Cisco Systems (CSCO)

Cisco is the enterprise-networking incumbent, with a durable installed base and a successful shift to software + subscription (56% of revenue, product ARR $17.0bn). The second engine is AI infrastructure (Ethernet for clusters, Silicon One), which accelerated growth to +12.5% in the TTM. But it is a mature franchise: the networking core is declining, the margin is compressing on the AI mix, subscription softened, and on the fastest-growing front (AI networking) it competes with Arista/Nvidia/Broadcom where its moat is weak. Durable growth is ~6%.

A mature technology business with material stock-based compensation is valued on EV/NOPAT (owner earnings), which expenses stock-based compensation and handles net debt in the bridge. Adjusted EBIT (adding back Splunk's non-economic intangible amortization, ~$0.94bn) is ~$15.1bn; at a normalized tax rate of 19% (FY25's 8.3% included a one-off tax benefit), NOPAT is ~$12.3bn. With net debt (~$14.7bn) in the bridge, EV/NOPAT is ~37× — an earnings yield of just 2.7%, below the risk-free rate. The 'non-GAAP ~26×' the market watches is misleading: the non-GAAP figure adds back stock-based compensation (~$3.85bn, which jumped with Splunk), a real cost.

The base scenario projects NOPAT growing ~6%/year (AI networking + software, diluted by the mature core + margin compression) toward ~$16bn over five years, at an exit multiple of 16.5× (the center of mature quality software, disciplined by the modest growth + the competitive dispute in AI). That gives ~$68/share → a total return of -7%/year. At ~37× today for ~6% growth, the multiple has to compress far more than the growth offsets.

No margin of safety: the price already discounts a demanding scenario. At ~$111 (near all-time highs, +62% year-to-date) Cisco trades at ~37× EV/NOPAT — a price that does not look cheap once stock-based compensation is expensed. The illusion of cheapness ('non-GAAP ~26×') comes from adding back the stock-based compensation that Splunk pushed up to ~$3.85bn. The verdict is Overvalued: the franchise is high quality (durable installed base, the shift to subscription, dividend aristocrat), but the price discounts a growth rate and durability that the business — with a mature core, a compressing margin, and a weak moat on the AI front — is unlikely to deliver. The strongest signal: not even the bull case (~9% growth + a 20× multiple, the top of the band) reaches today's price. The bear case (AI share loss + a declining core + multiple compression to 13×) takes the value well below. The asymmetry is unfavorable — limited upside, wide downside.

Four signals. Share in AI networking — whether Cisco captures Ethernet-for-AI from hyperscalers (Silicon One, the +111% purchase commitments) or loses it to Arista/Nvidia/Broadcom (the central disconfirmer of the bull case). Gross margin — it compressed −2.0pp on the AI/webscale mix + memory costs; if it keeps falling, NOPAT grows slower than revenue. Subscription and the backlog — subscription softened −2% and RPO has been flat for three quarters; if contracted revenue does not reaccelerate, durable growth is below 6%. And buyback discipline — Cisco buys back ~$6bn/year at ~37×, an expensive price that destroys value; if it keeps buying near highs instead of building cash, shareholder returns suffer. The base case already assumes the multiple compresses; if the AI re-rating holds for longer, the price can stay expensive longer than the model projects — but intrinsic value does not justify it.