourth quarter revenue guidance of $73 billion to $76 billion is concrete. The payoff from the AI products is an expectation and has not yet turned up as cash. I could not rebuild a multi-year record of guidance against results, so I give no accuracy percentage. I rate credibility 6/10 (my estimate) and management and allocation 7/10 inside business quality.

Stock Analyza scorecard

On the scorecard for non-financial companies Meta scores 7/12, a quality growth company: 0/2 for return on capital, because I could not build a reliable five-year series, 2/2 for five-year revenue growth ($85.97 billion in 2020 to $200.97 billion in 2025, 18.5% a year), 0/2 for five-year owner cash per share (3.9% a year, below 8%), 1/1 for balance sheet capacity, 1/1 for unit economics, 2/2 for repeat advertiser relationships and 1/2 for a runway of high-return reinvestment. The valuation score is 12/100: 0/40 for the discount to value, 0/20 for protection against the Bear, 4/15 for agreement between methods, 0/10 for history, 7/10 for what the price implies and 1/5 for data quality. Business quality and valuation are separate numbers and are never blended.

Confidence score

The low score describes valuation uncertainty, not a forecast that the business fails. The inputs I cannot verify are the live post-results price, the exact share count, the split of capital spending between renewal and growth, and the full third quarter filing. To reach grade B I would need the 10-Q commitments and lease liabilities, two quarters of owner cash trending towards positive, and a narrower spread between the cash and earnings methods.

What I would watch from here

The most important thesis breakers

• Capital spending that does not fall. If replacement cost stays above depreciation and 2028 owner cash lands below $20 billion, the exit value shrinks with it.
• Growth that leans on price. Price per ad rose 14% on a 3% rise in users. If impressions slow while price gains fade, the revenue path in the Base fails first.
• Costs that rise with revenue. The Base needs costs other than depreciation to grow about 4.9% a year. A margin below 30% for two quarters means they are not.

Stress test and independent check

I rebuilt the model from its stated inputs in separate code and reproduced the three primary values ($122.18, $465.59, $945.82), the three secondary values ($305.51, $717.80, $1,204.33), the weighted values ($177.18, $541.26, $1,023.37), the 42.6% divergence, the $324.75 buy price, both sensitivity grids, the reverse results ($102.66 billion and $100.72 billion), the 1.52% modeled return, the cash growth rates and the normalized earnings of $27.31 and $33.91. One small difference: 2027 earnings growth is 24.16%, not 24.15%, and the PEG stays 0.85. Arithmetic was otherwise correct throughout.

From Meta's third quarter release and its SEC exhibit I confirmed revenue, operating profit, the legal charge, earnings per share, cash flow lines, capital spending, stock compensation, depreciation, cash, restricted cash, debt, lease liabilities, share counts, segment results and guidance. From the 2025 results I confirmed free cash flow of $43.59 billion, stock compensation of $20.43 billion and the January guidance, and from the 2021 results the 2020 figures and the FTC payment. The reference price is the October 21 close of $697.52 on stockanalysis.com. After hours the shares were $712.61 (up 2.2%) at 7:59 PM Eastern, and a broker quote early on October 22 showed $696.09, 0.2% below the close. At $712.61 the shares would be 31.7% above my Base, a downside of 24.0%, and the verdict is the same at any of the three.

I could not re-read some inputs, so they come from the supplied analysis: the 2023 and 2024 cash flow and share figures, and the dividend and stock history behind the 2023 to 2025 share counts. I did not have the full third quarter 10-Q, only the earnings release exhibit.

Against the checks added in the newest version of my method: every selected multiple states the growth, durability, reinvestment and leverage it assumes (22 times needs about 12.7% cash growth for five years after 2029, and Meta ends 2029 with net cash of $76.6 billion in the Base); the common-basis table reconciles to the same net claims as the bridge; the 42.6% divergence was decomposed before weighting; the method choice does not flip the decision; and exit value shares are disclosed for both models. The cash flow level test for heavily indebted companies did not trigger, since debt is about 0.7 times 2026 EBITDA by my estimate and interest is well below 15% of pre-financing cash flow, though I show levels A and B anyway.

I changed the supplied analysis in four places. I cut the earnings and cash quality score from 9 to 6 of 15, because owner cash is negative over the last twelve months, which lowers business quality from 76 to 73. I stated the inputs of the generous case, which the first draft left undefined, and reproduced its $820.57. I added the three disclosures the newest version of my method asks for, namely what each multiple assumes, the two methods on one basis, and cash before and after financing. And I added the two-for-one effect of the Class C distribution on per-share figures. None of these changed a value or the verdict.

The strongest case against Hold is that the AI spending is building a durable asset base, so cash today understates what the business will earn. The earnings method already sits near the price, the Bull case is $1,023.37, and 2029 owner cash in the Bull is $117.3 billion. That is a serious case and I have not dismissed it. The strongest case against owning at all is that replacing compute and keeping up in models holds capital spending high indefinitely while taxes and legal costs drift up. Then my Base is too optimistic, and the Bear of $177.18 is 74.6% below the price and not a floor. My Base grants a recovery and does not assume the Bear is certain, and my corrections ran in both directions: the earnings method deliberately challenges an overly harsh cash-only view.

Main data limits: no live post-results quote beyond the after-hours print, no full 10-Q, so no exact share count, finance lease liabilities or commitments, no split of capital spending into renewal and growth, no synchronized analyst consensus or peer valuation, and no way to tell what the restricted cash secures. These are why Confidence is C.

Final verdict: hold at $697.52

Meta doesn't have to be a weak business for $697.52 to be the wrong price. Depreciation is already climbing, $7.2 billion in the third quarter against $5.0 billion a year earlier, and capital spending has not started to fall. The price is a bet on how wide the gap between those two lines will be in 2029. In my Base it is $50 billion a year.

Verdict: HOLD at $697.52, watchlist for new money. Triangulated Base value $541.26, downside 22.4%, Confidence C (13/25), required margin of safety 40%, maximum buy price $324.75, assumption-sensitive band $260 to $389.

For holders this is a review, not an instruction to sell, and portfolio circumstances differ. For new money it means waiting for a lower price with the thesis intact, or for owner cash to turn positive on a rolling basis without a cut to investment. Below the Base is not enough on its own: the buy price assumes a 40% margin of safety on a fragile estimate.

Sources

Primary sources include Meta's third quarter 2026 results and the SEC Exhibit 99.1 with the financial statements, the October 2026 Form 8-K on the Class C stock and the founder agreement, the 2025 fourth quarter and full year results with the original 2026 guidance, the 2025 Form 10-K, the 2024 results, the 2023 results, the 2021 results for the 2020 figures and the FTC payment, and the September 2026 dividend announcement.

The reference market price is $697.52, the October 21, 2026 regular session close, shown on the stockanalysis.com quote page, which also showed the $712.61 after-hours price. Per share values use 2.573 billion diluted shares, the third quarter weighted average. The forecasts, tax and financing assumptions, multiples, the 10% discount rate, the operating liquidity reserve, the legal reserve and the fourth quarter figures are analytical estimates, not company guidance.