This is investment research, not personal financial advice.
A 9.8% two-day fall separated the record from the economics
Cisco Systems (NASDAQ:CSCO) fell 8.40% to $113.47 on 13 August after reporting record FY2026 revenue of $63.3 billion and $9.3 billion of AI infrastructure orders from hyperscalers. The shares lost another 1.58% to $111.68 on 14 August, taking the two-session decline to 9.85%. That response was not a rejection of demand. It was a repricing of what the demand costs to serve.
Cisco recognised about $4 billion of AI infrastructure revenue during the year, yet operating cash flow was $14.18 billion, almost exactly the $14.19 billion produced in FY2025. Inventory rose from $3.16 billion to $5.69 billion and absorbed $2.54 billion of cash. Annual GAAP product gross margin slipped from 63.7% to 63.2%. The record came with more hardware in the mix, more components on the balance sheet and no increase in cash generated from operations (Cisco 2026).
The direction of the decline fits that evidence. Cisco proved that hyperscalers want its networking systems. It did not yet prove that the resulting growth can preserve margin and release cash at the rate embedded in a roughly $441 billion equity value. Even after the two-day fall, the owner-cash valuation still assumes a long and unusually fast conversion. The market reaction was directionally correct but did not remove that burden (Yahoo 2026; CompaniesMarketCap 2026).
This is a different owner question from whether artificial-intelligence infrastructure spending is genuine. The orders are filed facts. The question is whether Silicon One systems, optics, memory and customer financing convert those orders into durable owner cash before supply commitments and competition absorb the economics.
The AI tally is $9.3 billion ordered and about $4 billion delivered
Cisco still makes money through two related engines. Product revenue comes from Networking, Security, Collaboration and Observability. Services include support and other contracted work around that installed base. The reporting segments are geographic, so the product categories are revenue groupings rather than self-contained divisions with fully allocated operating profit.
Networking supplied the event. FY2026 Networking revenue rose 22% to $34.67 billion. It accounted for about 55% of group revenue and roughly 95% of Cisco's $6.67 billion year-on-year revenue increase. Security rose 2% to $8.23 billion, Collaboration rose 4% to $4.30 billion, Observability rose 4% to $1.10 billion, and services were flat at $15.03 billion. In Q4, product orders increased 35%; even excluding hyperscalers they rose 25%. Networking orders increased 40%, while Q4 Networking revenue rose 28% (Cisco 2026).
Those numbers show demand breadth beyond a single customer group, but the hyperscaler contribution changes the economics. Cisco recorded $4.0 billion of AI infrastructure orders in Q4 and $9.3 billion for the full year. Revenue recognised from those systems was about $4 billion. Orders, revenue and cash are three separate ledgers. Orders can be cancelled or rescheduled under contract terms; revenue waits for control to pass; cash depends on collections, inventory turns and any financing Cisco extends.
The installed base remains a barrier for competitors. Customers can combine switches, routing, security, observability, support and financing under one vendor relationship. Remaining performance obligations, or RPO, reached $46.73 billion, up 7%. Product RPO was $23.44 billion and services RPO was $23.30 billion. Deferred revenue was $29.78 billion (Cisco 2026).
But RPO is not annual recurring revenue. The Q3 filing showed that RPO included both deferred revenue and $14.86 billion of unbilled contract revenue, with only about half expected to be recognised over the following 12 months. Cisco's subscription-revenue definition also includes term software licences, some recognised upfront. Calling the whole balance software-like recurring revenue would overstate its quality (Cisco Q3 2026).
Splunk has not displaced the networking cycle
The $27.1 billion Splunk acquisition was meant to increase the weight of security, observability and contracted software. The purchase accounting shows the size of that wager. Cisco recorded $19.3 billion of goodwill and $10.55 billion of purchased intangibles. Splunk contributed about $1.4 billion of revenue and a $557 million net loss during the post-close portion of FY2024. On a pro-forma basis, the combined businesses would have generated $56.76 billion of FY2024 revenue and $9.28 billion of net income (Cisco 2025).
The product logic is credible. Cisco can combine network telemetry, Splunk Enterprise Security, XDR, ThousandEyes and application observability. A large enterprise buyer gains one set of data and control surfaces across networks, applications and threats. That integration adds switching costs when it works.
The financial evidence is less settled. Subscription revenue grew from $22.39 billion in FY2022 to $31.53 billion in FY2025, but it fell 2% in Q3 FY2026 and was almost flat over the first nine months. Cisco attributed some Splunk weakness to customers moving from fewer on-premise transactions toward cloud subscriptions. Security revenue fell 2% over those nine months before a Q4 rebound left the full year up only 2%. Observability finished 4% higher (Cisco Q3 2026; Cisco 2025).
That is the anti-thesis to the software transition. The FY2026 acceleration came from physical networking systems while the acquired software categories grew slowly. The recurring base still supports the franchise, but it did not drive this year's incremental economics. Splunk cross-sell remains a future operating proof rather than a conclusion available from the reported numbers.
Arista Networks supplies a useful peer check. Its Q2 release showed another strong period for cloud and AI networking, confirming that Cisco is participating in a wider spending cycle rather than creating the cycle alone (Arista 2026). Cisco's advantage is breadth across enterprise and hyperscaler systems. Arista's presence limits the case for treating every AI order as franchise rent.
Product mix took margin while supply claims took cash
Cisco's Q3 margin bridge makes the cost of the growth visible. Productivity added 1.2 percentage points to product gross margin, but product mix removed 4.4 points. Pricing removed another 0.1 point. Higher Networking revenue was the main adverse mix item, and higher memory cost diluted the productivity benefit. Q3 GAAP product gross margin fell to 61.9% from 64.4% (Cisco Q3 2026).
Q4 looked better under GAAP and worse on the cleaner comparison. GAAP product gross margin rose to 62.6% from 61.5%, partly because the prior-year quarter carried a $355 million supplier-settlement charge. On Cisco's non-GAAP presentation, Q4 product margin fell to 64.8% from 67.5%. For the full year, GAAP product margin declined to 63.2% from 63.7%, while total gross margin slipped about 41 basis points to 64.5% (Cisco 2026).
The balance sheet moved faster. Inventory increased 80% to $5.69 billion. At Q3, firm purchase commitments had more than doubled from $7.60 billion at FY2025 year-end to $16.03 billion. Of that amount, $14.15 billion was due within one year. Cisco linked the commitments to Silicon One, hyperscaler demand and memory availability. Many were firm, non-cancellable and unconditional, although the booked loss reserve for quantities judged above expected demand was only $209 million (Cisco Q3 2026).
The difference between a $16.03 billion gross commitment and a $209 million accounting reserve matters. The reserve reflects management's demand forecast. The contractual exposure reflects what Cisco has promised suppliers. If demand arrives as expected, the commitments secure capacity. If orders shift, component prices fall or a tailored system becomes harder to reuse, the same commitments can become write-downs or charges.
This is not Cisco's first inventory cycle. Revenue fell 5.6% in FY2024 after customers absorbed equipment shipped during the earlier shortage. The FY2024 annual report recorded $53.80 billion of revenue, down from $57.00 billion in FY2023, and operating cash fell almost 45% to $10.88 billion (Cisco 2024; Cisco 2023). The current cycle is different in product and customer, but the prior cycle shows why orders, shipment timing and customer digestion cannot be treated as a straight line.
Revenue rose 12%; operating cash did not move
Five years of reported figures put the event in context. All amounts below are USD millions. Gross margin is reported GAAP gross profit divided by revenue. ROIC and interest cover are author-computed rather than Cisco-reported metrics.
| Fiscal year | Revenue | Net income | Gross margin | Operating cash | Capex | Computed ROIC | Computed interest cover |
|---|---|---|---|---|---|---|---|
| 2022 | 51,557 | 11,812 | 62.5% | 13,226 | 477 | 39.5% | 38.8x |
| 2023 | 56,998 | 12,613 | 62.7% | 19,886 | 849 | 44.2% | 35.2x |
| 2024 | 53,803 | 10,320 | 64.7% | 10,880 | 670 | 27.6% | 12.1x |
| 2025 | 56,654 | 10,180 | 64.9% | 14,193 | 905 | 20.9% | 7.4x |
| 2026 | 63,325 | 13,267 | 64.5% | 14,177 | 1,410 | 21.4% | 10.5x |
The annual filings supply the reported history (Cisco 2022; Cisco 2023; Cisco 2024; Cisco 2025; Cisco 2026). The ROIC calculation uses GAAP operating profit after the reported effective cash-tax proxy, divided by average invested capital. Invested capital is debt plus equity less cash and short-term investments. This treatment lifts the early-year returns because Cisco then held net cash. Splunk increased goodwill, debt and purchased intangibles, pushing average invested capital higher from FY2024 onward.
The result is not a claim that Cisco's underlying franchise suddenly halved in quality. It shows that acquisition capital entered the denominator before acquired software produced a comparable increase in group profit. FY2026 computed ROIC of about 21% remains above a normal cost of capital, but it is far below the 40%-plus figures generated by the pre-Splunk, net-cash balance sheet. A simple incremental-ROIC ratio is omitted because the FY2024-FY2026 change in invested capital is dominated by Splunk purchase accounting and acquisition timing; presenting that quotient as organic reinvestment economics would mislead.
Owner cash produces a second bridge. Reported operating cash minus all capital expenditure was $12.77 billion in FY2026:
| FY2026 owner-cash bridge | USD bn |
|---|---|
| GAAP net income | 13.27 |
| Net non-cash and working-capital adjustments | 0.91 |
| Operating cash flow | 14.18 |
| Less capital expenditure | (1.41) |
| Reported owner-cash proxy | 12.77 |
| Less share-based compensation added back in cash flow | (3.83) |
| SBC-charged owner-cash proxy | 8.94 |
The second line is a shareholder-cost view, not a second GAAP measure. Net income already expenses share-based compensation, while the cash-flow statement adds it back. Subtracting the $3.83 billion again from operating cash restores that economic cost. A model using this cash deduction must not also assume the same dilution, or it counts the cost twice.
Cisco spent $6.11 billion on programme buybacks and $6.55 billion on dividends during FY2026. Another $1.87 billion of shares were withheld for tax on vested awards. Yet diluted weighted shares fell only 0.28%. Q4 programme repurchases averaged $111.53 a share, close to the $111.68 post-results close. The cash labelled as repurchase did much more work offsetting employee equity than shrinking the denominator (Cisco 2026). The proxy statement's pay structure confirms that equity awards remain central to executive incentives (Cisco Proxy 2025).
The balance sheet can fund the cycle, but funding is not the crux
At year-end Cisco held $7.22 billion of cash and $8.70 billion of investments against $29.53 billion of short- and long-term debt. Total equity was $50.29 billion. This is not a near-term solvency story. Computed interest cover improved to about 10.5 times in FY2026 as operating profit recovered, and nearly $29.8 billion of deferred revenue provides a contracted operating base.
The relevant balance-sheet question is allocation. Inventory, financing receivables and supplier commitments use capital before cash collection. Splunk has already moved the company from net cash to net debt. Buybacks and dividends consumed $12.66 billion, equal to 89% of operating cash flow, while programme repurchases accelerated at much higher share prices.
Management's capital record is mixed rather than uniformly poor. The installed base was built through decades of internal development and acquisitions. Splunk gives Cisco an asset that would have been difficult to reproduce quickly. But $19.3 billion of Splunk goodwill means the future cash return must carry a large acquisition premium. Smaller FY2026 deals in observability and non-human identity security continue the portfolio approach.
The FY2025 filing also contains a caution about clean quarterly comparisons. After the August results release, Cisco settled a supplier dispute, forfeited roughly $450 million of prepayments, recovered $563 million from escrow and recorded the $355 million product-cost charge in Q4. The 10-K superseded the earlier release (Cisco 2025). That episode does not repeat in FY2026, but it explains part of the apparent Q4 GAAP margin improvement and shows why the annual filing matters more than a headline margin.
$111.68 requires a faster owner-cash path
A cash-flow valuation fits Cisco better than a simple sales multiple because hardware mix, working capital and share-based pay are central to the event. The model starts with FY2026 operating cash less capex of $12.77 billion. It discounts equity cash directly at the required equity return, so it does not subtract net debt again. The share count is held at 3.95 billion. Acquisitions are not included as capex, which makes this measure generous if portfolio renewal requires continuing deal spend.
Four driver-built cases produce these ranges:
| Case | Operating path | Value per share |
|---|---|---|
| Severe contraction | Owner cash declines or stalls; 10.5%-11.5% equity return | $27-$35 |
| Slow growth | 1%-3% initial growth; 9.75%-10.5% equity return | $37-$47 |
| Execution | 4%-6% initial growth; partial working-capital release; 8.75%-9.5% equity return | $51-$68 |
| AI conversion | 7%-10% initial growth; stable product margin; 8.0%-8.75% equity return | $71-$115 |
The high end of the last range uses the five-year mean owner cash of $13.61 billion rather than the FY2026 figure, assumes double-digit early growth and gives the equity a lower return requirement. The 10-year Treasury yield was 4.63% on 13 August, which makes an 8% equity return a demanding low-discount assumption rather than a neutral default (FRED 2026).
A central calculation using 5% owner-cash growth for five years, 3.5% for the next five, 3% perpetual growth and a 9% equity return gives about $61 a share. Charging share-based compensation as an owner cost reduces that value to about $43. With the reported cash treatment, a sensitivity around the central operating path ranges from about $43 at an 11% equity return and 2% terminal growth to $86 at an 8% return and 4% terminal growth.
The reverse calculation is more revealing. At $111.68, and using a 9% equity return with 3% perpetual growth, the current price requires reported owner cash to grow about 12% annually for ten years. If the growth rate halves after year five, the required path is about 16% for five years and 8% for the next five. Starting from the SBC-charged cash base pushes the required early growth above 20%.
Those figures are mathematical implications, not forecasts. They explain why a 9.85% two-session fall can be economically justified without settling the valuation question. The market is capitalising a long conversion of AI orders into cash, not merely the $4 billion of AI revenue already recognised.
Three disclosures will close the AI ledger
The first is inventory. A decline in inventory as a percentage of revenue, together with purchase commitments below the $14 billion due-within-one-year level disclosed at Q3, would show that secured supply is moving through customers rather than accumulating. Inventory staying above 9% of revenue would point the other way.
The second is product gross margin. Two consecutive quarters below 63% would indicate that hyperscaler mix and component costs are outrunning productivity. Stabilisation above that line, especially without supplier-settlement comparison effects, would support the cash conversion case.
The third is the acquired software base. Security and Observability need combined growth above 5% across successive quarters to show that Splunk and Cisco's native portfolio are adding a second engine. If those categories remain slow while Networking carries most growth, the business will continue to deserve hardware working-capital assumptions in its valuation.
Cisco's $9.3 billion order tally is strong evidence of demand. The $5.69 billion inventory balance and flat operating cash show its current cost. The next four cash-flow statements will reveal whether those numbers converge. Until then, the 9.85% two-session decline reads as a justified recognition of conversion risk, while the remaining valuation still prices much of the conversion as already under way.
Source notes and confidence
The trigger release, latest 10-Q, five annual-result periods, proxy, peer release, market series and macro observation were fetched and read. Reported financials remain in USD. ROIC, interest cover, owner cash, market capitalisation and all valuation ranges are author calculations from the cited inputs. The Finance API sidecar authenticated successfully, resolved Cisco's identity and filing history, but its accepted daily-price series stopped at 7 August 2026; Yahoo's fetched daily series supplied the 13-14 August closes. That lag is why verification is marked partial rather than full. Cisco did not publish a separate FY2026 investor slide deck with the 8-K, so the SEC-filed release and 10-Q provide the operating tables.
The independent results-day report attributed the decline to margin concern and corroborated the timing of the reaction (Motley Fool 2026). The market-cap history source reported about $440 billion around the event; the frontmatter uses $441.0 billion calculated from the 14 August close and 3.949 billion Q4 basic weighted shares, a roughly 0.2% difference caused by source timing and share-count convention.
References
- SEC 2026: SEC ticker registry entry for Cisco Systems, Inc. (CSCO), used for legal identity.
- Cisco 2026: Cisco Q4 and FY2026 results release, filed 12 August 2026.
- Cisco Q3 2026: Cisco Q3 FY2026 Form 10-Q, including RPO, product-margin and purchase-commitment notes.
- Cisco 2025: Cisco FY2025 Form 10-K, including Splunk purchase accounting and the supplier settlement.
- Cisco 2024: Cisco FY2024 Form 10-K, used for the prior inventory cycle and financial history.
- Cisco 2023: Cisco FY2023 Form 10-K, used for financial history.
- Cisco 2022: Cisco FY2022 Form 10-K, used for financial history.
- Cisco Proxy 2025: Cisco 2025 proxy statement, used for management incentives and equity-pay context.
- Yahoo 2026: Yahoo Finance daily CSCO prices through 14 August 2026.
- CompaniesMarketCap 2026: Cisco market-capitalisation history around the event.
- Motley Fool 2026: Independent account of the 13 August results-day market reaction.
- Arista 2026: Arista Networks Q2 2026 earnings release, used for peer AI-networking context.
- FRED 2026: Federal Reserve 10-year Treasury constant-maturity yield, 13 August 2026.