PepsiCo PEP

Reference table

1. What the business is, and how the disclosure actually works

PepsiCo describes itself in the 10-K as "a leading global beverage and convenient food company with a complementary portfolio of brands, including Lay's, Doritos, Cheetos, Gatorade, Pepsi-Cola, Mountain Dew, Quaker and SodaStream." Incorporated in Delaware in 1919, reincorporated in North Carolina in 1986. Chief executive Ramon Luis Laguarta; 306,000 employees; headquartered in Purchase, New York; listed on NASDAQ Global Select; CIK 0000077476.

The six reportable segments — and the vendor is one taxonomy behind

This is the single most important disclosure-structure point in the file, and it is a live defect. The 10-K filed 2026-02-03 states at Note 1: "We are organized into six reportable segments" — and lists them:

1. PepsiCo Foods North America (PFNA) — all convenient food businesses in the US and Canada. Cheetos, Doritos, Fritos, Lay's, Pearl Milling Company, Quaker, Ruffles, Tostitos, plus Sabra (the Strauss Group's 50% acquired December 2024).

2. PepsiCo Beverages North America (PBNA) — all beverage businesses in the US and Canada. Aquafina, Bubly, Gatorade, Mountain Dew, Pepsi, Propel. Also distributes Celsius, Alani Nu and Rockstar for Celsius Holdings, and acts as exclusive small-format/foodservice distributor for Tropicana Beverages Group.

3. International Beverages Franchise (IB Franchise) — international franchise beverage businesses plus SodaStream. 7UP, Aquafina, Gatorade, Mirinda, Mountain Dew, Pepsi, Sting Energy.

4. Europe, Middle East and Africa (EMEA) — convenient foods and company-owned-bottler beverages across Europe, the Middle East and Africa. Chipsy, Sasko, Spekko, Walkers, White Star; plus the Russian dairy brands Agusha, Chudo and Domik v Derevne.

5. Latin America Foods (LatAm Foods) — convenient foods in Latin America. Emperador, Marias Gamesa, Sabritas, Saladitas Gamesa.

6. Asia Pacific Foods — convenient foods in Asia Pacific including China, Australia, New Zealand and India. BaiCaoWei, Kurkure, Smith's.

The vendor's seg_prod block reports FY2025 on this NEW six-segment basis and FY2024 and every prior year on the SUPERSEDED seven-segment basis — with one line carrying the wrong name entirely. The FY2025 line labelled "Africa, Middle East and South Asia" at $18.025B is not AMESA; AMESA was $6.217B the year before. It is the new EMEA segment, and the label is a leftover. The vendor also supplies a line called "International Beverage Franchise" that appears only in FY2025 — correctly, because the segment is new — and a "PepsiCo Foods North America" line that replaces the old Frito-Lay North America and Quaker Foods North America lines.

The arithmetic, however, is clean, and we verified every year.

No double-counting, no vanished segment, no over-summing. The defect is purely one of labelling and of comparability across the reorganisation.

FY2025 revenue by reportable segment, relabelled to the 10-K's own vocabulary:

North America is 59.3% of the company. That is the half where unit volumes are falling.

Year-on-year segment comparison is NOT available from this file and we are not going to estimate it. The FY2024 lines are on the old taxonomy (Frito-Lay North America $24.755B, Quaker Foods North America $2.676B, Europe $13.874B, AMESA $6.217B, Latin America $11.718B, PBNA $27.769B, APAC $4.845B) and PepsiCo did not recast them in the extracted text. The only observation that survives the reorganisation is that PFNA at $27.528B in FY2025 compares with Frito-Lay plus Quaker Foods at $27.431B in FY2024 — a 0.4% increase — and PBNA at $28.197B against $27.769B, a 1.5% increase. Both roughly flat, both consistent with the volume declines discussed below.

Geography — and Russia is bigger than the screen suggests

seg_geo is clean and sums exactly. FY2025 lines total $93,925M against reported revenue of $93,925M.

The 10-K corroborates the vendor exactly: "Our operations outside of the United States generated 44% of our consolidated net revenue in 2025, with Mexico, Russia, Canada, China, the United Kingdom, Brazil and South Africa, collectively, comprising 25% of our consolidated net revenue in 2025." The vendor's non-US share is 44.4%; the seven named countries sum to $23.756B, or 25.3%. This is one of the cleanest vendor-to-filing reconciliations in the batch and it is worth recording.

Russia deserves its own paragraph, because it is larger on the balance sheet than in revenue. The 10-Q filed 2026-07-09 discloses: "Our operations in Russia accounted for 6% and 5% of our consolidated net revenue for the 12 and 24 weeks ended June 13, 2026, respectively. Russia accounted for 6% of our consolidated assets, including 21% of our consolidated cash and cash equivalents, and 38% of our accumulated currency translation adjustment loss as of June 13, 2026." The 10-K puts the cash figure at 20% at 2025-12-27. A fifth of PepsiCo's cash sits in a jurisdiction under currency controls, and more than a third of the accumulated translation loss is attached to it. The 10-K adds that cash outside Russia subject to currency controls was "not material" — i.e. Russia is essentially the entire currency-control exposure. That is a real qualification on the headline cash balance and it is why the corrected net-debt figure in Section 4 matters more than usual.

Customer concentration

The 10-K is explicit: "In 2025, sales to Walmart Inc. (Walmart) and its affiliates, including Sam's Club (Sam's), represented approximately 14% of our consolidated net revenue, with sales reported across all of our segments... The loss of this customer would have a material adverse effect on our PFNA and PBNA segments." Roughly $13.1B of revenue through one buyer, disclosed by the company as material to the two largest segments.

The fiscal calendar, and the 53-week trap

PepsiCo's fiscal year ends on the last Saturday of each December, per the 10-K: "Our fiscal year ends on the last Saturday of each December, resulting in a 53rd reporting week every five or six years." Quarters run 12 / 12 / 12 / 16 weeks; the 8-K of 2026-07-09 confirms the second quarter as "the 12 and 24 weeks ended June 13, 2026." International operations are reported on a monthly calendar basis, one month in arrears — so the 12 weeks to 2026-06-13 include March, April and May outside North America.

We checked every year in the file for the extra week, using the true fiscal dates from seg_prod (the income statement dates are normalised and wrong — see Section 8):

FY2022 was the 53-week year. Its reported revenue growth of +8.70% therefore contains roughly 1.9 percentage points of extra calendar — the 10-K's own organic-revenue definition adjusts this out "every five or six years." Any five-year compound growth rate struck through FY2022 is inflated. We use FY2023-FY2025 for trend and say so.

2. The single most important thing to understand about these numbers: the vendor's operating income is not PepsiCo's operating profit

The vendor reports FY2025 operating income of $13.491B, up 4.7% on FY2024's $12.887B. The 10-K states, in plain words: "Operating profit decreased 11%."

Both statements are internally consistent and one of them is not PepsiCo's income statement. The vendor's operatingIncome is computed as gross profit less selling, general and administrative expenses. PepsiCo's income statement carries a separate "Impairment of intangible assets" line below SG&A and inside operating profit — and the vendor drops it into the residual bucket nonOperatingIncomeExcludingInterest instead.

We rebuilt the line and it reconciles to the filing three separate ways.

The identity operatingIncome − nonOperatingIncomeExcludingInterest = incomeBeforeTax + interestExpense holds in every period, which tells us the residual bucket contains everything between the two. For FY2025 the residual is $2.126B. The 10-K discloses pre-tax impairment charges of $1.9 billion ($1.5 billion after-tax, or $1.11 per share) "in impairment of intangible assets primarily comprised of the Rockstar brand in our PBNA, EMEA, and IB Franchise segments", plus a separately-described Be & Cheery brand impairment in Asia Pacific Foods, and separately states that "Other pension and retiree medical benefits expense increased $111 million." Splitting the residual into an impairment line of ~$2.022B and a pension expense of ~$104M gives:

> Reported FY2025 operating profit = $13.491B − $2.022B = $11.469B

> Reported FY2024 operating profit ≈ $12.858B

> Change = −10.8%, i.e. the 10-K's "decreased 11%."

Check two — the June quarter. The 10-Q states "Operating profit increased 125%" for the 12 weeks. Our rebuild gives Q2 FY2026 operating profit of $4.023B (no impairment in the period) against Q2 FY2025 of $1.789B (a $1.860B impairment inside a vendor operatingIncome of $3.649B). $4.023B ÷ $1.789B = +124.9%.

Check three — the half year. The 10-Q states "Operating profit increased 65%" for the 24 weeks. Our rebuild gives $7.236B against $4.372B = +65.5%.

Three independent confirmations. The rebuild is correct and the vendor's operating-income series is not PepsiCo's operating profit in any year that carries an impairment.

Why this matters more than a footnote. Read straight off the vendor file, PepsiCo's operating margin looks like this: 14.32%, 14.04%, 13.33%, 13.10%, 14.03%, 14.36% — a business that bottomed in FY2023 and has been expanding margin for two years. Read off the filings, FY2025's reported operating margin is 12.21%, the lowest in the file, and the direction is down. The vendor's version manufactures a two-year margin recovery that did not happen.

Two caveats we state rather than paper over. First, the reported-operating-margin column for FY2020-FY2023 is derived from the same residual and the decomposition is less clean in those years — FY2023's residual is negative $250M even though the 10-K discloses $0.6B of brand and $0.3B of goodwill impairments in that year, which means the FY2023 charges sat somewhere the residual does not capture. We therefore do not treat the pre-FY2024 reported-margin column as precise, and no conclusion in this dive rests on it. Second, FY2022's growth includes the 53rd week.

What survives all of that is the part that is verified from three angles: FY2025 reported operating profit fell about 11%, net income fell 14.0% from $9.578B to $8.240B, GAAP diluted EPS fell 13.7% from $6.95 to $6.00, and net margin fell from 10.43% to 8.77%. That is the honest picture of last year.

3. Growth and margin trajectory — and how much of it is currency and acquisitions

Quarterly, from the vendor income series, with the vendor's operating income and our rebuilt reported operating profit shown side by side:

Five consecutive adjusted-EPS beats, all of them small (+4.4%, +1.3%, +0.9%, +4.5%, +0.5%). The June quarter beat by half a cent on a $2.19 estimate. Revenue beat by 1.0%.

Now the part that matters: the revenue acceleration is not organic.

Reported first-half FY2026 revenue was $43.624B against $40.645B, +7.33%. The 10-Q's own attribution, quoted directly:

> "In the 24 weeks ended June 13, 2026, favorable foreign exchange contributed to net revenue performance by 3 percentage points primarily due to an appreciation of the Mexican peso and Russian ruble."

That leaves roughly 4.3 points before acquisitions. And the segment commentary attributes most of what is left to deals and price, not units:

• PFNA (24 weeks): "Net revenue increased slightly, primarily driven by an increase in organic volume and a favorable impact of an acquisition, partially offset by unfavorable net pricing." Unit volume +1%. In the 12 weeks, net revenue decreased 2% on unfavourable pricing with unit volume even.
• PBNA (24 weeks): "Net revenue increased 8%, primarily driven by a favorable net impact of acquisitions and divestitures and effective net pricing, partially offset by an organic volume decline." Unit volume −3%, with carbonated soft drinks −3% and non-carbonated −3.5%. In the 12 weeks, unit volume −4%.
• IB Franchise (24 weeks): net revenue +10%, of which 3 points FX; unit volume +3%, led by India.
• EMEA (24 weeks): net revenue +13%, of which 6 points FX; convenient foods volume +6%, beverages +1%.
• LatAm Foods (24 weeks): net revenue +16%, of which 12 points FX from the Mexican peso; unit volume −1%.
• Asia Pacific Foods (24 weeks): net revenue +12%, of which 3 points FX; unit volume +10%, led by India and China.

Read that list and the shape of the company is clear. The two North American segments — 59.3% of revenue — are flat-to-down on units, and PFNA is now taking negative pricing. The international foods businesses are genuinely growing on volume. And a large slice of the headline number is the peso and the rouble.

The exact consolidated organic revenue percentage is in a table that was stripped from the extracted filing text and we do not have it. We state that rather than estimate it. What we do have is the 10-K's statement that in FY2025 "unfavorable foreign exchange had a net nominal impact on net revenue performance" — i.e. FX was roughly neutral for the full year — which, combined with the Siete (January 2025, $1.2B) and poppi (May 2025, $1.9B cash plus $0.2B contingent) acquisitions ADDING revenue, means FY2025 organic growth was below the reported +2.3%.

The acquisitions, from the 10-Q: Siete, a Mexican-American foods business, acquired 2025-01-17 for $1.2B cash, goodwill recorded in PFNA, purchase price allocation finalised in Q1 FY2026. poppi (VNGR Beverage, LLC), a prebiotic soda business, acquired 2025-05-19 for $1.9B cash plus contingent consideration with an acquisition-date fair value of $0.2B, goodwill in PBNA, approximately $2.0B of goodwill and other intangibles, allocation to be finalised in Q2 FY2026. The contingent consideration is now marked at $117M against an original $300M maximum, a fair-value decrease of $45M in the 12 weeks and $161M in the 24 weeks, recorded in SG&A — a $161M non-cash credit inside the first-half operating profit, which management itself calls out as "a favorable net impact of acquisition and divestiture-related charges/credits related to our poppi acquisition."

Gross margin is going the wrong way, two quarters running. 55.19% against 55.77% in the March quarter; 54.22% against 54.66% in the June quarter. The 10-Q names the cause in the segment discussion: a 5-percentage-point impact of higher commodity costs in PBNA over 24 weeks (6 points in the 12 weeks), and the 10-K attributes 6 points of PBNA's FY2025 commodity headwind to tariffs specifically.

On tariffs, the 10-Q discloses a live and genuinely unresolved item:

> "During the 24 weeks ended June 13, 2026, the U.S. Supreme Court ruled that many of the tariffs previously imposed under the International Emergency Economic Powers Act (IEEPA) were invalid. Following this ruling, we have submitted, and expect to continue to submit, tariff recovery claims through the U.S. Customs and Border Protection (CBP) seeking a refund of certain eligible IEEPA tariffs. While we have begun to receive, and may continue to receive, refunds of such tariffs, the ultimate recoverability, timing and amount of any such refunds remain uncertain."

That is an unquantified potential recovery against a cost line the company has already told us was worth 6 points of segment commodity inflation. It is not in any estimate in this file, and it is the one genuinely asymmetric positive in the near term.

The tax rate is rising and the file lets us verify it exactly. The 10-Q says "The reported tax rate increased 3.4 percentage points" in the 12 weeks and "1.5 percentage points" in the 24 weeks, "primarily reflecting the prior-year release of federal interest accruals and the impairment of the Rockstar brand, as well as the current-year impact of the OECD model global minimum tax." Computing from the vendor income statements: Q2 FY2026 tax rate 22.01% against Q2 FY2025's 18.59% = +3.4 points; 24 weeks 21.70% against 20.21% = +1.5 points. Exact match on both. The 10-K adds that in 2026 "widespread implementation of the OECD model rules for a global minimum tax rate of 15% came into effect in various countries in which we do business, including European Union member states, resulting in an increase in our income tax provision." A 2-point structural rise on roughly $13B of pre-tax income is about $260M, or $0.19 of annual EPS — roughly 2.2% of the FY2026 consensus.

Research and development does not appear in this file at all — researchAndDevelopmentExpenses is $0 in all six annual periods and researchAndDevelopementToRevenueTTM is 0. PepsiCo does not break R&D out as a separate income-statement line; it sits inside SG&A. The zero is a disclosure artefact, not a data error, but the field is unusable and no R&D figure appears in this dive.

4. Balance sheet and the enterprise-value rebuild

The last balance sheet in the payload is 2025-12-27 — the fiscal 2025 year end, filed 2026-02-03. Two 10-Qs have been filed since. The most recent, covering the 24 weeks to 2026-06-13, was filed 2026-07-09. Its balance sheet table was stripped from the extracted text, so we cannot rebuild it line by line — and we say so rather than estimate around it.

Correction 1 — the vendor's netDebt omits short-term investments, but the error is small here. netDebt of $40.742B nets total debt against cash and equivalents only, ignoring $371M of short-term investments. Corrected: $40.371B. That is a 0.9% adjustment — far smaller than the META ($45.7B) or BRK-B ($321.4B) precedents, but the check is mandatory and we report the result.

Correction 2 — and this is the material one — the balance sheet in the payload is six months stale, and the vendor's own trailing metrics use a fresher one. Two figures in the same file disagree:

• bal_a[0] (2025-12-27) implies net debt of $40.371B corrected.
• km_ttm.enterpriseValueTTM of $232.9736B less market cap of $190.0106B implies net debt of $42.963B, and netDebtToEBITDATTM of 2.3050 against the vendor's own trailing EBITDA of $18.639B independently confirms the same $42.963B.

A $2.592B discrepancy inside one payload. The filings settle the direction unambiguously — debt has grown:

• The 10-Q states the fair value of PepsiCo's debt obligations was $49 billion at 2026-06-13 against $46 billion at 2025-12-27. That is a $3B increase in the market value of the debt stack in six months, during a period the 10-Q also says saw lower interest rates on average debt balances — so it is principal, not discount rate.
• Commercial paper outstanding rose from a FY2025 position embedded in $6.861B of short-term debt to $5.0 billion at 2026-03-21 and $6.1 billion at 2026-06-13.
• PepsiCo issued notes in the first half of fiscal 2026. The 10-Q says: "In the 24 weeks ended June 13, 2026, we issued the following notes" — and the table listing them was stripped from the extraction. What survives is the footnote and the hedge disclosure, which give us a floor: "These notes, issued in euros, were designated as net investment hedges", and separately "we designated $4.5 billion of existing euro denominated debt and $2.9 billion of euro denominated debt issued in February 2026 as net investment hedges." So at least $2.9B of euro-denominated notes were issued in February 2026, and the total issuance is larger and is not recoverable from this file.
• Against that, "$1.6 billion of U.S. dollar-denominated senior notes and €0.5 billion of euro-denominated senior notes matured and were paid" in the 24 weeks.

Standing rule applied: the vendor's bal_a[0] reports net debt of $40.371B corrected; the vendor's own trailing metrics and the 10-Q's fair-value, commercial-paper and issuance disclosures all show it is higher; we use $42.963B and flag that the exact 2026-06-13 balance sheet is not recoverable from the extracted filing text.

The multi-currency stack is real and is visible in the registered-securities list. The 8-K cover pages list fifteen separate registered note series under Section 12(b): 0.750% due 2027; 0.875%, 0.500% and floating-rate due 2028; 3.200% due 2029; 1.125% due 2031; 0.400% due 2032; 0.750% due 2033; 3.550% and 3.300% due 2034; 3.450% due 2037; 3.700% due 2038; 0.875% due 2039; 4.150% due 2047; 1.050% due 2050; 4.050% due 2055. The sub-1% coupons on 2027-2039 maturities are the euro tranches, issued into a negative-rate European market and now sitting on the balance sheet at coupons no dollar issuer could replicate — which is precisely why the fair value of the total stack ($49B) is below its carrying value.

The revolvers were refreshed in May and are entirely undrawn. The 8-K of 2026-05-22 records that PepsiCo terminated both its 2025 facilities and entered two new ones on the same day:

• $5.0B five-year unsecured revolving credit agreement, expiring 2031-05-22, borrowable in US dollars and/or euros, including a $1.2B same-day swing-line subfacility for euro borrowings, expandable to $5.75B, with up to two one-year extensions.
• $5.0B 364-day unsecured revolving credit agreement, expiring 2027-05-21, dollars and/or euros, expandable to $5.75B, renewable or convertible into a one-year term loan.

Both replaced identically-sized 2025 facilities with no outstanding borrowings at termination, and the 10-Q confirms: "As of June 13, 2026, there were no outstanding borrowings under the 2026 Five-Year Credit Agreement or the 2026 364-Day Credit Agreement." $10.0B of committed, undrawn liquidity, expandable to $11.5B.

The credit rating is not stated in the file but the collateral trigger tells you where it sits. The 10-Q: "Certain of our agreements with our counterparties require us to post full collateral on derivative instruments in a net liability position if our credit rating is at A2 (Moody's Investors Service, Inc.) or A (S&P Global Ratings) and we have been placed on credit watch for possible downgrade or if our credit rating falls below either of these levels." The exposed amount was $53M at 2026-06-13 and no collateral has been posted, so no trigger has been hit.

Interest-rate mix is drifting floating. "As of June 13, 2026, approximately 17% of total debt was subject to variable rates, after the impact of the related interest rate swap contracts, compared to approximately 11% as of December 27, 2025." That is consistent with the commercial-paper build, and it means a larger slice of the stack reprices with short rates than three quarters ago.

The rebuilt enterprise value:

> Market cap $190.011B (vendor; 1,364,891,558 cover-count shares × $139.10 gives $189.856B, a 0.08% difference — see Section 8)

> + net debt $42.963B (trailing basis, filing-corroborated as rising)

> + noncontrolling interests $0.141B

> = corrected enterprise value $233.115B

Correction 3 — trailing EBITDA. The vendor's ebitdaMarginTTM of 19.2345% on trailing revenue of $96.904B implies EBITDA of $18.639B, built bottom-up from net income. We build it top-down from the operating line: trailing operating income $14.495B + trailing depreciation and amortisation $4.358B = $18.853B (19.5% margin). The gap is 1.1% and arises from noncontrolling interests and the residual bucket; unlike the KLAC case, neither figure is absurd and the difference is immaterial to any conclusion. We use $18.853B and disclose both.

What the balance sheet actually says. Net debt of $42.963B against $18.853B of trailing EBITDA is 2.28x — genuine leverage, and it is being used. In FY2025 the company issued $8.2B of long-term debt against $4.1B of repayments (10-K), a net $4.1B increase, while returning $8.6B to shareholders. In FY2024 issuances were $4.0B against $3.9B of repayments — essentially neutral — against $8.2B of returns. The step change in net borrowing happened in FY2025 and it funded the poppi and Siete acquisitions ($3.15B of net cash) plus a shareholder return that exceeded free cash flow.

Three further balance-sheet observations that are genuine and not accounting artefacts.

Working capital is negative by design, and the vendor's measure of it is wrong. workingCapitalTTM is −$2.331B and the current ratio is 0.934 — normal for a staple with fast inventory turns. But daysOfPayablesOutstandingTTM of 200.5 days and the resulting cashConversionCycleTTM of −94.5 days are computed against averagePayablesTTM of $24.415B, which is not trade payables — it is trade payables plus accrued expenses plus other current liabilities. Against actual accounts payable of $11.704B and trailing cost of sales of $44.618B, days payable outstanding is roughly 96 days, not 200, and the true cash conversion cycle is roughly +10 days, not −94.5. We reject the vendor's payables-based ratios as lenses.

There is a reverse-factoring programme, disclosed and sized. The 10-Q: "We maintain voluntary supply chain finance agreements with several participating global financial institutions... As of March 21, 2026 and December 27, 2025, $1.6 billion and $1.7 billion, respectively, of our accounts payable are to suppliers participating in these financing arrangements." That is roughly 14% of accounts payable sitting in a structure that lengthens PepsiCo's payment terms without appearing as debt. It is disclosed, it is not hidden, and it is not large enough to change the credit picture — but it is the kind of item that belongs in an EV rebuild discussion and it is not in any vendor field.

Off-balance-sheet: the company denies material exposure, and we checked. The 10-Q states: "We do not have guarantees or off-balance sheet financing arrangements, including variable interest entities, that we believe could have a material impact on our liquidity." The only unrecorded contingent item we found is the poppi earn-out, capped at $300M and marked at $117M, payable on performance milestones by the third quarter of 2027. Purchase, marketing and other contractual commitments are referenced in the liquidity discussion but the table was stripped from the extraction and no figure is available. Stated as a gap.

Tangible book value is deeply negative and the vendor's field disagrees with its own balance sheet. Goodwill of $18.916B plus other intangibles of $15.066B is $33.982B, or 31.6% of total assets, against total shareholders' equity of $20.406B — tangible equity is approximately −$13.6B. The vendor's tangibleBookValuePerShareTTM of $1.4568, implying positive $1.99B, cannot be reconciled with any combination of the reported lines and is rejected. priceToBookRatioTTM of 8.60x and priceToFairValueTTM (identical, 8.60x — the vendor's "fair value" field is simply book value) are arithmetically correct and economically meaningless against $41.832B of treasury stock. returnOnInvestedCapitalTTM of 13.21% and returnOnCapitalEmployedTTM of 18.80% are the honest return measures and are what we use; returnOnEquityTTM of 50.4% is a leverage artefact.

5. Cash flow and capital returns

Read that table before reading anything else about this stock. In four consecutive fiscal years PepsiCo has returned more cash to shareholders than it generated after capital spending — 137%, 97%, 115% and 113% of free cash flow. The gap was funded by borrowing, which is exactly what the balance sheet shows.

The 10-K corroborates every line of the FY2025 and FY2024 financing picture verbatim: "In 2025, net cash used for financing activities was $5.0 billion, primarily reflecting the return of operating cash flow to our shareholders through dividend payments and share repurchases of $8.6 billion, as well as payments of long-term debt borrowings of $4.1 billion, partially offset by proceeds from the issuances of long-term debt of $8.2 billion." And for 2024: "$8.2 billion" of returns against "$4.0 billion" of issuance and "$3.9 billion" of repayment.

A definitional note that matters, because it changes the coverage ratio. PepsiCo defines free cash flow as "net cash from operating activities less capital spending, plus sales of property, plant and equipment", and the 10-K reports "net capital spending of $3.9 billion" for 2025 against the vendor's gross capex of $4.415B. On the company's own definition FY2025 free cash flow was approximately $8.2B and the $7.638B dividend was covered 1.07x. On the vendor's gross-capex definition it was $7.672B and the dividend was covered 1.00x. Both are defensible; neither leaves room. We use the gross-capex figure throughout for consistency with the vendor series and disclose the company's alternative every time it matters.

Trailing free cash flow has improved to $9.282B on $13.456B of operating cash flow and $4.174B of capex — capital intensity has fallen from 6.0% of revenue in FY2022 to 4.3% trailing. That improvement is the single best number in this dive and it is what gives the 2026 plan its 96% coverage rather than 112%. Whether it holds is the medium-term watch item.

The dividend, verified from the filings — and the vendor's field is mislabelled. The vendor reports lastDividend: 5.7475 and dividendPerShareTTM: 5.7475, yielding 4.132%. That $5.7475 is the trailing-twelve-month total, not the last declared dividend. The 10-Q gives the actual sequence:

> "Cash dividends declared per common share were $1.4800 and $1.4225 for the 12 weeks ended June 13, 2026 and June 14, 2025, respectively, and $2.9025 and $2.7775 for the 24 weeks ended June 13, 2026 and June 14, 2025, respectively."

So the declared quarterly rate was $1.4225 in Q1 FY2026 and $1.4800 in Q2 FY2026. Three quarters at $1.4225 plus one at $1.4800 is $5.7475 — exactly the vendor's TTM figure, correctly computed and wrongly labelled. The 10-K states the forward rate directly: "on February 3, 2026, we announced a 4% increase in our annualized dividend to $5.92 per share from $5.69 per share, effective with the dividend expected to be paid in June 2026."

Forward dividend: $5.92 annualised. Forward yield: 4.256%, not the vendor's 4.132%.

The buyback is a token. $1.0B a year for three consecutive years, 0.53% of market capitalisation, and the 10-K's own 2026 plan holds it at ~$1.0B. Diluted share count went from 1.392B (FY2020) to 1.373B (FY2025) — 1.4% of shrinkage in five years, roughly 0.3% a year. Against $288M of stock-based compensation (0.33% of revenue, low by any standard), the buyback barely does more than offset dilution. The 2026 Share Repurchase Program was announced on 2026-02-03; the authorisation size is in a table that was stripped from the extraction and is not available.

Total shareholder yield is therefore approximately 4.78% — 4.26% dividend plus 0.53% net of the buyback. That is the return you are being paid to wait, and it is genuinely the largest single component of the expected return on this stock.

6. Valuation — priced in or room?

At $139.10 (market cap $190.011B, corrected enterprise value $233.115B):

Estimate coverage is adequate on the near years and thin beyond. 17 analysts on FY2026 and FY2027 revenue, 13 on both EPS lines. FY2028 EPS rests on seven, FY2029 on four, FY2030 on two. FY2029 and FY2030 are excluded from every conclusion in this dive.

And the FY2028 dispersion is the single most informative number in the estimate block. Consensus adjusted EPS of $9.465 carries a low of $7.612 and a high of $10.512 — a 38% spread, with net income ranging from $10.451B to $14.433B. The street has no shared view of this company two years out, which is unusual for a mega-cap staple and is a direct measure of how live the GLP-1 and volume questions are.

Per the data contract, est.ebitdaAvg and est.ebitAvg are treated as unreliable and are not used. We ran the detection signatures and all three are clean: no negative EBITDA in any forward year, no year where ebitAvg exceeds ebitdaAvg, and no ebitdaAvg == −revenueAvg pattern. But the rows are still unusable, for a different reason. The FY2025 ebitAvg of $12.647B sits between PepsiCo's actual reported operating profit of $11.469B and the pre-impairment figure of $13.491B — so the row is struck on neither basis and we cannot tell which. All forward valuation in this dive runs on epsAvg and revenueAvg.

Basis note — and it is not labelled anywhere in the file. The est.epsAvg line for FY2025 is $8.114 against actual GAAP diluted EPS of $6.00. That is a 35% gap. The estimates are struck on PepsiCo's "adjusted for items affecting comparability" basis — the company's own non-GAAP measure, which the 10-K says excludes "mark-to-market gains and losses on centrally managed commodity derivatives that do not qualify for hedge accounting, restructuring and impairment charges related to our 2019 Multi-Year Productivity Plan, charges associated with our acquisitions and divestitures, impairment and other charges/credits, indirect and income tax impacts, product recall-related impact and the impact of settlement, curtailment and certain other gains and losses related to pension and retiree medical plans." The est block never says this. Every forward multiple in this dive is struck on the adjusted basis and the trailing multiple is shown on both (18.23x GAAP, 16.64x adjusted).

The gap is large and it is worth quantifying. FY2025 GAAP diluted EPS was $6.00; the impairment alone was $1.11 per share after tax (10-K, verbatim); the balance of roughly $1.00 per share is restructuring under the 2019 Productivity Plan, acquisition and divestiture charges, commodity mark-to-market, indirect tax settlements and pension items. That is a 26% adjustment, which is high for a consumer staple and which has been recurring for years — the 2019 Productivity Plan alone is a multi-year programme expected to cost approximately $6.15 billion in total, of which $3.6 billion has been incurred through 2025-12-27 and approximately $900 million more is expected in 2026. A restructuring charge that recurs for eleven consecutive years is a cost of doing business. We use the adjusted basis because that is what consensus is struck on, and we flag that the honest economic earnings number sits somewhere between $6.00 and $8.36 and is closer to the middle than to the top.

Peer context. The vendor peer set is reasonable for once — Coca-Cola, Keurig Dr Pepper, Monster Beverage, Coca-Cola Europacific, Coca-Cola Consolidated, Coca-Cola FEMSA, Anheuser-Busch InBev, Unilever, Philip Morris, British American Tobacco. No peer multiples, earnings or revenue are supplied in the file, so no peer-multiple comparison is drawn. One structural fact does survive: PepsiCo's market capitalisation of $190.0B is now 51% of Coca-Cola's $372.4B, and the knowledge base's oldest and highest-conviction entity claim on this name argues that "on beverages Coke is twice Pepsi's operating income and 4x its nearest pure-beverage peer." On the beverage half of the business the claim base thinks PepsiCo is the number two, and the market capitalisations now say the same thing — but PepsiCo also owns a $27.5B North American snack business Coca-Cola does not have, so the comparison is not apples to apples and we do not draw a valuation conclusion from it.

6a. What today's price assumes (the inversion)

At $139.10 — 16.26x FY2026 consensus adjusted EPS and 15.52x FY2027 — the price embeds the following falsifiable claims, each with a number and a date:

• Revenue reaches $99.007B in fiscal 2026 (+5.4% on FY2025's $93.925B) and $101.854B in fiscal 2027 (+2.9%). (Consensus-derived; 17 analysts on both.) The nearest test is 2026-10-08: consensus revenue of $25.003B for the September quarter would be +4.5% year on year. The fragile part is not the total, it is the composition. First-half FY2026 revenue grew 7.3% with 3 points of that from currency by the company's own attribution, and the Mexican peso and Russian rouble do not appreciate forever. If FX turns neutral in the second half, consensus needs organic acceleration to hit $99.0B.

• Adjusted EPS reaches $8.554 in FY2026 and $8.965 in FY2027 — roughly 5% a year. (Consensus; 13 analysts each.) This must happen through a rising effective tax rate: the 10-Q confirms +3.4 points in the June quarter and +1.5 points across 24 weeks on the OECD 15% global minimum tax, worth roughly $0.19 of annual EPS, or 2.2% of the FY2026 estimate. It must also happen while gross margin declines, which it has now done for two consecutive quarters.

• North American unit volume stops falling, or pricing continues to more than offset it. (Our number, from the 10-Q's segment disclosures.) PBNA units fell 3% in the 24 weeks and 4% in the 12 weeks; PFNA units rose 1% in the 24 weeks and were flat in the 12 weeks, on UNFAVOURABLE net pricing. Those two segments are 59.3% of revenue. This is the most fragile assumption in the price. Every other line in this valuation is a price or currency assumption; this one is the actual demand for the actual product, and it is the direct point of contact with the GLP-1 and whole-foods theses in Section 7.

• The dividend is not cut and is raised roughly in line with earnings. (Company-guided, half-weighted as management framing; the 10-K's 2026 plan of ~$7.9B of dividends and ~$1.0B of buybacks.) At 4.26% the yield is the single largest component of the expected return, and the payout has consumed 96-113% of free cash flow in each of the last four years. A cut is not the base case and nothing in the file suggests one — but the price assumes coverage improves, not deteriorates.

• The market keeps paying roughly 15-16x forward adjusted earnings. (Our number.) At 14x FY2027E the stock is $125.5; at 18x it is $161.4. The multiple is worth more than four years of earnings growth, and we have no edge on it.

6b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: adjusted EPS growth (+4.8%, from FY2026E $8.554 to FY2027E $8.965) + multiple drift (HOLD at roughly 16.3x forward, 0%) + shareholder yield (+4.78%) ≈ +9.6%.

Our base case assumes the multiple HOLDS, and we should say why — because on a name in a two-year de-rating that is a real assumption, not a neutral one. The stock has fallen from a 52-week high of $170.49 to $139.10 and its twelve-month total price return is −0.13% against SPY's +24.26%. Something has been re-rating this name downward continuously. We assume it stops here for one specific reason: at 15.5x FY2027 with a 4.26% dividend yield and a 4.88% free cash flow yield, PepsiCo is priced at a level where the income buyer becomes the marginal buyer. A staple yielding 4.3% with an investment-grade balance sheet and $10B of undrawn revolvers does not usually de-rate to a 5% yield without a dividend cut or a recession. That is the floor argument, and it is the whole reason this is a Hold rather than an Avoid.

Note carefully what this bridge does NOT contain: any meaningful earnings growth. Consensus adjusted EPS growth is 5.4%, 4.8% and 5.6% for FY2026-FY2028. Roughly half the expected return is the dividend cheque. That is a completely legitimate way to make money and it is the correct way to own this stock — but it means the position is an income position, not a growth position, and it should be sized and judged as one.

Sensitivity, stated openly. If the multiple simply held at today's 15.52x and applied to FY2028E of $9.465, the price would be $146.9 (+5.6%). If it expanded to 17x on FY2027E the price would be $152.4 (+9.6%). If it compressed to 14x on FY2027E the price would be $125.5 (−9.8%). A three-turn band on the multiple is worth roughly 20 points of price and the entire earnings growth of two years is worth 10. The multiple dominates, and we do not control it.

The bull case at $175 requires multiple EXPANSION and we flag that as its fragile leg — 19x FY2027E, a re-rating of roughly 22% from today, which is what would happen if the GLP-1 fear proves overdone and volumes inflect. The bear case at $116 is the mirror: 13.5x on the FY2027 consensus LOW, i.e. both a de-rating and a downgrade, which is what a confirmed structural volume decline looks like.

6c. Variant perception (where we differ, what would surprise)

• We differ from the vendor's version of this company far more than we differ from the street, and that is the honest finding. Consensus is $155.64 across 46 analysts with a "Hold" label, 29 of the 46 at hold and only one strong buy. Our base of $148 is 4.9% BELOW the crowd. When our own arithmetic lands below the street on a name where the knowledge base contributes six claims and none is about the security, we have no edge on the equity and we say so. No-edge names default toward Hold or Watch, and the 4.26% yield is what makes this a Hold rather than a Watch.

• We differ materially on what the reported margin trend is. Every screen built on this vendor file shows PepsiCo's operating margin expanding from 13.10% in FY2023 to 14.36% in FY2025. The 10-K says operating profit decreased 11% and the true FY2025 reported operating margin is 12.21% — the lowest in the file. That is not a rounding difference; it is the difference between a business quietly recovering and a business whose GAAP profitability just made a five-year low on a $2.0B brand write-down. Anyone screening this name on vendor margin data is looking at the wrong picture, and it is the reason our growth-quality score is a 3 rather than a 5.

• We differ on how much of the 2026 revenue acceleration deserves credit. The +7.3% first-half headline is what most commentary will anchor on. The 10-Q says 3 points of it is currency, and the segment discussion attributes most of the rest to the poppi and Siete acquisitions and to pricing, with the largest segment's unit volume down 3%. (Watchable number: PBNA unit volume in the 2026-10-08 print. A third consecutive quarterly decline, or a decline steeper than 4%, converts the GLP-1 sector claim from theory into a P&L line.)

• Positive surprise that would force a re-rate. Three candidates, in order of likelihood. (1) Tariff refunds. The 10-Q discloses that the Supreme Court invalidated many IEEPA tariffs and that PepsiCo has filed recovery claims through Customs and Border Protection and "begun to receive" refunds, with amount and timing uncertain. This is in no estimate in the file and it lands against a cost line the company sized at 6 points of PBNA commodity inflation. (2) North American volume inflecting positive in the October print, which would kill the structural bear case in one number. (3) A dividend increase above 5% in February 2027, which would signal that management sees the free-cash-flow improvement from $7.7B to $9.3B as durable.

• Negative surprise that would break the thesis. (1) PBNA unit volume worse than −4%, or PFNA turning negative on units with negative pricing alongside. (2) Any deterioration in the dividend coverage story — free cash flow reverting toward the FY2025 $7.7B while the payout is $7.9B. (3) A SodaStream goodwill impairment: the 10-K states plainly that at 2025-12-27 "the estimated fair value of the SodaStream reporting unit narrowly exceeded its carrying value" and warns of "further impairment... if future sales and operating profit results are not in line with the forecasted future cash flows." That is a company telling you where the next write-down comes from. (4) Russia. Six percent of revenue, 6% of assets, 21% of consolidated cash and 38% of the accumulated translation loss, in a jurisdiction with currency controls.

Synthos fair values

All three anchors are multiples of the FY2027 consensus adjusted-EPS distribution (mean $8.965, low $8.617, high $9.195, 13 analysts), cross-checked against FY2026E of $8.554.

• Bear ~$116 — 13.5x the FY2027 consensus LOW of $8.617. Cross-check: 13.6x FY2026E; at $116 the forward dividend yield is 5.10%, which is roughly where a staple trades when the market begins to question the payout. The scenario: North American volume decline accelerates, FX turns from a 3-point tailwind to a headwind, free cash flow reverts to $7.7B against a $7.9B dividend, and the market re-rates a shrinking business. −16.6%, offset by 4.78% of shareholder yield.
• Base ~$148 — 16.5x the FY2027 consensus MEAN of $8.965. Cross-check: 17.3x FY2026E of $8.554 gives $148.0 — the same answer from a different year, which is the best kind of confirmation available here. At $148 the forward dividend yield is 4.00%, the free cash flow yield is 4.59%, and corrected enterprise value of $245.1B is 13.0x trailing EBITDA and 2.41x FY2027E revenue. Sensitivity, stated openly: 15.5x gives $138.9, essentially spot; 17.5x gives $156.9, essentially the street. The entire answer lives in a 15.5-17.5x band and we claim no more precision than that. The scenario: revenue lands roughly on consensus, adjusted EPS compounds at ~5%, the multiple holds because the yield floors it, and the investor is paid 4.78% to wait. +6.4%.
• Bull ~$175 — 19.0x the FY2027 consensus HIGH of $9.195. Cross-check: 20.5x FY2026E gives $175.4. At $175 the dividend yield falls to 3.38%. This bull case requires a genuine re-rating of roughly 22% and we flag that as its fragile leg. The scenario: tariff refunds land, North American volume inflects, GLP-1 fear is judged overdone, the free-cash-flow improvement proves durable, and the market pays for a staple again. +25.8%.

Base is 6.4% above spot; asymmetry roughly 1.55:1 to the upside (16.6% down, 25.8% up) before the 4.78% shareholder yield, which applies in every scenario and materially improves the arithmetic on a two-year view. What makes this a Hold rather than a Buy is that the base case is thin, sits below the street, depends on a multiple holding on a chart that has been de-rating for a year, and offers a total return in which the dividend does half the work.

7. Knowledge base — six claims touch PepsiCo, none is a view on the stock, and the one lane that matters never says its name

Raw hits across all sweeps: 426. Used: 12. Discarded: 414. Entity matches naming PepsiCo: 5. Text-only matches naming PepsiCo usefully: 1. Discarded for collision: 47 of the 48 text-only hits in the primary sweep.

Searches run. Primary sweep on PEP, PepsiCo and Pepsi across 51,928 distilled claims: 53 raw hits (5 entity, 48 text-only). Brand and management sweeps on Frito-Lay, Gatorade, Quaker, SodaStream, Doritos and Ramon Laguarta: zero hits, all six terms. Sector sweeps on consumer staples (43), packaged food (4), GLP-1 (315), MAHA (8), soda (3), snacking (0) and food dyes (0).

The three-letter collision is severe and we triaged all 48 text-only matches individually, as required. The breakdown:

The five entity matches, verbatim.

> 2024-02-18 · bullish · conviction 76 · horizon: thesis · entities: Coca-Cola, KO, PepsiCo, Pepsi · channel: business_breakdowns · no named speaker

> "On beverages Coke is twice Pepsi's operating income and 4x its nearest pure-beverage peer, out-spending rivals 3-5x on marketing and taking value share."

> 2026-02-13 · bullish · conviction 65 · horizon: principle · entities: Coca-Cola, PepsiCo · channel: compound_and_friends · no named speaker

> "Market is splitting into disruptible vs non-disruptible; physical, non-replicable businesses like Coke and Pepsi are safe from chatbots."

> 2026-02-03 · bullish · conviction 60 · horizon: thesis · entities: Pepsi, Coca-Cola, Walmart · channel: compound_and_friends · no named speaker

> "Asset-heavy, low-margin staples are ideal AI beneficiaries; an AI layer streamlines operations and drives margin improvement even without volume growth."

> 2025-07-01 · bullish · conviction 75 · horizon: principle · entities: Vuori, Coca-Cola, Pepsi · channel: invest_like_the_best · no named speaker

> "Distinctive sensory experience (fabric feel, first-sip taste) is a durable moat as powerful as the Coke-vs-Pepsi difference."

> 2024-01-31 · bearish · conviction 45 · horizon: principle · entities: Coca-Cola, PepsiCo · channel: dwarkesh · no named speaker

> "Concentrated passive ownership (3-5 funds holding the market portfolio) risks implicit collusion—e.g. picking lazy managers so co-owned rivals like Coke and Pepsi compete less."

And the one text-only claim that survived triage, because it names the company AND carries a number:

> 2026-07-09 · bearish · conviction 85 · horizon: thesis · channel: eurodollar_university · no named speaker · skill weight 0.7

> "US consumer is 'worse than expected'; Pepsi North American beverage volume fell 4%, revolving credit dropped most in ~2 years (>$5B paid down) — demand destruction from the gas-price shock, not resilience."

We verified that datapoint against the filing and it is correct. The 10-Q filed the same day states PBNA "Unit volume declined 4%, primarily driven by a 4% decline in noncarbonated beverage (NCB) volume and a 3% decline in carbonated soft drink (CSD) volume" for the 12 weeks ended 2026-06-13. A claim in this knowledge base that cites a company number, dated the day the number was published, and that reconciles exactly. That is the standard the whole lane should meet and only this one does. Note, however, that the claim is not an equity view — it uses PepsiCo as macroeconomic evidence about the US consumer.

What this lane is, stated without inflation.

Not one of the five entity claims is a view on PepsiCo the security. Four of them use Pepsi as the second half of a Coca-Cola comparison — and in the one that makes a quantitative statement, PepsiCo comes off worse ("Coke is twice Pepsi's operating income" on beverages). The fifth is an antitrust argument that uses Coke and Pepsi as an illustration of common ownership. Two of them — the "safe from chatbots" claim and the "asset-heavy staples are ideal AI beneficiaries" claim — are the closest thing to a bull case the file contains, and both are category claims about staples generally. They support the margin argument in the exponential-potential score and nothing else.

FIVE of the six claims carry speaker: null and are attributed only to a channel slug. Under the 4-lane attribution policy that is the weakest form of sourcing we accept: a channel tells you where a claim was said, not who said it or what their record is. Not one claim in this lane has a named speaker. No claim carries a management speaker_role, so the half-weight management discount does not apply to the name-level lane.

The sector lane that actually matters — and why it is quoted at reduced weight

The GLP-1 and ultra-processed-food cluster is the single most relevant long-horizon claim class for this company, and not one claim in it names PepsiCo. Under the brief's own standard that is sector colour, not name-level conviction. But it would be dishonest to omit it, because PepsiCo's own 10-K names the same two forces in its own demand risk factor. So we quote it, label it, and give it no weight in the conviction rating.

> 2025-11-11 · bearish · conviction 55 · thesis · channel: eli_lilly_ceo · no named speaker

> "GLP-1s cut roughly 800 calories a day, so widespread adoption structurally threatens food companies as users eat meaningfully less."

> 2024-12-04 · bearish · conviction 45 · thesis · channel: eli_lilly_ceo · no named speaker

> "GLP-1 anti-hedonic effect suppresses appetite for food and alcohol; long-term a headwind for food/alcohol companies, though supply must scale first."

> (undated in the returned record) · consumer-staples / alcohol / packaged-food

> "GLP-1 appetite/urge suppression could eventually shrink alcohol and food demand, but only 7-8M Americans currently on drugs—supply constraints limit near-term impact."

> 2026-01-27 · bearish · conviction 60 · thesis · channel: natalie_brunell · no named speaker

> "Ultra-processed foods engineered to be addictive drive the chronic-disease epidemic; consumer shift toward single-ingredient whole foods pressures packaged-food industry."

> 2025-05-04 · bullish · conviction 85 · thesis · channel: all_in · no named speaker

> "Taxpayer SNAP dollars (~$15B/yr on soda) should be redirected from sugary drinks/junk food toward whole foods; state waivers will reshape nutrition spending."

> 2026-01-27 · neutral · conviction 60 · thesis · channel: natalie_brunell · no named speaker · entities: FDA

> "MAHA food-dye/additive and pharma-ad reforms will be tied up in industry lawsuits and reversed by later administrations; mostly symbolic concessions."

Three disclosures on this lane, all of which reduce its weight.

First, and most important: three of these claims are attributed to a channel slug that is the chief executive of a GLP-1 manufacturer. A pharmaceutical chief executive arguing that his company's drug class will structurally shrink food demand is making a claim with a direct commercial interest attached. Under the promoter-class rule this is exactly the voice we weight low with the frame disclosed, and we are disclosing it: these are the strongest bearish statements in the lane and they come from the most conflicted source in it.

Second, the lane contains its own rebuttal, and we quote it deliberately: adoption is described as only 7-8 million Americans with supply constraints limiting near-term impact. The bearish case is structural and slow, not imminent.

Third, one of the six is explicitly a claim that the policy risk is overstated — that MAHA reforms will be litigated and reversed and are "mostly symbolic concessions." We include it because leaving it out would make the lane look more one-sided than it is.

Corroboration from the company's own filing, which is the reason this lane cannot simply be dismissed. The 10-K's demand risk factor lists, in a single sentence, consumer-preference shifts driven by "diet (whether due to changes in consumer behavior and eating habits, increasing use of weight-loss drugs, such as GLP-1 medications, or other factors)"; "changes in funding for or restrictions on the inclusion of our products in benefit programs, such as the Supplemental Nutrition Assistance Program (SNAP) in the United States"; and "concerns or perceptions regarding the processing, nutrition profile and health effects of... ingredients or substances in our products... including due to public statements by government officials." The same filing also lists, as a demand risk, "use of artificial intelligence shopping agents that autonomously select products that may not be ours" — the same technology the claim base says will improve PepsiCo's margins is, in the company's own words, a threat to its shelf position.

Conclusion, stated as a finding rather than dressed up. The Synthos knowledge base has no view on PepsiCo, Inc. It has four Coca-Cola comparisons in which PepsiCo is the smaller party, one antitrust illustration, one macro claim that uses a PepsiCo volume figure as evidence about the US consumer, and a genuinely relevant but company-anonymous bearish sector cluster whose loudest voices are commercially interested. Conviction rating Low, breadth 5, net conviction negative-low, and the thinness of the name-level lane is a direct input to the Hold verdict.

8. Data integrity — what we rejected from the vendor file and why

Fourteen findings, plus five defect classes tested and found clean. PepsiCo's file is not the worst in this batch on the derived metrics, but it carries the single most consequential unlabelled-definition defect we have seen: a vendor operating-income series that points the opposite way to the company's own annual report.

1. inc_a.operatingIncome is NOT PepsiCo's reported operating profit — REJECTED and rebuilt. This is the headline finding. The vendor reports FY2025 operating income of $13.491B, +4.7%; the 10-K filed 2026-02-03 states "Operating profit decreased 11%." The vendor computes gross profit less SG&A and excludes the "impairment of intangible assets" line, which it dumps into nonOperatingIncomeExcludingInterest — a residual bucket which also contains the pension line and whose sign is inverted (a positive value is an expense). The vendor reports operating income of $13.491B; the filing implies $11.469B; we use $11.469B. The rebuild is confirmed three ways: it reproduces the 10-K's "decreased 11%" (−10.8% against a rebuilt FY2024 of $12.858B); it reproduces the 10-Q's "Operating profit increased 125%" for the 12 weeks ($4.023B ÷ $1.789B = +124.9%); and it reproduces the 10-Q's "Operating profit increased 65%" for the 24 weeks ($7.236B ÷ $4.372B = +65.5%). Every operating-margin figure in this dive is computed on the rebuilt basis and labelled. (Defect class 4 — inconsistent computation across adjacent years, in its most consequential form.)

2. The residual decomposition does NOT hold in FY2023 and earlier — the pre-FY2024 operating-margin series is DISCARDED for trend purposes. FY2023's nonOperatingIncomeExcludingInterest is −$250M even though the 10-K discloses $0.6B of brand and $0.3B of goodwill impairments in that year, primarily on SodaStream. The mapping is therefore not stable across the six-year series and we do not use the pre-FY2024 reported-operating-margin figures for any conclusion. Stated as a limit, not estimated around.

3. bal_a[0] is the 2025-12-27 balance sheet — SIX MONTHS STALE, with two 10-Qs filed since and a note issuance inside the gap. The 10-Q filed 2026-07-09 states "In the 24 weeks ended June 13, 2026, we issued the following notes" — and the table listing them was stripped from the extraction. What survives establishes a floor: at least $2.9B of euro-denominated notes were issued in February 2026 (disclosed in the net-investment-hedge footnote), against $1.6B of USD notes and €0.5B of euro notes that matured and were paid. Corroborating evidence that debt rose: fair value of debt obligations $49B at 2026-06-13 against $46B at 2025-12-27; commercial paper $6.1B at 2026-06-13 against $5.0B at 2026-03-21; variable-rate share of total debt 17% against 11% at 2025-12-27. We use the trailing net-debt figure of $42.963B rather than the $40.371B the stale balance sheet implies, and we state that the exact 2026-06-13 balance sheet is not recoverable from this file. (Defect class 1.)

4. The payload contains TWO mutually inconsistent net-debt figures and does not label either. bal_a[0] gives $40.742B (uncorrected) at 2025-12-27; km_ttm.enterpriseValueTTM less market cap gives $42.963B, independently confirmed by netDebtToEBITDATTM of 2.3050. A $2.592B, 6.4% discrepancy inside one file. We use the higher, filing-corroborated figure.

5. bal_a.netDebt omits short-term investments — real but small. $40.742B nets total debt against cash and equivalents only, ignoring $371M of short-term investments. Corrected: $40.371B, a 0.9% adjustment. Reported as a verified instance of the class, materially smaller here than in the META, AAPL or BRK-B precedents. We also considered and rejected netting the $4.564B of long-term investments — the 10-Q describes these as including the Celsius Series A and B convertible preferred shares valued using Monte Carlo simulation with Level 3 unobservable inputs, which is not liquidity. (Defect class 2.)

6. seg_prod mixes TWO segment taxonomies across years and mislabels one FY2025 line — corrected in presentation, accepted in substance. FY2025 is reported on the current six-segment basis (PFNA, PBNA, IB Franchise, EMEA, LatAm Foods, Asia Pacific Foods) and FY2024 and earlier on the superseded seven-segment basis (Frito-Lay NA, Quaker Foods NA, PBNA, Europe, AMESA, Latin America, APAC). The vendor's FY2025 line labelled "Africa, Middle East and South Asia" at $18.025B is the new EMEA segment, not AMESA — AMESA was $6.217B the year before. The 10-K states at Note 1: "We are organized into six reportable segments." We verified the sums for every year and all six match reported revenue exactly (FY2025 $93,925M, FY2024 $91,854M, FY2023 $91,471M, FY2022 $86,392M, FY2021 $79,474M, FY2020 $70,372M). No double-counting. The figures are used, under the 10-K's own vocabulary, and no year-on-year segment growth rate is drawn because the taxonomies are not comparable. (Defect class 8.)

7. seg_geo is CLEAN and corroborated by the 10-K to the percentage point — reported as a verified pass. FY2025 lines sum to $93,925M, exactly reported revenue. The vendor's non-US share of 44.4% matches the 10-K's "Our operations outside of the United States generated 44% of our consolidated net revenue in 2025", and the seven named countries sum to 25.3% against the 10-K's "comprising 25%." The vendor's Russia line of $4.768B (5.1%) is consistent with the 10-Q's disclosure of 6% and 5% of consolidated net revenue for the 12 and 24 weeks to 2026-06-13. This is one of the cleanest geographic reconciliations in the batch.

8. The income-statement, balance-sheet and cash-flow DATES are normalised and wrong for every year except FY2025. inc_a reports FY2024 as 2024-12-31, FY2023 as 2023-12-31, FY2022 as 2022-12-31, FY2021 as 2021-12-31 and FY2020 as 2020-12-31. PepsiCo's fiscal years actually ended 2024-12-28, 2023-12-30, 2022-12-31, 2021-12-25 and 2020-12-26 — the true dates survive in seg_prod. This matters because it destroys the ability to detect the 53-week year from the income statement. Using the correct dates, FY2022 ran 371 days (53 weeks) against 364 for every other year, so its +8.70% reported revenue growth contains roughly 1.9 points of extra calendar — which the 10-K's own organic-revenue definition adjusts out "every five or six years." Flagged, and no five-year compound growth rate through FY2022 appears in this dive.

9. est.epsAvg is struck on PepsiCo's "adjusted for items affecting comparability" basis and NEVER says so. FY2025 epsAvg of $8.114 against actual GAAP diluted EPS of $6.00 — a 35% gap. Every forward multiple in this dive is struck on the adjusted basis and the trailing multiple is shown on both (18.23x GAAP, 16.64x adjusted). The 10-K lists the excluded items: commodity mark-to-market, 2019 Productivity Plan restructuring and impairment, acquisition and divestiture charges, impairment and other charges, indirect and income tax impacts, product-recall impact, and pension settlement and curtailment. The disclosed FY2025 impairment alone was $1.11 per share after tax; the residual ~$1.00 is the recurring items. (Defect class 7.)

10. est.ebitdaAvg and est.ebitAvg pass all three fabrication signatures but are still NOT USED. No negative EBITDA in any forward year; no year where ebitAvg exceeds ebitdaAvg; no ebitdaAvg == −revenueAvg pattern. The detection signatures are clean and we report that. They are nevertheless unusable because the basis is indeterminate: FY2025 ebitAvg of $12.647B sits between PepsiCo's reported operating profit of $11.469B and its pre-impairment figure of $13.491B, so we cannot tell which the row is struck on. All forward valuation runs on epsAvg and revenueAvg. (Defect class 9 — tested, signature-clean, rejected on a different ground.)

11. Forward-year analyst COUNTS collapse and the thin years are excluded. FY2026 and FY2027 carry 17 analysts on revenue and 13 on EPS. FY2028 drops to 16 and SEVEN. FY2029 is 11 and FOUR. FY2030 is 11 and TWO. FY2029 and FY2030 are excluded from every conclusion. Note separately that the FY2028 EPS dispersion is extraordinary — mean $9.465, low $7.612, high $10.512, a 38% spread — which is not a data defect but is a genuine finding about how unsettled the street is on this name.

12. Multiple vendor fields are ZERO or NULL where a real number exists — scanned and reported. researchAndDevelopmentExpenses is $0 in all six annual periods (PepsiCo reports R&D inside SG&A, so this is a disclosure artefact, but the field is unusable and no R&D figure appears in this dive). interestIncome is $0 in every year while netInterestIncome carries the full expense, so interest income is netted and invisible. cf_a.debtRepayment and cf_a.commonStockIssued are null in all four years, hiding the FY2025 issuance and repayment of $8.2B and $4.1B which the 10-K discloses in prose and which we use instead. capitalLeaseObligationsNonCurrent is $0 while capitalLeaseObligationsCurrent equals the full $719M — implausible; PepsiCo's finance leases are not all current. prepaids, taxPayables and deferredRevenue are all $0. (Defect class 11.)

13. tangibleBookValuePerShareTTM cannot be reconciled with the balance sheet — REJECTED. The vendor reports $1.4568 per share, implying positive tangible equity of ~$1.99B. Total shareholders' equity of $20.406B less goodwill of $18.916B and other intangibles of $15.066B is approximately −$13.6B. No combination of the reported lines produces $1.99B. The field is rejected. Alongside it we reject priceToBookRatioTTM and priceToFairValueTTM (both 8.60x — the vendor's "fair value" field is simply book value), returnOnEquityTTM (50.4%, a leverage artefact against $41.832B of treasury stock), grahamNumberTTM ($52.86) and grahamNetNetTTM (−$48.11). returnOnInvestedCapitalTTM of 13.21% is the honest return measure. Note that the vendor's own composite rating scores PepsiCo 1 out of 5 on both price-to-book and debt-to-equity, dragging the overall grade to B / 3 — a rating artefact of a company that has bought back $41.8B of stock, not a judgement about the business.

14. daysOfPayablesOutstandingTTM of 200.5 days and cashConversionCycleTTM of −94.5 days are computed on the wrong denominator — REJECTED as lenses. averagePayablesTTM of $24.415B is trade payables plus accrued expenses plus other current liabilities, not trade payables. Against actual accounts payable of $11.704B and trailing cost of sales of $44.618B, days payable outstanding is approximately 96 days and the true cash conversion cycle is approximately +10 days, not −94.5. Separately, and unrelated to the defect, the 10-Q discloses a voluntary supply-chain-finance programme covering $1.6B of accounts payable at 2026-03-21 and $1.7B at 2025-12-27 — roughly 14% of the payable balance — which is disclosed, is not debt, and appears in no vendor field.

15. lastDividend: 5.7475 is the trailing-twelve-month total, not the last declared dividend — corrected from the filings. The 10-Q states declared dividends of $1.4800 per share for the 12 weeks to 2026-06-13 and $2.9025 for the 24 weeks, implying $1.4225 in the first quarter. Three quarters at $1.4225 plus one at $1.4800 is $5.7475 — the vendor's figure, correctly computed and wrongly labelled. The 10-K gives the forward rate explicitly: "a 4% increase in our annualized dividend to $5.92 per share from $5.69 per share, effective with the dividend expected to be paid in June 2026." The forward yield is 4.256%, not the vendor's 4.132%. (Exactly the defect the prior PEP-class dive predicted.)

16. quote.yearHigh/yearLow disagree with tech.hi52/lo52 — we use tech and say so. The quote block reports a 52-week high of $171.48 and low of $133.73; the computed technical block reports $170.49 and $134.95, discrepancies of 0.58% and 0.91%. The moving averages disagree too: quote.priceAvg50 of $141.5084 against tech.sma50 of $141.1202, and quote.priceAvg200 of $149.74886 against tech.sma200 of $149.629. We use the tech figures throughout, because they are computed from the same six-year close series that produces the drawdown and relative-return figures, and mixing sources within the technical section would make the percentages inconsistent. (Defect class 12.)

Defect classes tested and found CLEAN — each is a real finding.

Currency (class 5) — CLEAN. inc_a[0].reportedCurrency is USD; profile.currency is USD; every bal_a, cf_a, seg_prod and seg_geo record is USD. No currency mixing. This check matters more than usual here because 44% of PepsiCo's revenue is generated outside the United States and the euro-denominated note tranches sit on the balance sheet in dollars after translation — but the statements themselves are struck in a single currency and every multiple in this dive is internally consistent. The FX exposure is economic, not a data defect, and it is discussed in Section 3 with the 10-Q's own 3-percentage-point attribution.

Share count (class 6) — CLEAN, with a 0.08% note. Market cap ÷ price gives exactly 1,366,000,000 — a rounded figure. The 10-Q cover count at 2026-07-02 is 1,364,891,558; the previous 10-Q cover at 2026-04-09 was 1,366,768,315; weighted-average diluted shares were 1,369M in Q2 FY2026 and 1,373M for FY2025. All five figures agree within 0.6% and the trend is a slow, consistent decline — no split, no share class, no partial-unit error. The vendor's rounded count is 1.1M shares above the freshest cover count, overstating market capitalisation by $154M, or 0.08%. Immaterial, corrected in the At a glance table, and stated because the check is mandatory.

est split corruption — CLEAN. Every historic estimate row reconciles to actuals within a sensible margin: FY2025 revenueAvg $93.459B against actual $93.925B (consensus low by 0.5%); FY2024 revenueAvg $92.077B against actual $91.854B. epsAvg × diluted shares reconciles to netIncomeAvg within 1% in every year. No 10x rows, no split artefacts.

Non-equity tripwire (class 13) — CHECKED and PASSED. PEP is common stock, NASDAQ Global Select, CUSIP 713448108, ISIN US7134481081, par value 1⅔ cents per share (from the 8-K registered-securities tag). The price of $139.10 is not par-like; the dividend is variable and was raised 4% on 2026-02-03, not fixed; the 52-week band of $134.95 to $170.49 is a 26.3% range, wider than a fixed-income instrument would show; and volume was 7.999M shares against a 7.837M average, roughly $1.1B of turnover. The one flag worth naming is beta of 0.368, which is close to the sub-0.3 warning level — but that is what a mega-cap consumer staple's beta looks like, and every other test says common equity. This is common equity.

8-K extraction quality — stated verbatim as a finding. The 8-K of 2026-07-09, which announced the second-quarter results, is present in the archive as the cover document ONLY. It states: "Attached as Exhibit 99.1 and incorporated by reference into this Item 2.02 is a copy of the press release issued by PepsiCo, Inc... reporting PepsiCo's financial results for the 12 and 24 weeks ended June 13, 2026." Exhibit 99.1 was not extracted and the file contains no financial figures whatsoever. Separately, the 8-K of 2026-05-08 is NOT a financing event — it is the Item 5.07 annual-meeting voting results (see Section 10). Numeric tables are stripped from all extracted filing text, which is why the 2026-06-13 balance sheet, the note-issuance table, the segment revenue tables, the organic-revenue bridge and the share-repurchase authorisation are unavailable and are stated as gaps rather than estimated.

Not defects, correctly reported and independently confirmed: the FY2025 and FY2024 capital-return figures ($8.6B and $8.2B) and the debt issuance and repayment figures ($8.2B/$4.1B and $4.0B/$3.9B) match the 10-K verbatim; the geographic revenue shares match the 10-K to the percentage point; the effective tax rate computed from inc_q reproduces the 10-Q's "+3.4 percentage points" and "+1.5 percentage points" exactly; stockBasedCompensationToRevenueTTM of 0.329% reconciles to $288M on FY2025 revenue; interestCoverageRatioTTM of 12.85x is consistent with the reported operating and interest figures; and incomeQualityTTM of 1.280 correctly reflects trailing operating cash flow of $13.456B against trailing net income of $10.485B.

9. Technicals

• Price $139.10, DOWN 0.38% (−$0.53) on the day from a previous close of $139.63. −18.41% from the 52-week high of $170.49, and only +3.08% above the 52-week low of $134.95. Position within the annual range: the 11.7th percentile — near the bottom, not the middle.
• Moving averages, both broken: −1.43% below a 50-day average of $141.12 and −7.04% below a 200-day average of $149.63. The 50-day is itself 5.69% BELOW the 200-day, which is a confirmed intermediate downtrend, not a wobble.
• Maximum drawdown from the six-year peak: −29.07%. The current price is 18.4% below the 52-week high, so the stock has recovered part of a deeper decline but remains well inside it.
• RSI 58.1 and MACD −0.39 — and the tension between these two and a −10.4% six-month return has to be described precisely rather than resolved in whichever direction suits the argument. RSI at 58.1 is above the neutral 50 line: over the recent look-back window, up-days have outweighed down-days. MACD at −0.39 is negative, but on a $139 stock that is −0.28% of price — the fast and slow moving averages have all but converged and the signal is close to crossing. Neither of those readings contradicts the six-month decline; they describe a different time window. The correct reading is: the stock fell roughly 10% between February and May, bottomed at $134.95, and has since gone sideways-to-slightly-up. RSI and MACD are measuring the stabilisation. The 50-day, the 200-day and the six- and twelve-month returns are measuring the decline. Both are true. What the combination says is "a base attempt inside a confirmed downtrend", and the level that adjudicates it is the 50-day at $141.12, 1.4% overhead.
• Relative performance is the worst set of numbers in this dive. 3-month −10.02% against SPY +7.59% and QQQ +7.67% — a 17.6-point deficit. 6-month −10.37% against SPY +11.09% and QQQ +15.71% — 21.5 points. 12-month −0.13% against SPY +24.26% and QQQ +30.80% — 24.4 points behind the index over a full year. PepsiCo has been a source of funds for twelve months and the price action has not stopped saying so.
• Beta 0.368 — by far the lowest in this batch. In a drawdown that is worth something; in a melt-up it is why the stock is 24 points behind.
• Sentiment: 1 strong buy, 15 buy, 29 hold, 1 sell across 46 analysts, with a vendor consensus label of "Hold". Consensus target $155.64 (+11.9%), median $155, high $183 (+31.6%), low $134 — 3.7% BELOW spot. A well-dispersed, genuinely multi-analyst target set. The distribution is the tell: 63% of the coverage is at hold.

Today's move and what it does to the entry

PEP closed 2026-08-04 at $139.10, down 0.38% or $0.53 from a previous close of $139.63. It opened at $137.77 — a 1.33% gap DOWN — traded as low as $137.13, and closed at $139.10, which is within 0.08% of the day's high of $139.21 and 0.97% ABOVE the open, on 7.999M shares against a 7.837M average.

Read the intraday shape rather than the headline. The stock gapped down 1.3%, extended the loss to −1.79% at the low, and then recovered the entire day and closed at the high. That is a red day the buyers won back — the opposite pattern to a distribution day, and it is consistent with the RSI and MACD readings that say the decline has stalled. It does not make the day bullish. The close is still red, still below both moving averages, and still only 3.1% above the 52-week low.

The batch context matters and it is the mirror image of this name. 2026-08-04 was a violent semiconductor melt-up: MRVL +12.8%, INTC +10.9%, LRCX +7.9%, MU +7.6%, KLAC +7.0%, AMD +7.0%, AMAT +5.5%, ASML +4.2%. PepsiCo fell. That is exactly what a 0.368-beta defensive does on a high-beta risk-on day, and it is a reminder of what this position is for: it is the thing that does not move when the semiconductors do, in either direction. No company-specific news is in this file for 2026-08-04; the last company event was the 2026-07-09 earnings release, twenty-six days earlier.

So the entry question here is the opposite of the one every other name in this batch poses. Elsewhere the reader is asked whether to pay up for a one-day gap. Here there is no gap to pay for. The question is whether being 18.4% below the 52-week high, 7.0% below the 200-day, 24 points behind the index over twelve months, and 3.1% above the low is a discount or a warning.

Our answer is: mostly a warning, partly a discount, and the split is roughly where the price is. The de-rating has been earned — GAAP earnings fell 13.7% last year, reported operating profit fell 11%, and North American units are shrinking. But at 15.5x FY2027, a 4.26% dividend yield and a 4.88% free cash flow yield, a good deal of that is now in the price, which is precisely why our base fair value of $148 is above spot rather than below it. What it is not is a bargain: our base is only 6.4% up and sits 4.9% below the street's.

The setup that would change this is one of two things, and they are the falsifiers:

• A price near $125 — roughly 14x FY2027E, a 5.1% forward dividend yield — which would put our base $148 at an 18% premium and make the risk/reward genuinely asymmetric on top of the yield. That is 10.1% below today's close and inside the range this stock has traversed in the last six months.
• Or the 2026-10-08 print showing North American unit volume stabilising — PBNA better than −2%, PFNA positive — which would validate the single most fragile assumption in Section 6a and take the structural bear case off the table. At that point paying 16x is defensible, because the uncertainty being paid for has been resolved.

10. Insiders — eight transactions, no purchases, and one small sale into the decline

Total sold: 9,400 shares. Total awarded: 2,883.325 shares. Open-market purchases: ZERO.

The reading, stated carefully. This is a small and unremarkable table and we are not going to overread it. Nine thousand four hundred shares across two officers is roughly $1.4M in total against a $190B company — noise. The Gibbs entries are a new director's initial Form 3 plus his first equity grants, which is a governance event, not a signal.

Two observations are worth making, and they are modest.

First, Eugene Willemsen sold 6,500 shares at approximately $164.45 on 2026-03-04. The stock is now $139.10 — 15.4% below where he sold. That is well-timed, but it is a single officer selling roughly 9% of a directly-held position, which is the profile of a scheduled diversification rather than a call on the stock. We treat it as neutral and note the price only because it usefully brackets how far the de-rating has run: the stock was $164 five months ago.

Second, and more informative in aggregate: there are no open-market purchases anywhere in this file. On a stock 18.4% below its 52-week high and 3.1% above its 52-week low, with a 4.26% yield, not one officer or director bought a share. That is not a bearish signal on its own — insiders at mega-cap staples rarely buy — but on a name whose entire investment case is "the de-rating has gone far enough", the absence of a single insider willing to act on that view is worth recording rather than omitting.

The 8-K of 2026-05-08 records the 2026 annual meeting and is the only governance disclosure in the file. Thirteen directors were elected; KPMG LLP was ratified as auditor for fiscal 2026; the advisory say-on-pay vote passed. Three shareholder proposals were voted DOWN — an independent board chair, a report on human rights oversight, and a report evaluating the treatment of animals within the supply chain.

On activism, the file is unambiguous and we state it as an absence rather than infer a presence. We searched the 10-K, both 10-Qs and all three 8-Ks for activist, Schedule 13D, cooperation agreement, nominee, strategic review, separation and spin-off. The only hit was a routine cross-reference in the 10-K to the proxy statement's procedures by which security holders may recommend director nominees. There is no Schedule 13D, no cooperation agreement, no board settlement, no announced strategic review and no separation proposal anywhere in this filing archive. The three shareholder proposals that were defeated are ESG-governance proposals of the kind that appear on most mega-cap ballots, not an activist campaign. If a shareholder-activism situation exists at PepsiCo, this file does not disclose it, and we are not going to assert one from outside knowledge.

11. Verdict, kill-criteria and flip conditions

Hold.

This is a legitimately ownable income asset trading at a fair price, and it is not a place we would put new money today.

What is genuinely solid, and none of it is in dispute: $96.9B of trailing revenue across six segments and more than 200 countries at a 54.2% gross margin; $13.456B of trailing operating cash flow and $9.282B of free cash flow on capital intensity that has fallen from 6.0% to 4.3% of revenue; a 4.26% forward dividend yield on a rate the board raised 4% on 2026-02-03 and has never cut in the period this file covers; $10.0B of committed, entirely undrawn revolving credit across facilities refreshed on 2026-05-22 and running to 2027 and 2031, both drawable in dollars or euros; a credit standing whose derivative collateral trigger sits at A2/S&P A; 13.2% return on invested capital; beta 0.368; and a 26% adjusted-to-GAAP gap that, whatever one thinks of it, is at least fully itemised in the 10-K.

What we are declining to pay up for: North American unit volumes going backwards — PBNA −3% over 24 weeks and −4% over 12, PFNA +1% on unfavourable pricing — in the 59.3% of the company that sits in North America. A reported revenue acceleration that the company itself attributes 3 points to currency and most of the rest to two acquisitions and to price. A dividend that consumed 99.6% of FY2025 free cash flow and whose 2026 plan, with buybacks, is 95.9% of trailing free cash flow. Net debt of $42.963B, 2.28x EBITDA, that has grown rather than shrunk since the last balance sheet in the file. A $2.0B intangible write-down that took GAAP operating profit down 11% and net income down 14% last year, on brands the company had bought. A base fair value of $148 that is 6.4% above spot and 4.9% below the street's $155.64. And a knowledge-base lane of six claims, none of which is a view on this security, sitting next to a genuinely relevant sector lane that is bearish and never names the company.

The distinction that matters. We are not saying PepsiCo is a bad business — it is a very good one having a bad few years. We are saying that the return on offer is roughly 4.8% of dividend and buyback plus about 5% of adjusted earnings growth, contingent on a multiple that has been falling for a year not falling further; that our own number sits below the street's; that the knowledge base contributes nothing at the name level and something distinctly negative at the sector level; and that the one thing that would resolve the argument — North American unit volume — is 65 days away. In the Synthos frame, a name where our fair value is barely above spot, sits below consensus, and where the conviction lane is empty is a Hold by construction. Manufacturing a differentiated view here would be dishonest. What separates this from a Watch is the 4.26% yield: you are paid to hold it while you wait, and you are not paid to watch it.

Pre-registered KILL criteria — what would take this to Avoid:

• Any indication that the dividend is at risk. Concretely: trailing free cash flow falling below $7.9B — the company's own 2026 dividend plan — for two consecutive quarters. That is the whole thesis, and it is a live number, not a hypothetical.
• PepsiCo Beverages North America unit volume worse than −5% in the 2026-10-08 print, or PepsiCo Foods North America turning negative on units with negative pricing alongside. That is the GLP-1 and whole-foods thesis arriving in the P&L, and this stock has no growth to absorb it.
• A SodaStream goodwill impairment. The 10-K states its fair value "narrowly exceeded" carrying value at 2025-12-27 and warns explicitly of further impairment. A second consecutive year of billion-dollar intangible write-downs would say the acquisition strategy is not working.
• Net debt rising above roughly 2.6x EBITDA while shareholder returns continue to exceed free cash flow. The company is currently funding part of its dividend with borrowing; there is a level at which that stops being financial policy and starts being a problem.
• A material Russia event — write-off, forced disposal or currency-control escalation — against 6% of revenue, 6% of assets and 21% of consolidated cash.
• A break below the 52-week low of $134.95 on company-specific news rather than market beta.

Pre-registered FLIP conditions — what would take this to Buy — Tactical:

• Price below approximately $125 (roughly 14x FY2027E, a 5.1% forward dividend yield) with no deterioration in the fundamental picture. This is the single most likely path to a Buy and it is only 10.1% below today's close. At $125 the base case is +18% plus a 5.1% yield and the risk/reward becomes genuinely asymmetric.
• The 2026-10-08 print showing North American unit volume stabilising — PBNA better than −2% and PFNA positive — which validates the most fragile assumption in the price and removes the structural overhang in a single number.
• Confirmation and sizing of the IEEPA tariff refunds. The 10-Q says PepsiCo has filed claims and "begun to receive" refunds against a cost the company sized at 6 points of PBNA commodity inflation. This is in no estimate in this file. If it is quantified and material, it is free upside.
• A reclaim of the 50-day moving average at $141.12 and then the 200-day at $149.63, converting the current base attempt into a trend change — though on its own this would be a technical flip only and would need the volume leg with it.
• Genuine name-level knowledge-base conviction appearing — a claim about PepsiCo rather than about the category containing PepsiCo, or about Coca-Cola with PepsiCo attached. At present there is none, and that alone is worth roughly one tier.

Where PEP fits in the Synthos Framework Portfolio. The defensive income sleeve, at 0% today with a 1.5-2% target on a fill near $125 or a confirmed stabilisation in the October print. On the batch overlap question: PepsiCo is the only genuinely low-beta name in this batch and is the natural counterweight to a book full of 1.4-beta semiconductor equipment — on 2026-08-04 it fell 0.38% while the wafer-fab complex rose 4-8%, which is precisely the correlation profile a portfolio uses it for. But the standing house philosophy is forward next-exponentials, not trailing compounders, and PepsiCo is not merely a trailing compounder — it is a trailing compounder whose earnings went backwards last year and whose largest market is shrinking in units. The argument for owning it is the 4.26% cheque and the beta; the argument against is that a 4.26% cheque is available from several places that are not being attacked by a new drug class. Logged as a tracked Synthos call (Hold) as of 2026-08-04 at $139.10.

Single biggest risk: the dividend is the reason to own this and it has no cover. FY2025 dividends of $7.638B against $7.672B of free cash flow is 99.6%. FY2024 was 114.5% including buybacks; FY2023 96.9%; FY2022 136.9%. Four consecutive years of returning more than the business generated, funded with $8.2B of debt issuance in FY2025 alone. Trailing free cash flow has improved to $9.282B and management's 2026 plan of ~$8.9B is 95.9% of it — an improvement, and still no margin. If free cash flow reverts toward the FY2025 level while the dividend runs at $7.9B, PepsiCo is borrowing to pay its shareholders in a fifth consecutive year, at 2.28x leverage, into a rising tax rate. Nothing in this file suggests a cut is imminent, and we are not forecasting one. We are saying that the single asset the equity story rests on has no spare capacity in it, and that is a different and more precise statement than "the yield is safe."

Most fragile assumption in the price: that North American unit volume stops declining. Every other assumption embedded in $139.10 is a price, a currency or a cost assumption, and each is observable and manageable. This one is the actual number of bags and bottles people buy, in the 59.3% of the company that sits in the United States and Canada, and it has fallen for consecutive quarters while the company's own 10-K names "increasing use of weight-loss drugs, such as GLP-1 medications" as a driver of shifting diet. PepsiCo has offset the decline with price for two years, and in the June quarter PFNA's net revenue fell 2% on unfavourable net pricing — the first evidence that the offset has a limit. If units keep falling and pricing can no longer cover it, the consensus path of +2.9% and +3.3% revenue growth in FY2027 and FY2028 does not happen, adjusted EPS stalls near $8.50, and at 14x that is $119 — below our bear case. That is what the $116 anchor is describing.

Provenance & disclosures

• Traceability: 6 tagged knowledge-base claims touch PepsiCo (raw hits across all sweeps 426; used 12; discarded 414; entity matches 5; usable text-only matches 1; discarded for collision 47 of 48 text-only hits in the primary sweep; breadth 5; net conviction negative-low). Sources of the six: business_breakdowns (1 claim, 2024-02-18, bullish, conviction 76 — Coke earns twice Pepsi's beverage operating income); compound_and_friends (2 claims: 2026-02-13, bullish, conviction 65, physical businesses safe from chatbots; and 2026-02-03, bullish, conviction 60, asset-heavy staples as AI margin beneficiaries); invest_like_the_best (1 claim, 2025-07-01, bullish, conviction 75, the Coke-vs-Pepsi taste difference as a sensory-moat metaphor); dwarkesh (1 claim, 2024-01-31, bearish, conviction 45, common-ownership antitrust); and eurodollar_university (1 text-only claim, 2026-07-09, bearish, conviction 85, skill weight 0.7, citing PBNA's −4% volume as evidence of US consumer demand destruction — verified against the 10-Q filed the same day). NOT ONE of the six carries a named speaker; all are attributed to a channel slug only, which is the weakest sourcing the 4-lane attribution policy admits. No claim in the name-level lane carries a management speaker_role. NONE of the six is a view on PepsiCo as a security, and this is stated as a finding rather than padded. A separate, clearly-labelled sector lane of 6 further claims (GLP-1 appetite suppression, ultra-processed-food backlash, SNAP soda redirection, MAHA reform durability) is quoted at reduced weight and excluded from the conviction rating because none names PepsiCo; three of those claims are attributed to a channel slug that is the chief executive of a GLP-1 manufacturer and are half-weighted with the commercial conflict disclosed. Searches: primary sweep on PEP, PepsiCo, Pepsi (53 hits); brand and management sweeps on Frito-Lay, Gatorade, Quaker, SodaStream, Doritos, Ramon Laguarta (zero hits, all six terms); sector sweeps on consumer staples (43), packaged food (4), GLP-1 (315), MAHA (8), soda (3), snacking (0), food dyes (0). The 47 discarded text-only matches break down as 38 "peptide" collisions, 4 "Pepe" memecoin/NFT collisions, 2 Coke-vs-Pepsi analogies that are not about the company, 1 "PEPP" collision (the ECB's Pandemic Emergency Purchase Programme — a genuine three-letter ticker collision), 1 "black pepper", and 1 claim naming Pepsi only as a prospective robotics customer of another company (a product/customer claim, not an equity claim). All quotes are verbatim from the stored claim text.
• Data as-of: income statement, balance sheet and cash flow for fiscal 2025 ended 2025-12-27 from the 10-K filed 2026-02-03; quarterly income statements through 2026-06-13 from the 10-Q filed 2026-07-09; the most recent BALANCE SHEET in the vendor payload is 2025-12-27 — six months stale, with the 2026-06-13 balance-sheet table stripped from the extracted filing text and therefore not rebuildable · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873601 = 2026-08-04T20:00:01Z ($139.10, −0.38%; open $137.77; day range $137.13–$139.21; 50-DMA $141.12; 200-DMA $149.63; RSI 58.1; MACD −0.39) · knowledge-base claims 2026-08-04. Fiscal/calendar note: PepsiCo operates a 52/53-week fiscal year ending on the last Saturday of December. "FY2025" means the 52 weeks ended 2025-12-27. Quarters run 12/12/12/16 weeks; international operations are reported on a monthly calendar basis one month in arrears. FY2022 was a 53-week year (371 days) and its reported growth is inflated by roughly 1.9 percentage points as a result. All figures come from the Synthos vendor data file for PEP or from the SEC filings in the PEP archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents and its gaps: 10-K filed 2026-02-03 (fiscal 2025, year ended 2025-12-27); 10-Q filed 2026-04-16 (12 weeks ended 2026-03-21); 10-Q filed 2026-07-09 (12 and 24 weeks ended 2026-06-13); 8-K filed 2026-05-08 (Item 5.07 — 2026 annual meeting voting results; not a financing event); 8-K filed 2026-05-22 (Item 1.01 — termination of the 2025 revolving credit agreements and entry into new $5.0B five-year and $5.0B 364-day unsecured revolvers); 8-K filed 2026-07-09 (Item 2.02 — second-quarter results). The 2026-07-09 8-K is present as the cover document ONLY — Exhibit 99.1, the press release carrying the financials, was not extracted and the file contains no numbers. Numeric tables are stripped from all extracted filing text, so the following are stated as gaps rather than estimated: the 2026-06-13 balance sheet; the table of notes issued in the 24 weeks ended 2026-06-13; the consolidated organic-revenue bridge; the segment operating-profit tables; the items-affecting-comparability reconciliation; the 2026 Share Repurchase Program authorisation size; and the purchase and marketing commitment schedule.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): operating income rebuilt from the vendor's $13.491B (+4.7%) to $11.469B (−10.8%) after establishing that the vendor excludes the ~$2.0B "impairment of intangible assets" line — the rebuild reproduces the 10-K's "Operating profit decreased 11%", the 10-Q's "increased 125%" for the 12 weeks and "increased 65%" for the 24 weeks, all three to within rounding; net debt taken from the stale balance sheet's $40.371B to $42.963B on the vendor's own trailing basis, corroborated by the 10-Q's fair-value-of-debt disclosure ($49B at 2026-06-13 against $46B at 2025-12-27), the commercial-paper build ($5.0B → $6.1B in one quarter), the variable-rate share (11% → 17%) and at least $2.9B of euro-denominated notes issued in February 2026, with enterprise value rebuilt to $233.115B; bal_a.netDebt corrected for $371M of omitted short-term investments; seg_prod identified as carrying two different segment taxonomies across years, with the FY2025 line labelled "Africa, Middle East and South Asia" relabelled to the 10-K's "Europe, Middle East and Africa" and the six-reportable-segment structure restored; the income-statement fiscal dates for FY2020-FY2024 identified as normalised and wrong, with the true dates recovered from seg_prod and FY2022 identified as the 53-week year; est.epsAvg identified as adjusted-basis and never labelled, with every forward multiple struck on that basis and the trailing multiple shown on both; est.ebitdaAvg/ebitAvg passed all three fabrication signatures but rejected because the FY2025 row sits between the reported and pre-impairment operating profit and the basis is indeterminate; lastDividend of $5.7475 identified as the TTM total rather than the last declared dividend, with the 10-Q giving $1.4800 declared for the 12 weeks to 2026-06-13 and the 10-K giving a forward annualised rate of $5.92 (4.256% yield); tangibleBookValuePerShareTTM rejected as unreconcilable; daysOfPayablesOutstandingTTM of 200.5 days rejected as computed on total current operating liabilities rather than trade payables (true figure roughly 96 days); and quote.yearHigh/yearLow ($171.48/$133.73) rejected in favour of tech.hi52/lo52 ($170.49/$134.95). Where vendor and filing AGREED — worth recording: the segment revenue lines sum exactly to reported revenue in all six years; the geographic lines sum exactly and the non-US share of 44.4% matches the 10-K's "44%"; the effective tax rate computed from inc_q reproduces the 10-Q's "+3.4 percentage points" and "+1.5 percentage points" exactly; the FY2025 and FY2024 capital-return totals ($8.6B and $8.2B) and the debt issuance and repayment figures match the 10-K verbatim; and the share count on the 10-Q cover (1,364,891,558 at 2026-07-02) agrees with the market-cap-implied count to 0.08%.
• Basis note: analyst estimates are PepsiCo's "adjusted for items affecting comparability" basis; reported diluted EPS is GAAP. The gap is large and recurring — FY2025 GAAP diluted EPS was $6.00 against a consensus adjusted figure of $8.114, a 35% difference, of which $1.11 per share is the disclosed after-tax impairment and the balance is 2019 Productivity Plan restructuring, acquisition and divestiture charges, commodity mark-to-market, indirect tax settlements and pension items. Forward multiples in this dive use consensus adjusted EPS, and the trailing multiple is shown on both bases (18.23x GAAP, 16.64x adjusted). The 2019 Productivity Plan is an eleven-year programme expected to cost approximately $6.15B in total, $3.6B incurred through 2025-12-27, ~$900M expected in 2026 — a recurring charge that the adjusted basis excludes every year.
• Estimate coverage: 17 analysts on FY2026 and FY2027 revenue and 13 on both EPS lines; 16 and SEVEN on FY2028; 11 and FOUR on FY2029; 11 and TWO on FY2030. FY2029 and FY2030 are excluded from every conclusion. The FY2028 EPS dispersion — mean $9.465, low $7.612, high $10.512, a 38% spread on seven analysts — is reported as a finding about market disagreement, not as a usable estimate.
• Peer note: the vendor peer set is largely appropriate — Coca-Cola ($372.4B), Philip Morris ($291.3B), Anheuser-Busch InBev ($166.6B), Unilever ($138.0B), British American Tobacco ($127.5B), Monster Beverage ($92.1B), Coca-Cola Europacific ($47.0B), Keurig Dr Pepper ($42.3B), Coca-Cola FEMSA ($22.8B), Coca-Cola Consolidated ($14.2B). No peer multiples, earnings or revenue are supplied in the file, so no peer-multiple comparison is drawn. The only cross-name observation made is that PepsiCo's $190.0B market capitalisation is 51% of Coca-Cola's, which is consistent with the knowledge base's highest-conviction entity claim on this name — and which is explicitly NOT used as a valuation argument, because PepsiCo also owns a $27.5B North American snack business Coca-Cola does not.
• Fair-value caveat: the $116 / $148 / $175 anchors are multiples of the FY2027 consensus adjusted-EPS distribution — 13.5x the low of $8.617, 16.5x the mean of $8.965, and 19.0x the high of $9.195. Stated arithmetic, not a discounted cash flow. The base is deliberately sensitivity-disclosed: 15.5x gives $138.9, essentially spot; 17.5x gives $156.9, essentially the street. The whole answer lives in a 15.5-17.5x band and no greater precision is claimed. The base is cross-checked from a different year — 17.3x FY2026E of $8.554 gives the same $148.0. The base case assumes the multiple HOLDS, which on a stock in a twelve-month de-rating is a real assumption and is defended in Section 6b on the grounds that a 4.26% dividend yield brings in the income buyer; the bull case requires roughly 22% of multiple EXPANSION and that is flagged as its fragile leg. All three scenarios carry an additional 4.78% of annual shareholder yield (4.26% dividend plus 0.53% net buyback), which is not included in the percentage returns quoted.
• Timing: second-quarter fiscal 2026 results were released 2026-07-09, twenty-six days before this dive, and the 10-Q covering the 12 and 24 weeks to 2026-06-13 is in the archive — so the most recent quarter is fully reflected. The next print is 2026-10-08, 65 days away (consensus revenue $25.003B, adjusted EPS $2.31) — there is no near-term company-specific catalyst, which is an explicit input to the Hold verdict rather than an oversight. 2026-08-04 was a violent sector-wide semiconductor up-day (MRVL +12.8%, INTC +10.9%, LRCX +7.9%, MU +7.6%, KLAC +7.0%, AMD +7.0%, AMAT +5.5%, ASML +4.2%); PepsiCo fell 0.38%, which is what a 0.368-beta defensive does on a risk-on day and is treated as such.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.