Cisco Systems CSCO

EARNINGS BANNER — RESULTS IN EIGHT DAYS

Cisco reports fiscal fourth-quarter and full-year 2026 results on 2026-08-12, eight days after this dive. Vendor consensus is revenue $16.819B and non-GAAP EPS $1.17, which would be +14.6% revenue growth year on year — an acceleration from the +12.0% delivered in the April quarter — and +6.2% sequentially. Because this is the fiscal year end, FY2027 guidance arrives with it, and the company has said it expects to recognise approximately $450 million of a $1 billion restructuring charge in this quarter (8-K, 2026-05-13). The verdict below is set with that timing as an explicit input. A reader with no position has no reason to establish one in the eight days before a print that carries a full-year guide, at RSI 71.6, into a fair value that sits below the price.

Reference table

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

Cisco describes itself in the 10-Q as designing and selling "a broad range of technologies that help to power, secure, and draw insights from the Internet." Chief executive Charles H. Robbins; approximately 86,200 employees as of 2025-07-26 per the 10-K, of whom approximately 25,600 are in worldwide sales and marketing. Listed on NASDAQ; IPO February 1990; fiscal year ends the last Saturday of July, so "FY2025" means the twelve months to 2025-07-26 and "Q3 FY2026" means the quarter to 2026-04-25.

The reportable segments are geographic, and the vendor mislabels one of them

The 10-K, Note 19, is unambiguous: Cisco is "primarily managed on a geographic basis consisting of three segments: the Americas, EMEA, and APJC." Performance is measured on segment revenue and segment gross margin only — research and development, sales and marketing, and general and administrative expenses are not allocated to segments.

The vendor's seg_geo block supplies exactly three lines and they sum exactly to revenue — $33.656B + $14.824B + $8.174B = $56.654B for FY2025, matching reported revenue to the dollar. But the largest line is labelled "UNITED STATES", and it is not the United States. The 10-K states plainly:

> "Revenue in the United States was $30.4 billion, $28.7 billion, and $29.9 billion for fiscal 2025, 2024, and 2023, respectively."

The vendor reports $33.656B under "UNITED STATES"; the filing shows US revenue of $30.4B; the $33.656B figure is the AMERICAS segment. We use the filing. The gap is $3.3 billion, or 10.9% of the line. The corroboration is in the 10-Q's own commentary, which breaks Americas product revenue growth into "the United States, Canada and Mexico by 22%, 9% and 32%, respectively, partially offset by a decline in Brazil of 10%" — four countries inside one segment that the vendor calls a country.

A second, unrelated defect in the same block: the FY2018 row contains a single line, "UNITED STATES: $25.500B", summing to $25.5B against reported FY2018 revenue of $49.3B. That row is broken and is not used anywhere in this dive.

The product categories, and a taxonomy that has changed three times in five years

The seg_prod table reconciles cleanly for the years that matter. FY2025 sums to $56.653B against reported revenue of $56.654B; FY2024 sums to $53.804B against $53.803B. No double-count, no missing line. But the category names change in FY2021→FY2022 and again in FY2023→FY2024, so only FY2024 and FY2025 are comparable to each other and to the 10-Q's quarterly commentary. Everything before FY2024 is on a superseded taxonomy and is presented here for context only.

Note what that annual table says and what the quarterly commentary says, because they point in opposite directions. Across FY2025 as a whole, Networking DECLINED 3.2% and the growth came from Splunk annualising into Security. In the April 2026 quarter, the 10-Q reports the reverse: "Revenue from the Networking product category increased by 25%, or $1.7 billion primarily driven by our AI Infrastructure and Campus Networking solutions", while "Revenue in our Security product category was flat, primarily driven by declines in our prior generation products and Splunk offerings." The engine changed over. That is the single most important fact about the last twelve months of this company, and it is why the annual data alone is misleading.

Customer concentration: none. The 10-K states "No single customer accounted for 10% or more of revenue in fiscal 2025, 2024, and 2023." That is a genuine structural strength and it distinguishes Cisco from most of the AI-infrastructure complex.

2. The inflection, quarter by quarter — and what the filing says is driving it

Every headline in that final row is confirmed verbatim by the 10-Q, and the reconciliation is unusually clean. The filing states "Total revenue increased by 12% compared with the third quarter of fiscal 2025" — the vendor gives 12.0%. It states "Total gross margin decreased by 2.0 percentage points" — 65.6% to 63.6% is exactly 2.0 points. It states "Operating income as a percentage of revenue increased by 2.4 percentage points" — 22.6% to 25.0% is exactly 2.4 points. It states "Diluted earnings per share increased 37%" — the vendor's net income and diluted share counts give +36.0%, within diluted-share rounding. This is one of the cleanest vendor-to-filing ties in the batch and we record it as such.

The growth is one category and the filing says so. From the April-quarter MD&A, verbatim:

> Networking: "increased by 25%, or $1.7 billion primarily driven by our AI Infrastructure and Campus Networking solutions... led by double digit revenue growth in Campus Switching, Data Center Switching, Wireless and Service Provider Routing."

> Security: "was flat, primarily driven by declines in our prior generation products and Splunk offerings."

> Collaboration: "decreased by 1%, or $7 million."

> Observability: "increased by 3%."

> Services: "Services revenue decreased by 1% in the third quarter of fiscal 2026... with the decline primarily driven by lower revenue from support services."

And the backlog is not growing. "Total remaining performance obligations as of April 25, 2026 were flat compared to the end of fiscal 2025." "Total deferred revenue decreased 1%." Total software revenue was $5.7 billion, up 1%, and total subscription revenue DECREASED 2%.

That combination — 12% revenue growth, flat RPO, declining services and declining subscription revenue — is the signature of a hardware product cycle, not a platform transition. It is why the growth score is a 6 and not an 8.

The Q1 FY2025 tax benefit — a one-off that inflates every FY2025 comparison

The vendor's own quarterly data contains the tell and it is worth stating explicitly. In the quarter ended 2024-10-26, Cisco reported pre-tax income of $2.267B and net income of $2.711B — net income exceeded pre-tax income, because the income tax line was a benefit of $444M. The 10-Q corroborates and sizes it: nine-month FY2026 EPS growth of 24% was "partially offset by the income tax benefit of $720 million we had in the first nine months of fiscal 2025."

Consequence: FY2025 GAAP net income of $10.180B and the FY2025 effective tax rate of 8.3% ($920M on $11.100B of pre-tax income) are not clean operating results. Stripping the $720M benefit gives roughly $9.46B of clean FY2025 net income and about $2.37 of clean GAAP diluted EPS against the $2.55 reported — a 7% overstatement. We flag this rather than smooth it, and we note that our trailing-twelve-month window (Q4 FY2025 through Q3 FY2026) does NOT contain the benefit, so the trailing EPS of $3.02 and the 40.3x trailing multiple used throughout this dive are clean. It is the year-on-year comparisons against FY2025 that are flattered.

3. Gross margin, memory, and the working-capital commitment

Product gross margin fell 2.5 percentage points year on year in the April quarter and 1.0 point over nine months. The 10-Q's own attribution table names three drivers — product mix, pricing and productivity — and defines productivity to include "overall manufacturing-related costs, such as component costs (including memory)". The commentary is direct:

> "Product gross margin decreased by 2.5 percentage points primarily driven by negative impacts from product mix and to a lesser extent, pricing, partially offset by productivity improvements. The negative impacts from product mix were primarily due to higher Networking revenue. Productivity benefits were adversely impacted by higher memory costs."

Read that carefully: the mix hit is caused BY the growth. Networking is the fastest-growing and the lowest-margin category, so the better the AI-infrastructure quarter, the worse the blended product margin. That is a structural feature of this cycle, not a temporary cost problem, and it will not reverse if the growth continues.

The memory cost line is the same physical scarcity that runs through this entire batch, and Cisco has responded by committing the balance sheet to it. From the 10-Q, verbatim and in full because it is the most important disclosure in the file:

> "Inventory as of April 25, 2026 increased by 49% and inventory purchase commitments with contract manufacturers and suppliers increased by 111% from our balances at the end of fiscal 2025. The combined increase of 93% in our inventory and inventory purchase commitments as compared with the end of fiscal 2025 was primarily related to commitments with contract manufacturers and suppliers related to manufacturing Cisco Silicon One and other products to meet the demand from hyperscalers and other customers. In the third quarter of fiscal 2026, we increased our purchase commitments with a certain supplier to help secure memory components."

And on the nature of those commitments:

> "Certain of our inventory purchase commitments are entered into directly with suppliers and relate to fixed-dollar commitments to secure supply and pricing for certain product components for multi-year periods. In addition, certain of these inventory purchase commitments are related to long-term supply agreements for fixed quantities of certain memory components for which pricing is variable. A significant portion of our reported purchase commitments arising from these agreements are firm, noncancelable, and unconditional commitments."

And the risk, in the company's own words:

> "This exposure includes potential material excess and obsolete or other charges if product demand significantly decreases for a sustained duration, we are unable to generate demand for certain products, or we are otherwise unable to mitigate this exposure."

We cannot size these commitments in dollars. The 10-Q presents them in tables, and numeric tables are stripped from the current filing extraction — a known limit of this data pipeline, stated rather than estimated around. What we can say with certainty is the direction and the magnitude of change: +49% inventory, +111% commitments, +93% combined, firm and noncancelable, much of it in multi-year fixed-dollar memory contracts struck during a memory shortage. That is the single largest risk on this page.

One favourable item, correctly not taken as income. On 2026-02-20 the US Supreme Court ruled the IEEPA tariffs unauthorised, and Cisco "may be eligible for a refund of tariffs previously paid" — but the company "ha[s] not recorded a benefit for any potential refund and will not until such amounts are realizable" and believes any refund "will not have a material impact." No number in this dive assumes a tariff refund.

4. Balance sheet — where the vendor file is three quarters out of date

This is the most consequential data defect on this name and it is a stale-balance-sheet defect of exactly the class the batch has been finding.

The vendor's bal_a block ends at 2025-07-26. The most recent filing-verified balance sheet is 2026-04-25, nine months later, and in those nine months Cisco's financing position changed materially.

The 10-K ties the FY2025 debt exactly: "As of July 26, 2025, we had $24.8 billion in principal amount of senior fixed-rate notes outstanding. The carrying amount of the senior notes was $24.6 billion", plus "$3.5 billion in commercial paper notes outstanding under this program as of July 26, 2025." Carrying $24.6B + CP $3.5B = $28.1B, against the vendor's $28.093B. Vendor and filing agree on the fiscal-2025 debt figure.

Correction 1 — netDebt omits short-term investments. The vendor nets $28.093B of debt against $8.346B of cash only, ignoring $7.764B of short-term investments which the 10-Q describes as classified short-term "based on their nature and their availability for use in current operations." The vendor reports FY2025 net debt of $19.747B; the balance sheet shows $11.983B; the filing's own language supports the lower figure; we use it. The FY2023 error is more dramatic still: the vendor reports net debt of −$1.732B when the corrected figure is net CASH of $17.755B — a $16.0 billion sign-and-magnitude error on the pre-Splunk balance sheet.

Correction 2 — the balance sheet is three quarters stale and misses $4.9B of net issuance. The 10-Q (2026-04-25) states:

> "We had $8.4 billion and $3.5 billion in commercial paper notes outstanding as of April 25, 2026 and July 26, 2025, respectively."

and the liquidity bridge:

> "The net increase in cash and cash equivalents and investments in the first nine months of fiscal 2026 was primarily driven by net cash provided by operating activities of $8.8 billion, $4.9 billion net issuance of commercial paper, and the release to us of approximately $0.6 billion of restricted cash previously held in escrow. These sources of cash were partially offset by cash returned to stockholders in the form of cash dividends of $4.9 billion and repurchases of common stock of $4.6 billion, repayment of debt of $1.8 billion, shares repurchased for tax withholdings on vesting of restricted stock units of $1.4 billion and capital expenditures of $1.0 billion."

Rebuilt from the filing, as of 2026-04-25:

> Senior notes, principal $24.8B less $1.8B repaid ≈ $23.0B

> + commercial paper $8.4B

> = total debt ≈ $31.4B

> Cash and investments $16.110B plus the ≈$0.6B net increase the MD&A bridge implies ≈ $16.7B

> = net debt ≈ $14.7B

Both of the last two figures are DERIVED from the MD&A's cash-flow bridge, not stated in it, because the balance-sheet tables are stripped from the extraction. We label them as derived and do not present them as filing-stated. The direction, however, is not in doubt: debt is higher and net debt is higher than any figure in the vendor payload.

Correction 3 — the vendor's own enterprise value cannot be reconciled to its own balance sheet. km_ttm.enterpriseValueTTM is $504.050B against a market cap of $479.830B — an implied net debt of $24.220B, which matches neither the vendor's own netDebt of $19.747B nor the corrected $11.983B. The residual is $4.473B, or 22.7% of the stated figure. The vendor's EV is struck off a balance sheet that is not in this payload and we cannot reconcile it. We reject it and rebuild.

> Market cap $479.830B

> + filing-derived total debt ≈$31.4B

> − cash and investments ≈$16.7B

> = corrected enterprise value ≈ $494.5B

Correction 4 — the vendor's trailing EBITDA is not an operating figure. ratios_ttm.ebitMarginTTM of 25.58% implies trailing EBIT of $15.540B, which is $1.349B ABOVE the sum of the last four quarters' operating income ($14.191B). The difference is exactly reconstructible: trailing pre-tax income of $14.075B plus trailing interest expense of $1.465B equals $15.540B. The vendor's "EBIT" is pre-tax income grossed back up for interest, and therefore includes non-operating and investment income. That is a defensible definition but it is not operating EBITDA, and it makes the multiple look 6% cheaper than it is. We use operating income of $14.191B plus trailing depreciation and amortisation of $2.528B = $16.719B (27.5% margin) and compute every EV/EBITDA figure in this dive on it.

What the corrected balance sheet actually says. Net debt of roughly $14.7B against $16.7B of operating EBITDA and $13.288B of fiscal-2025 free cash flow is manageable but no longer trivial — Cisco was net-cash to the tune of $17.8B before the Splunk acquisition and is now net-debt. Interest coverage is 9.69x against a revolver covenant of not less than 3.0 to 1.0, and the company confirms it "was in compliance with all associated covenants" at 2026-04-25 with no funds borrowed under the $5.0B revolving facility. What deserves flagging is not the leverage but its shape: current ratio 0.92, working capital NEGATIVE $2.982B, and $8.4B of the debt sitting in commercial paper that must be rolled continuously.

Goodwill and intangibles are $68.311B — 55.9% of total assets — almost all of it created by the Splunk acquisition, which shows in the cash-flow statement as $25.994B of net acquisition outflow in FY2024. The vendor's intangiblesToTotalAssetsTTM of 53.5% corroborates. Tangible book value per share is NEGATIVE $4.62. That makes priceToBookRatioTTM of 9.85x and debtToEquityRatioTTM of 0.64x low-information metrics here, and it is why the vendor's own composite rating scores Cisco 1 out of 5 on debt-to-equity and 2 out of 5 on price-to-book, dragging the grade to B / 3. Those sub-scores are rejected as lenses; returnOnInvestedCapitalTTM of 12.3% is the honest return measure.

5. Cash flow and capital returns

Capital intensity is trivial and that is the whole point of the model — capital expenditure was 1.6% of revenue in FY2025 and capexToOperatingCashFlowTTM is 3.1%. Cisco's competitive asset is silicon design, software and channel, and it does not build fabs.

The nine-month FY2026 picture, from the 10-Q: operating cash flow $8.8 billion "includes our final U.S. transition tax payment of $2.3 billion" — a real cash outflow that will not recur, so the FY2026 headline understates the run-rate by that amount. Against it: $4.9B of dividends, $4.6B of buybacks, $1.4B of shares repurchased for RSU tax withholding, $1.0B of capital expenditure and $1.8B of debt repayment — $13.7B of outflows against $8.8B of operating cash flow, with the gap bridged by $4.9B of net commercial paper issuance. That is worth stating plainly: over the last nine months Cisco's capital return exceeded its free cash flow, and the difference was funded with short-term debt.

Buyback authority remaining is $9.6 billion at 2026-04-25, "with no termination date" — about 2.0% of market capitalisation in dry powder.

Dividend, and a vendor labelling point. lastDividend: 1.66 and dividendPerShareTTM: 1.66 are the trailing-twelve-month total, not the last declared payment. The 10-Q gives the actual rate: "On May 13, 2026, our Board of Directors declared a quarterly dividend of $0.42 per common share to be paid on July 22, 2026." Forward run-rate is therefore $1.68, a 1.38% forward yield, against the vendor's trailing 1.364%. Payout ratio on trailing GAAP earnings is 53.8%, and the company states its policy directly: "we target to return a minimum of 50% of our free cash flow annually to our stockholders through cash dividends and repurchases of common stock."

Total shareholder yield is approximately 2.6% — 1.38% dividend plus roughly 1.27% from a $6.1B annualised buyback pace. That is a meaningful support and it is the strongest single argument for holding rather than selling.

6. Valuation — priced in or room?

At $121.74 (market cap $479.830B, corrected enterprise value ≈$494.5B, trailing revenue $60.746B, trailing net income $11.958B):

The epsAvg row passes an independent check and we record it, because the data contract says to test rather than assume. FY2026E epsAvg of $4.276 should equal the four quarterly non-GAAP figures. From earn_cal: Q1 actual $1.00, Q2 actual $1.04, Q3 actual $1.06, Q4 estimate $1.17 — sum $4.27. An exact tie. That also confirms the estimate rows are non-GAAP, and that the reported quarterly beats have been consistently small (+1.8%, +2.0%, +2.9%, +1.9% across the last four prints — never more than 3%).

The est EBITDA and EBIT rows are NOT arithmetically impossible here — we checked. ebitAvg is below ebitdaAvg in every forward year and neither is negative, so this file trips none of the known corruption signatures. FY2026 ebitdaAvg of $18.980B against our computed trailing operating EBITDA of $16.719B is directionally sensible. We nonetheless run all forward valuation on epsAvg per the data contract, and note that FY2026 ebitAvg of $16.429B is closer to the vendor's non-operating "EBIT" definition than to operating income.

One estimate row IS rejected. The FY2024 row reports epsAvg of 2.29216 and niAvg of $9.187B. Actual FY2024 GAAP diluted EPS was $2.541 ($10.320B / 4.062B shares) and actual non-GAAP was materially higher still. The row matches neither basis and is discarded; no conclusion in this dive touches it. By contrast the FY2025 row — epsAvg 3.7889 — ties closely to the four reported non-GAAP quarters of FY2025, so the defect is confined to FY2024.

Peer context. The vendor peer set is largely unusable for this purpose: Applied Materials, Lam Research and Micron are semiconductor capital equipment and memory; Salesforce, SAP and Uber are software and platforms; Qualcomm is mobile silicon. The only genuine comparable in the list is Arista Networks ($239.9B market cap), and IBM ($221.5B) is the relevant "cheap incumbent re-rating on AI" analogue — which is exactly the pairing the knowledge base made in March 2025. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn.

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

At $121.74 — 28.5x FY2026 consensus and 25.4x FY2027 — the price embeds the following falsifiable claims:

• Revenue reaches $62.901B in the fiscal year ending 2026-07-26 (+11.0%) and $68.755B in FY2027 (+9.3%). (Consensus-derived; 18 analysts each.) The near test is eight days away: the fourth quarter must do $16.819B, +14.6% year on year, which would be an acceleration from the +12.0% just delivered. This is the most immediately falsifiable bullet on the page.
• Non-GAAP EPS of $4.276 in FY2026 and $4.787 in FY2027, i.e. roughly 12% annual EPS growth for two years. (Consensus; 14 analysts each.) Cisco's fiscal-2021-to-2025 EPS compound rate on a GAAP basis was negative — $2.50 to $2.55 diluted over four years. The price assumes a business that has not grown earnings per share in five years now compounds them at 12% for two.
• The market keeps paying 23–25x forward at every date between now and FY2028. (Our number.) Cisco's own knowledge-base record says it traded near 16x in March 2025. At 20x FY2027E the stock is $96; at 25x it is $120; at 30x it is $144. The multiple, not the earnings, is where the variance lives — and it has already travelled from 16x to 25x.
• Product gross margin stabilises. (Our assumption; the filing points the other way.) It fell 2.5 points year on year in the April quarter, and the 10-Q attributes the mix component of that decline to "higher Networking revenue" — i.e. the growth itself compresses the margin. This is the most fragile assumption in the price, because it is the only one that gets worse if the volume assumptions get better.
• The $93% combined increase in inventory and purchase commitments converts to revenue rather than to charges. (Filing-stated increase; our assumption on the outcome.) The 10-Q explicitly warns of "potential material excess and obsolete or other charges."

Where our own signal has a reading: the knowledge base's specific, dated call on this name — Visser, 2025-03-02 — was that Cisco should re-rate to a market multiple from ~16x. At 25.4x FY2027E that has been delivered in full. The priced-in gap our own library identified is closed.

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

Expected return over the next twelve months decomposes as: EPS growth (+12.0%, FY2026E $4.276 to FY2027E $4.787) + multiple drift (COMPRESSION, from 25.4x to roughly 23x on the respective forward year, −9%) + shareholder yield (+2.6%) ≈ +5% to +6%, against a base fair value that is 1.4% below spot.

Our base case assumes multiple compression, and the reason is in the consensus estimates themselves, not in a view of our own. The forward EPS growth path is +12.9%, +12.0%, +10.1%, then +1.7%. The revenue path is +11.0%, +9.3%, +6.8%, then +0.4%. A multiple appropriate to 12% growth is not appropriate to 2% growth, and the market normally begins discounting that a year ahead of the deceleration. By FY2028 the market is looking at FY2029, and FY2029 consensus is a flat year.

To be explicit about the guardrail: 28.5x FY2026 becoming 25.4x FY2027 at a constant price is mechanical rolldown, not compression. The compression we assume is a genuine change in what the market pays at each date — from 25.4x forward today to roughly 23x forward in twelve months — and we assume it because the growth being purchased is visibly shorter-dated than it was.

Almost none of the expected return is re-rating, and that is a change. For the last eighteen months, re-rating was the whole return: the stock rose 81.4% while EPS rose roughly 12%. The re-rating leg is now spent. What is left is 12% EPS growth against 9% assumed compression and 2.6% of yield. If instead the multiple simply HELD at 25.4x on FY2028E of $5.272, the price would be $134 (+10.1%); if it compressed to 20x, $105 (−13.7%).

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

• We differ from the street by very little, and we differ from our OWN prior view by a lot — which is the honest headline. The street is at $126.44 (+3.9%) with 38 buys and 34 holds; we are at $120 (−1.4%). The interesting divergence is internal: the Synthos knowledge base was correctly and specifically bullish on this name from March 2025, and that thesis has been realised. We are not reversing it. We are retiring it. A closed gap is not a sell signal; it is the absence of a buy signal.
• We differ on how to read the growth quality, and this is our one genuine variant perception. The consensus narrative is "Cisco is an AI infrastructure winner." The filings say something narrower: one product category grew 25% while services revenue FELL 1%, subscription revenue FELL 2%, remaining performance obligations were FLAT and deferred revenue FELL 1%. A platform company converting to recurring revenue does not show that pattern; a hardware supplier in a strong product cycle does. Watchable number: services revenue growth and total RPO in the 2026-08-12 print. Services returning to growth with RPO up would genuinely change our reading and would justify a higher multiple. A second quarter of declining services with flat RPO confirms it.
• We differ on the working capital, and we think it is under-discussed. Every screen shows Cisco with a strong balance sheet. The 10-Q shows inventory +49%, purchase commitments +111%, a combined +93% increase, in firm noncancelable multi-year fixed-dollar memory contracts, plus commercial paper up from $3.5B to $8.4B. That is a company taking real physical and financing risk to serve a hyperscaler order book, and it is not visible in any ratio in the vendor file. Watchable number: the inventory and purchase-commitment disclosure in the FY2026 10-K, due roughly early September.
• Positive surprise that would force a re-rate upward: fourth-quarter revenue meaningfully above $16.819B together with FY2027 revenue guidance above roughly $70B — which would be above consensus and would argue the cycle extends past the modelled decay; or services revenue returning to growth with RPO up; or product gross margin flat to up year on year, killing the memory-cost concern.
• Negative surprise that would break the thesis: FY2027 guidance implying revenue below roughly $66B; a second consecutive quarter of declining services revenue with RPO still flat; an excess-and-obsolete charge against the enlarged inventory and commitment book; product gross margin down more than 2.5 points again; or a restructuring charge materially above the announced $1B. Any of these hits a stock 6.4% off its high at RSI 71.6 with a target band that tops out at $137.

Synthos fair values

All three anchors are multiples of the FY2027 consensus non-GAAP EPS distribution (mean $4.787, low $4.636, high $5.007, 14 analysts — the deepest coverage in the file), cross-checked against FY2028E of $5.272 (8 analysts, used only as a check).

• Bear ~$88 — 19x the FY2027 consensus LOW of $4.636. Cross-check: 16.7x FY2028E; roughly the 200-day moving average of $90.01. The scenario: the AI product cycle proves to be a two-year pull-forward, services keeps shrinking, an inventory charge lands against the enlarged commitment book, and the multiple returns toward the 16x the stock carried before the re-rating. −27.7%.
• Base ~$120 — 25x the FY2027 consensus MEAN of $4.787. Cross-check: 22.8x FY2028E of $5.272 gives $120.2 — the two anchors agree to within 0.2%, which is unusual and worth stating. Sensitivity, stated openly: 23x gives $110 (the street's target low); 27x gives $129. The entire answer lives in a 23–27x band. The scenario: FY2026 and FY2027 land roughly on consensus, the multiple drifts down modestly as the FY2029 flat year comes into view, and the shareholder yield of 2.6% carries the rest. −1.4%.
• Bull ~$140 — 28x the FY2027 consensus HIGH of $5.007. Cross-check: 26.6x FY2028E. Slightly above the street's target high of $137. This bull case requires BOTH an earnings beat and a further re-rating from a multiple that has already moved from 16x to 25x — which is why it is the least likely of the three. The scenario: AI Infrastructure demand extends beyond the modelled decay, Silicon One takes share, services returns to growth and the market pays a platform multiple. +15.0%.

Base is 1.4% BELOW spot; asymmetry is roughly 0.54:1 — unfavourable (27.7% down against 15.0% up). That negative skew, at RSI 71.6, eight days before a print carrying a full-year guide, is the arithmetic behind the Hold.

7. Knowledge base — 29 raw hits, 21 discarded, and a concentration to disclose

Raw hits: 29. Entity matches: 13. Text matches: 16. Used as name-level conviction: 8. Discarded: 21. Search terms: CSCO, Cisco, Cisco Systems, word-boundary matched across entity lists and claim text over all 51,928 distilled claims.

The discards are the story, and they are a textbook case of why raw hit counts must never be reported as coverage.

Discard class 1 — Cisco as a 1999-2000 analogy in an argument about a different company (16 claims). The single most common use of the word "Cisco" in this knowledge base is as the archetype of a networking stock that peaked with a capital cycle. Examples, all discarded:

• "Buying anything over 40x sales, which Nvidia is at, is generally a very bad idea — like Cisco, which grew ~20%/yr 2000-2010 while the market priced 35% growth." (2023-07-21, compound_and_friends, no named speaker)
• "Nvidia is a bigger bubble than Cisco was in 2000." (2025-09-15, Henrik Zeberg, guest on anthony_pompliano_show)
• "Nvidia is bulletproof... it's the Cisco of this but without a 100 PE." (2025-11-16, jordi_visser — and this exact claim appears THREE separate times in the store with near-identical text, which would have triple-counted a single utterance)
• "AI is the 'railroad business' — early capital-cycle infra investors get wrecked (Cisco 2001, railroads 1907)." (2025-12-06, arthur_hayes)
• "Nvidia's extraordinary margins are manufactured by vendor-financing customers (like Cisco in the late 1990s)." (2026-07-08, Mike Green)

None of these is a claim about Cisco Systems in 2026. Every one of them would have been counted as coverage by a naive entity search, and one of them would have been counted three times.

Discard class 2 — a hard name collision (1 claim). 2026-05-01, money_of_mine, speaker David Franklin: "Great development asset; Cisco maiden resource is global-scale, well-located, close to infrastructure; won't stay independent long — a takeover target, possibly Rio." This is a lithium exploration project named Cisco. It has nothing to do with Cisco Systems and is discarded in full.

Discard class 3 — already quarantined by the store (1 claim). The 2026-02-28 record carries thinker: quarantine_misattributed and quarantine_class: SUMMARY_OF_VISSER. The knowledge base itself caught a misattributed summary of a Visser claim and flagged it. We honour that flag and exclude the record. It is also, on its content, another Nvidia-not-Cisco claim.

Discard class 4 — basket and principle claims where Cisco is a set member (3 claims). The 2025-04-23 forward_guidance bearish claim on US tech exceptionalism; the 2025-09-26 compound_and_friends "price matters" principle claim listing Microsoft, Cisco and Sun Microsystems; and the 2024-07-03 real_vision claim that Nvidia echoes late-90s Cisco and Dell. All list Cisco as an entity; none assigns it a view.

The 8 claims that survive

> 2023-07-15 · bullish · conviction 45 · horizon: thesis · entities: CSCO, SBUX, KO · speaker: Ray Dalio · attribution: host · skill 0.7

> "Recommends holding Cisco, Starbucks and Coca-Cola as reliable long-term holdings, citing dividends and low beta."

> Weight: low. Three years old, conviction 45, and a basket-of-three defensives rather than a Cisco argument. Retained because it names the security and assigns a stance, not because it carries information.

> 2025-03-02 · bullish · conviction 58 · horizon: thesis · entities: IBM, CSCO · thinker: jordi_visser · skill 2.0

> "IBM and Cisco benefit from AI/datacenter buildout, trade near ~16 PE (half the Mag 7); should re-rate to a market multiple as Mag 7 multiples compress."

> Test: "IBM/Cisco re-rate toward a market multiple"

> This is the call, and it worked. It is dated, falsifiable, carried its own test, and the test has passed: Cisco went from approximately 16x to 25.4x forward. The highest-skill speaker in this lane (2.0) made a specific re-rating call seventeen months ago and it has been delivered in full.

> 2025-12-20 · bullish · conviction 75 · horizon: thesis · entities: CSCO · speaker: jordi_visser · skill 2.0

> "Cisco is the traditional-finance AI and stablecoin/tokenization play — enterprise on-prem AI bots need routers, traffic, cyber; also benefits from data-center and stablecoin volume."

> 2025-12-21 · bullish · conviction 85 · horizon: thesis · entities: Cisco, CSCO · thinker: jordi_visser · skill 2.0

> "Infrastructure layer benefits as AI shifts from cloud training to real-time inference across physical systems; not yet priced in."

> Highest conviction in the lane (85), and the "not yet priced in" clause is now eight months old with the stock up materially since. We treat the priced-in half of that claim as superseded by price action and say so.

> 2026-03-25 · bullish · conviction 70 · horizon: thesis · entities: Cisco · thinker: cognitive_revolution · skill 1.0

> "Cisco's distributed-systems DNA (scale-out networking, security, observability) positions it to enable horizontal scaling of AI, not just vertical model-building."

> The most valuable claim in the lane after Visser's, because it is from an INDEPENDENT source making a structurally similar argument. Two independent voices converging is worth more than one voice repeating.

> 2026-03-25 · bullish · conviction 78 · horizon: thesis · thinker: cognitive_revolution · skill 1.0

> "Multi-agent systems will agentify enterprise operations; Cisco's CAPE (20 agents) cut SRE load 30%, automated 40% of tasks, response hours-to-instant."

> A specific, checkable product fact rather than a market view.

> 2026-07-26 · bullish · conviction 48 · horizon: thesis · entities: UPS, CSCO, JBL, CNC, WMT · thinker: jordi_visser · skill 2.0

> "Low-margin, labor-heavy firms become big AI winners — a 5% expense cut yields ~20-25% EPS jump; agents deployed as infrastructure unlock margin benefit not yet priced."

> A five-name basket claim, conviction 48, and a poor fit for Cisco specifically — Cisco is a 64.9%-gross-margin company, not a low-margin labour-heavy one. Retained but weighted near zero. This is the freshest claim in the lane, nine days before the dive, and it is the weakest.

Half-weighted commercial-partner voice — 2 claims, quoted separately, excluded from the conviction pool:

> 2026-02-04 and 2026-02-08 · bullish · conviction 80 and 82 · thinker: jensen_huang

> "NVIDIA's Vera Rubin stack ships with Cisco integrating NVIDIA AI networking into the Nexus control plane for enterprise manageability and security."

> "The full AI stack — compute (Vera Rubin), networking, security — must be reinvented; Cisco integrates NVIDIA AI networking into Nexus for enterprise control."

> These are management voice from a commercial partner describing a joint product. Under house rules management framing is half-weighted; a partner's chief executive promoting a joint stack is at least that conflicted. The FACT — that Cisco is integrated into NVIDIA's enterprise reference architecture — is corroborated by the 10-Q's AI Infrastructure revenue attribution and is genuinely useful. The BULLISHNESS is discounted to zero. A third claim from the same source (2025-03-18, AI-RAN with T-Mobile) is treated identically.

The concentration problem, stated plainly

Jordi Visser supplies 4 of the 8 usable claims — 50% of the lane. He is the highest-skill speaker in it (2.0 against 1.0 for everyone else), his central call was specific, dated, tested and correct, and there is no evidence in the record of him reversing position on this name. But 50% of a lane from one speaker is not consensus, and this dive does not present it as consensus. The only genuinely independent corroboration is the two cognitive_revolution claims of 2026-03-25.

Note also that Visser's voice appears under two different thinker labels in the raw hits — jordi_visser on most, and one record under quarantine_misattributed with quarantine_class: SUMMARY_OF_VISSER. Collapsing by speaker rather than by channel is what prevented a single view from being counted as two.

Net conviction: positive-moderate. Breadth: 6 distinct speakers or channels after collapsing. The lane says buy Cisco at 16x. It does not say buy Cisco at 25x, and no claim in it postdates the re-rating with a fresh target.

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

Eleven findings. Cisco's file is mixed: the income statement ties to the filing with unusual precision, and the balance-sheet and geographic blocks contain the two most serious errors in this dive.

1. seg_geo labels the AMERICAS segment as "UNITED STATES" — REJECTED, corrected from the 10-K. The vendor reports $33.656B under "UNITED STATES" for FY2025. The 10-K states: "Revenue in the United States was $30.4 billion, $28.7 billion, and $29.9 billion for fiscal 2025, 2024, and 2023, respectively." The vendor reports $33.656B; the filing shows $30.4B; the vendor's line is the Americas segment, not the country. We use the filing. The error is $3.256B, or 10.9%, and it would materially overstate US concentration for any reader building a geographic exposure view. Corroborated by the 10-Q, which decomposes Americas growth across the United States, Canada, Mexico and Brazil.

2. The FY2018 seg_geo row is broken — DISCARDED. It contains one line, "UNITED STATES: $25.500B", summing to $25.5B against FY2018 revenue of approximately $49.3B — 51.7% of the total. Not used.

3. bal_a is three quarters stale — the largest financing change of the period is absent. The vendor's newest balance sheet is 2025-07-26; the most recent filing-verified balance sheet is 2026-04-25. In between, the 10-Q documents commercial paper rising from $3.5B to $8.4B ($4.9B of net issuance) and $1.8B of debt repayment. We rebuild total debt to approximately $31.4B and net debt to approximately $14.7B from the filing, labelling both as derived because the balance-sheet tables are stripped from the extraction.

4. netDebt omits short-term investments — REJECTED and rebuilt. FY2025: the vendor nets $28.093B of debt against $8.346B of cash only, ignoring $7.764B of short-term investments. The vendor reports $19.747B; the balance sheet shows $11.983B; we use $11.983B. The FY2023 case is worse: the vendor reports net debt of −$1.732B when the corrected figure is net CASH of $17.755B, a $16.0B error that would have made the pre-Splunk Cisco look far weaker than it was.

5. enterpriseValueTTM cannot be reconciled to the vendor's own balance sheet — REJECTED. $504.050B implies net debt of $24.220B, matching neither the vendor's $19.747B nor the corrected $11.983B; the residual is 22.7%, above the 15% threshold at which the data contract says a multiple should be withheld. We rebuild EV to approximately $494.5B from filing-derived debt and cash, and state the derivation.

6. Trailing EBITDA is not an operating figure — RECOMPUTED. ebitMarginTTM of 25.58% implies EBIT of $15.540B, which is $1.349B above the sum of the last four quarters' operating income of $14.191B. The gap reconstructs exactly as trailing pre-tax income ($14.075B) plus trailing interest expense ($1.465B). The vendor's EBIT includes non-operating and investment income. We use operating income $14.191B + D&A $2.528B = $16.719B and compute EV/EBITDA at ≈29.6x rather than the vendor's 27.90x.

7. FY2025 GAAP earnings contain a $720M discrete tax benefit — flagged, not capitalised as run-rate. The quarter ended 2024-10-26 shows pre-tax income $2.267B, tax of NEGATIVE $444M, net income $2.711B — net income above pre-tax income. The 10-Q sizes the full effect: "the income tax benefit of $720 million we had in the first nine months of fiscal 2025." FY2025's 8.3% effective tax rate and $10.180B of net income are therefore not clean; roughly $9.46B and $2.37 of diluted EPS is the honest operating figure against the $2.55 reported. Our trailing-twelve-month window excludes the benefit, so the 40.3x trailing multiple used throughout is clean — but every comparison against FY2025 is flattered by it.

8. The FY2024 est row reconciles to neither GAAP nor non-GAAP — DISCARDED. epsAvg 2.29216 and niAvg $9.187B against actual GAAP diluted EPS of $2.541 and actual net income of $10.320B. Not used. The FY2025 row (epsAvg 3.7889) does tie closely to reported non-GAAP, so the defect is isolated.

9. est EBITDA/EBIT rows tested and found NOT corrupt — reported because the contract says to test. No forward year has negative ebitdaAvg; ebitAvg is below ebitdaAvg in every year; neither equals −revenueAvg. This file trips none of the three known corruption signatures. We still run valuation on epsAvg per the contract, and we note that epsAvg passed an independent check: FY2026E $4.276 against the four quarterly non-GAAP figures from earn_cal summing to $4.27, an exact tie.

10. lastDividend: 1.66 is the trailing-twelve-month total, not the last declared dividend — corrected from the filing. The 10-Q states: "On May 13, 2026, our Board of Directors declared a quarterly dividend of $0.42 per common share." Forward run-rate is $1.68 (1.38% yield), not $1.66 (1.364%).

11. quote.yearHigh/yearLow disagree with tech.hi52/lo52 — we use tech. The quote block gives $130.37 and $65.75; the computed technical block gives $130.00 and $66.20. Discrepancies of 0.3% and 0.7%. We use tech throughout, for consistency with the moving averages and relative-return figures computed from the same close series.

Share-count check — passed, with a caveat. Market capitalisation of $479,830,256,117 at $121.74 implies 3,941,353,838 shares, against 3,987 million diluted in the April quarter — a 1.2% gap consistent with the difference between shares outstanding and the diluted weighted average. revenuePerShareTTM of 15.363 against trailing revenue of $60.746B independently implies 3.954B. No share-class or partial-unit error; CSCO is a single class of common stock on NASDAQ and the 10-Q confirms no preferred stock has been issued. We use 3.987B diluted for per-share earnings and the vendor's market cap for enterprise value, and note the 1.2% inconsistency rather than forcing it.

Not defects, correctly reported and independently confirmed: the Q3 FY2026 revenue growth (+12.0% vendor, "12%" filing), gross margin change (−2.0 points both), operating margin change (+2.4 points both), FY2025 senior notes and commercial paper ($24.6B carrying + $3.5B = $28.1B against the vendor's $28.093B), employee count (86,200 both), no-customer-above-10% concentration, seg_prod summing exactly to reported revenue in FY2024 and FY2025, and interestCoverageRatioTTM of 9.69x consistent with the stated covenant compliance at a 3.0x floor.

Non-equity tripwire — checked and passed. CSCO is common stock, NASDAQ Global Select, CUSIP 17275R102, ISIN US17275R1023. Price of $121.74 is not par-like; beta is 1.007; the dividend is variable and has been raised; the 52-week band of $66.20 to $130.00 is a 96% range; volume was 22.27M shares. Common equity.

Extraction limits, stated rather than estimated around. Numeric tables are stripped from the current filing text, so the dollar values of inventory, purchase commitments, remaining performance obligations, deferred revenue and the April-quarter balance sheet are not available — only the percentage changes and prose the MD&A states. Earnings-call transcripts are not available on this data plan, so management Q&A is absent. The 8-K of 2026-05-13 is present but its Exhibit 99.1 — the press release carrying the actual quarterly financials — was not extracted; what the 8-K does contain, and what we use, is the restructuring announcement.

9. Technicals

• Price $121.74. −6.4% from the 52-week high of $130.00; +83.9% above the 52-week low of $66.20. Position within the annual range: 87th percentile.
• Moving averages, both supportive: +3.4% above a 50-day average of $117.77 and +35.3% above a 200-day average of $90.01. Both trends intact and rising.
• RSI 71.55 — overbought. MACD +0.31 — positive but thin, a fraction of a percent of the share price.
• Maximum drawdown from peak over the trailing year: −6.4% — i.e. the stock has essentially not corrected.
• Relative performance, and this is the whole argument: 3-month +31.4% vs SPY +7.6% and QQQ +7.7%; 6-month +51.0% vs SPY +11.1% and QQQ +15.7%; 12-month +81.4% vs SPY +24.3% and QQQ +30.8%. Cisco has beaten the S&P by 57 points over twelve months. For a $480 billion, forty-year-old networking incumbent with a 1.0 beta, that is an extraordinary result and it is the strongest possible evidence that the re-rating has occurred.
• Sentiment: 38 buy, 34 hold, 1 sell across 73 analysts — a genuine, deeply covered distribution. Consensus target $126.44 (+3.9%), median $130, high $137 (+12.5%), low $110 (−9.6%). The band from low to high is only 22 points wide on a $122 stock. That is a sell side with no disagreement and no conviction.

Today's move and what it does to the entry

CSCO closed 2026-08-04 at $121.74, up 5.08% or $5.88 from a previous close of $115.86. It opened at $119.39 — a 3.05% gap — traded $118.73 to $121.90, and closed within 0.13% of the day's high on 22.27M shares against a 24.7M average. The move was sector-wide: the same session saw large gains across semiconductors and AI hardware. No company-specific news for Cisco is in this file for 2026-08-04; the last company event was the 2026-05-13 earnings release and restructuring announcement, and the next is eight days away.

The honest read. Cisco gapped into the top decile of its own 52-week range, at RSI 71.6, eight days before a fiscal-year-end print. Unlike a name gapping out of a drawdown, there is no compression here to be released — the stock is 6.4% from its high and 35% above its 200-day average. Buying today means paying a 5% one-day move for a base case that sits 1.4% below the closing price, immediately ahead of the single largest information event of the company's year.

What we are telling the reader plainly: hold what you own, do not add ahead of 2026-08-12, and if you want to establish a position, the two setups that would make it defensible are below.

10. Insiders — two transactions, thin, and mildly negative

The file contains only two insider transactions, which is unusually sparse for a company with 86,200 employees and a large executive equity programme, and we note that as a data limitation rather than as evidence of anything.

The reading. The 2026-07-10 transaction is coded F-InKind — shares withheld by the issuer to satisfy tax on vesting restricted stock units. That is not a sale and carries no signal. The 2026-06-16 transaction is coded S-Sale — a genuine open-market disposal of 5,219 shares, approximately $625,000, by the EVP of Operations, who retained 142,765 shares worth $17.4M at today's price. A 3.5% trim of a large retained position is a routine liquidity event, not a statement.

What the file contains is therefore: zero open-market purchases and one small open-market sale. That is mildly negative and no more than that. What would have been informative — insider buying into the AI-infrastructure inflection — is absent, and its absence in a stock up 81% is unsurprising rather than telling.

11. Verdict, kill-criteria and flip conditions

Hold.

This is a name where our research process worked and the correct response to that is discipline, not a victory lap.

What is genuinely good, and none of it is in dispute: trailing revenue of $60.746B, a record; April-quarter revenue up 12.0% with product up 17% and Networking up 25%, or $1.7B, on AI Infrastructure; operating margin up 2.4 points to 25.0%, every one of those figures confirmed verbatim in the 10-Q; a 64.3% trailing gross margin; free cash flow of $13.288B on capital expenditure of 1.6% of revenue; return on invested capital of 12.3%; interest coverage of 9.69x against a 3.0x covenant with a $5.0B revolver undrawn; no customer above 10% of revenue; and a 2.6% shareholder yield with $9.6B of buyback authority remaining.

What we are declining to pay for: 40.3x trailing GAAP earnings and 25.4x FY2027 consensus on a business whose consensus revenue growth decays to 0.4% by FY2029; growth confined to one of five product categories while services revenue falls 1%, subscription revenue falls 2%, remaining performance obligations are flat and deferred revenue falls 1%; a 93% combined increase in inventory and firm noncancelable purchase commitments entered into to secure memory at cycle-high prices, with the filing itself warning of "potential material excess and obsolete or other charges"; product gross margin down 2.5 points, compressed by the growth itself; commercial paper up from $3.5B to $8.4B funding a capital return that exceeded free cash flow over the last nine months; and a base fair value of $120 that is 1.4% below spot and 5.1% below the street.

The distinction that matters. We are not saying Cisco is a bad business or that the AI-networking cycle is fake. We are saying that the specific, dated, testable claim our own knowledge base made about this security — that it would re-rate from 16x toward a market multiple — has been delivered in full, and that no claim in the lane postdates the re-rating with a fresh case. In the Synthos frame, a realised thesis with a fair value below spot and negative skew is a Hold by construction. Selling a compounding franchise 6% off its high because the multiple caught up is not obviously right either; holding, collecting 2.6%, and waiting for either a better price or better evidence is.

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

• FY2027 revenue guidance on 2026-08-12 implying below roughly $66B, against the $68.755B consensus. That is the cycle ending early and the multiple has nothing underneath it.
• A second consecutive quarter of DECLINING services revenue with remaining performance obligations still flat or down. A hardware company with a shrinking annuity is a different and cheaper security.
• A material excess-and-obsolete or purchase-commitment charge against the inventory and commitment book that grew 93% in nine months. The 10-Q names this risk in terms.
• Product gross margin down more than 2.5 points year on year again, which would make the memory and mix pressure structural rather than cyclical.
• Networking growth falling to single digits in any quarter, which is the whole thesis reversing.
• Restructuring charges materially above the announced $1 billion, or a second restructuring, which would say the cost base is not being fixed.

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

• Price below approximately $100 — roughly 21x FY2027E — with no deterioration in the operating picture. At $100 the base case is +20% and the skew inverts. That is 18% below today's close and inside the range this stock traversed in the last six months.
• The 2026-08-12 print delivering revenue above $16.819B TOGETHER WITH FY2027 revenue guidance above roughly $70B. Both legs, not one. That would falsify the consensus decay path in 6a and justify paying today's multiple.
• Services revenue returning to growth with total RPO up year on year, which is the single cleanest evidence that this is a platform transition rather than a product cycle — and it is the one number that would change our growth score from 6 to 8.
• Product gross margin flat or up year on year, killing the memory-cost and mix concern in one print.
• A fresh name-level knowledge-base claim postdating the re-rating — someone arguing Cisco is cheap at 25x, not at 16x. At present no such claim exists, and its absence is worth roughly one tier.

Where CSCO fits in the Synthos Framework Portfolio. The AI-infrastructure sleeve, incumbent expression, at a maintain-only weight with no addition ahead of 2026-08-12. On batch overlap: Cisco is the lowest-beta, highest-yield, most-diversified way to own the AI networking buildout, and it is the one where the least of the return is still ahead. A reader already holding semiconductor capital equipment or AI silicon in this batch should understand that Cisco is a different point on the same curve, not a diversifier — the shared dependency is hyperscaler capital expenditure, and it is the same dependency. Logged as a tracked Synthos call (Hold) as of 2026-08-04 at $121.74.

Single biggest risk: that the AI-networking cycle is a pull-forward and the balance sheet is already committed to it. Cisco has taken firm, noncancelable, multi-year, fixed-dollar positions in memory and contract-manufacturing capacity to serve hyperscaler demand, growing inventory 49% and purchase commitments 111% in nine months. If that demand normalises on the schedule consensus itself models — 11.0% revenue growth this year decaying to 0.4% by FY2029 — Cisco will be holding the inventory when it does. At 40.3x trailing earnings there is no valuation cushion to absorb the charge, and the stock is 6.4% from its high.

Most fragile assumption in the price: that product gross margin stabilises. It is the only assumption in 6a that gets WORSE if the volume assumptions get BETTER, because the 10-Q attributes the mix component of the 2.5-point decline directly to "higher Networking revenue" — the growth category is the low-margin category. Add the memory cost line the filing names explicitly, and the company is in the position of needing its best product line to sell less in order to protect its blended margin. Operating leverage has covered this so far — operating margin rose 2.4 points while gross margin fell 2.0 — but that is a one-time benefit from expense discipline and a $1 billion restructuring, and it cannot be repeated indefinitely.

Provenance & disclosures

• Traceability: 29 raw knowledge-base hits on CSCO, Cisco, Cisco Systems across 51,928 distilled claims (13 entity matches, 16 text matches). 8 used as name-level conviction; 21 discarded. Discards by reason: 16 use Cisco as a 1999-2000 ANALOGY inside an argument about Nvidia or Apple and assign no view to Cisco (speakers include Henrik Zeberg, Mike Green, Gavin Baker, Jim Bianco, Arthur Hayes and Jordi Visser — one Visser claim of 2025-11-16 appears THREE times in near-identical form and would have triple-counted a single utterance); 1 is a hard name collision (2026-05-01, money_of_mine, David Franklin — a lithium exploration asset named Cisco); 1 is already flagged by the store itself as quarantine_class: SUMMARY_OF_VISSER; 3 are basket or principle claims listing Cisco as a set member. Of the 8 retained, Jordi Visser supplies 4 — 50% of the lane from one speaker, disclosed rather than presented as consensus; the independent corroboration is 2 claims from cognitive_revolution dated 2026-03-25. A further 3 claims from Jensen Huang are quoted separately at half weight as commercial-partner management voice and are EXCLUDED from the conviction pool, with the underlying fact (Cisco integrated into NVIDIA's Nexus/Vera Rubin enterprise stack) retained and the bullishness discounted to zero. Breadth after collapsing by speaker: 6. All quotes are verbatim from the stored claim text.
• Data as-of: annual fundamentals fiscal 2025 ended 2025-07-26; quarterly fundamentals through 2026-04-25; most recent filing-verified financials 2026-04-25, from the 10-Q filed 2026-05-19 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873600 = 2026-08-04T20:00:00Z ($121.74, +5.08%; 50-DMA $117.77; 200-DMA $90.01; RSI 71.6; MACD +0.31) · knowledge-base claims 2026-08-04. Fiscal note: Cisco's fiscal year ends the last Saturday of July. "FY2025" means the twelve months to 2025-07-26; "Q3 FY2026" means the quarter to 2026-04-25. All figures come from the Synthos vendor data file for CSCO or from the SEC filings in the CSCO archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2025-09-03 (fiscal 2025); 10-Q filed 2026-02-17 (January 2026 quarter); 10-Q filed 2026-05-19 (April 2026 quarter); 8-K filed 2026-04-06; 8-K filed 2026-05-01; 8-K filed 2026-05-13 (third-quarter results and the restructuring announcement). The 2026-05-13 8-K's Exhibit 99.1, carrying the actual quarterly financials, was not extracted — what the document itself contains, and what this dive uses from it, is the non-GAAP policy discussion and the restructuring disclosure of "up to $1 billion" with "approximately $450 million... in the fourth quarter of fiscal 2026." Numeric tables are stripped from all extracted filings in this archive, so dollar values for inventory, purchase commitments, RPO, deferred revenue and the April-quarter balance sheet are unavailable; only percentage changes and prose figures the MD&A states in text are used, and every derived figure is labelled as derived.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): the seg_geo line labelled "UNITED STATES" ($33.656B) is the AMERICAS segment, against the 10-K's explicit "Revenue in the United States was $30.4 billion" — a $3.256B, 10.9% mislabel, corrected to the filing; bal_a is three quarters stale and misses $4.9B of net commercial-paper issuance documented in the 10-Q ($8.4B outstanding at 2026-04-25 against $3.5B at 2025-07-26) plus $1.8B of debt repayment, with total debt rebuilt to ≈$31.4B and net debt to ≈$14.7B; netDebt omits $7.764B of short-term investments, restating FY2025 from $19.747B to $11.983B and FY2023 from net debt of $1.732B to net CASH of $17.755B; enterpriseValueTTM of $504.050B implies net debt of $24.220B that reconciles to neither vendor figure (22.7% residual) and is rejected and rebuilt to ≈$494.5B; trailing EBITDA recomputed from $18.068B to $16.719B because the vendor's "EBIT" is pre-tax income plus interest and includes non-operating income; the FY2025 income statement is shown to contain a $720 million discrete tax benefit the 10-Q names, making the 8.3% FY2025 effective tax rate and $10.180B net income not clean; the FY2024 est row is discarded as reconciling to neither GAAP nor non-GAAP actuals; the FY2018 seg_geo row is discarded as broken; lastDividend of $1.66 is the TTM total against a $1.68 forward run-rate confirmed by the 10-Q's $0.42 quarterly declaration; and quote.yearHigh/yearLow ($130.37/$65.75) are rejected in favour of tech.hi52/lo52 ($130.00/$66.20). Where vendor and filing AGREED — worth recording: Q3 FY2026 revenue growth (+12.0% vendor, "12%" filing), gross-margin change (−2.0 points both), operating-margin change (+2.4 points both), FY2025 total debt ($28.093B vendor against $24.6B carrying senior notes + $3.5B commercial paper = $28.1B filing), employee count (86,200 both), the absence of any 10%-plus customer, seg_prod summing exactly to reported revenue in FY2024 and FY2025, and the FY2026 epsAvg of $4.276 tying exactly to the four quarterly non-GAAP figures in earn_cal.
• Basis note: analyst estimates are non-GAAP; reported diluted EPS is GAAP, and for Cisco the gap is large — FY2025 GAAP diluted EPS was approximately $2.55 against a non-GAAP consensus of $3.79, a 49% difference driven mainly by $3.641B of share-based compensation and acquisition-intangible amortisation. Forward multiples in this dive use consensus non-GAAP EPS and the trailing multiple is shown on GAAP (40.3x), and the two are NOT comparable to each other. This is stated because for Cisco, unlike for some names in this batch, the choice of basis changes the answer materially.
• Estimate coverage: 18 analysts on FY2026 through FY2029 revenue; 14 on FY2026 and FY2027 EPS, but only 8 on FY2028 and 5 on FY2029. The FY2028 and FY2029 EPS rows are used as cross-checks only, never as anchors. The FY2024 row is discarded outright.
• Peer note: the vendor peer set contains one genuine comparable — Arista Networks ($239.9B) — plus IBM ($221.5B) as the relevant cheap-incumbent-re-rating analogue, alongside names that are not comparables (Applied Materials, Lam Research, Micron, Salesforce, SAP, Uber, Motorola Solutions, Qualcomm). No peer multiples are supplied in the file, so no peer-multiple comparison is drawn.
• Fair-value caveat: the $88 / $120 / $140 anchors are multiples of the FY2027 consensus non-GAAP EPS distribution — 19x the low of $4.636, 25x the mean of $4.787, and 28x the high of $5.007. Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 23x gives $110 (the street's target low), 27x gives $129. The whole answer lives in a 23–27x band. Cross-checked against FY2028E: 22.8x $5.272 gives $120.2, agreeing with the base to 0.2%.
• Timing: fiscal fourth-quarter and full-year results are due 2026-08-12, EIGHT DAYS after this dive, carrying FY2027 guidance and approximately $450M of restructuring charge. The verdict is set with that timing as an explicit input. 2026-08-04 was a broad technology up-day; CSCO's +5.08% carried no company-specific news in this file.
• 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.