CISCO SYSTEMS, INC. (CSCO) stock intrinsic value fair value estimate & owner earnings analysis

Cisco sells the switches, routers, wireless gear and firewalls that run most corporate, government and telecom networks, and wraps an increasingly large software and subscription business around that installed base: security, the Splunk data and observability platform bought in 2024, Webex collaboration, and multi-year support contracts. Fiscal 2026 (ended July 25, 2026) revenue was $63.3 billion, up 12%, with product revenue up 16% and Networking up 22% on AI data-centre orders and campus refresh. Sales of AI infrastructure to the largest cloud operators reached about 6% of revenue, from under 2% the year before. Services revenue was flat, and despite headline software revenue of $23.2 billion, subscription revenue grew only 1% once Splunk was in the comparison base. Contracted but unrecognised revenue stands at $46.7 billion, up 7%. How we measure the earnings. We start from operating cash flow and remove three things: swings in working capital, the cost of the stock Cisco pays its employees (now $3.8 billion a year, roughly 6% of revenue and more than double its level a decade ago), and the capital spending needed to keep the business running. Capital spending is small for Cisco, around 1.7% of revenue. On that basis owner earnings were remarkably steady from fiscal 2019 to fiscal 2025, between $10.6 billion and $12.6 billion a year, and reached $13.4 billion in fiscal 2026. Removing working capital matters here: fiscal 2023 looked like a record and fiscal 2024 like a collapse in raw cash flow, but almost all of that was tax payments and deferred-revenue timing, and the underlying earnings in both years were about $11.5 billion. Fiscal 2018 is left out because the one-time US tax-reform charge distorts it. Large amortisation of acquired intangibles, mostly from Splunk, is added back because it is a sunk purchase cost. The fact that acquired technology wears out is instead reflected in the cases below, where earnings fade over time and Cisco has to keep spending on acquisitions. To value the business before its lenders, we add back interest paid (about $1.4 billion in fiscal 2026) and take out the interest Cisco earns on its cash, because we count that cash directly in the balance sheet. On that basis the business has generated a median of about $11.7 billion a year over fiscal 2019 to fiscal 2026, and $13.9 billion in fiscal 2026. Over the same years Cisco has never grown this figure much on its own. Total owner earnings in fiscal 2025 were no higher than in fiscal 2019, and per-share gains came almost entirely from buying back stock. We have not projected the buybacks forward. We also deduct an ongoing allowance for the small acquisitions Cisco makes almost every year (an average of about $1.5 billion a year over the last decade, excluding Acacia and Splunk), because those purchases are how the company replaces technology that ages. Base case. Fiscal 2026 business cash flow rose 21% while the long-run median barely moved. Our first year of the base case therefore steps down rather than up, to $12.8 billion, before the acquisition allowance. That reflects the visible contribution of the AI and campus refresh cycle and the $46.7 billion of contracted revenue, offset by Cisco's own warning that cloud AI orders are large, sporadic and concentrated in a few customers, and by the higher memory costs that already trimmed product gross margin in fiscal 2026. It then settles at about $12 billion a year, a little above the long-run median, because Splunk's recurring revenue is now a permanent part of the base. After acquisitions, that leaves about $11 billion a year. We credit no further growth. Cisco's organic growth record is close to zero, AI networking demand is barely two years old, and market growth or share gains do not count as growth that is already secured. Beyond year 10, the stream holds level. It is supported by an installed base that is expensive and disruptive to rip out, by support contracts renewed year after year for more than two decades, and by continued tuck-in acquisitions paid for in the cash flow. The business is worth about $184 billion. Total debt is $29.5 billion, and cash and investments are $15.9 billion, of which we treat about $3 billion as needed to run the business, leaving net debt of $16.6 billion. That leaves about $168 billion for shareholders, or $42.07 per share. Bear case. Networking equipment has had sharp, real downturns. In the 2001-2002 telecom bust Cisco's revenue fell about 15%, and in the 2009 recession it fell about 9%. In the fiscal 2022 supply squeeze, cash conversion broke down even as reported profit rose. The bear case combines the loss of the hyperscaler AI business (about $3.8 billion of fiscal 2026 revenue concentrated in a handful of buyers) with a 2009-style enterprise spending recession. Business cash flow drops to about $9 billion in years two and three, recovers to about $10.3 billion, and then fades 2% a year. That fade reflects white-box hardware, merchant silicon and lower-priced Asian competitors, which Cisco names as risks, and it continues at 3% a year after year 10. Acquisition spending runs at $1.5 billion a year. The business is worth about $109 billion, less $16.6 billion of net debt, leaving about $92 billion, or $23.12 per share. Interest of about $1.4 billion is covered several times over even at the trough, and $10.2 billion of debt maturing in the coming year can be refinanced comfortably by an investment-grade borrower with this level of cash flow. Bull case. Cisco's identified levers execute: AI networking and the campus refresh hold, Splunk's installed base keeps expanding inside existing Cisco accounts, and cash flow starts at $14 billion and grows 2% a year, roughly Cisco's long-run revenue growth rate, with tuck-in acquisitions at $0.8 billion a year. Beyond year 10 the stream holds level. The business is worth about $254 billion, less $16.6 billion of net debt, or $59.57 per share. What changed since our previous analysis. The previous figures are not comparable because they were built on a different basis. In the business itself, fiscal 2026 brought double-digit growth for the first time in years, a fourfold rise in hyperscaler AI sales, and record contracted revenue. It also brought rising memory costs, a sharp inventory build (inventory rose from $3.2 billion to $5.7 billion) and flat services and subscription revenue, so the news is mixed. Margin-of-safety price. We are fairly sure of the earnings level. Seven of the last eight years fall within about 10% of the median, and the business throws off cash in every year we can see. The worst plausible outcome is not a collapse but a long, slow commoditisation of the core switching and routing franchise while a large acquisition fails to earn its price. Cisco has paid about $46 billion for acquisitions in ten years without lifting total owner earnings. The balance sheet would survive that: net debt is barely more than one year of cash flow, there is no controlling owner, and no customer is dominant outside the cloud AI segment. The things we are most likely to be wrong about are whether fiscal 2026 marks a new, higher level or a cyclical peak, how fast stock-based pay keeps rising, and whether Splunk retains its customers. If fiscal 2026 proves to be a peak and stock pay keeps climbing, the shares are worth something close to the bear case, in the low-to-mid $20s. During the holding period, Cisco pays a dividend of about $1.64 a share, which has been raised every year for over a decade and absorbs roughly half of owner earnings. Measured against $27 that is a cash yield of about 6%. A 20% annual return from $27 requires the price to reach the low $40s within about two and a half years, counting the dividend. That is where the base case already sits, and it does not depend on any growth. If the bear case happens instead, the implied loss from $27 is limited and the dividend is still paid. The margin-of-safety price is $27. Above about $30, the base case alone no longer implies a 20% return in a reasonable time, and the estimate would depend on the AI cycle persisting, which cannot yet be evidenced.