Cisco Grew 18%, and Five Points of It Came From Price Hikes

The core results:

You only see the shape of the acceleration when you string the year's quarters together. The first nine months of FY2026 produced $46.07B of revenue, up 9.7%, with the third fiscal quarter alone at $15.84B, up 12.0%. Then the closing quarter jumped straight to $17.3B, up 18%. The acceleration is real, and almost all of it is packed into that final quarter. On a full-year basis: revenue $63.3B up 12%, GAAP net income $13.3B up 30%, GAAP EPS $3.33 up 31%, adjusted EPS $4.33 up 14% (Cisco 8-K, 2026-08-12). GAAP profit growth running far ahead of adjusted profit growth is a base-effect story: the prior year carried more one-time acquisition and restructuring items, which depressed the GAAP comparison base. So that 51% net income growth should not be read as a leap in underlying earning power.

How much of the 18% is genuine demand? CFO Mark Patterson said the quiet part out loud on the call: pricing contributed 5 percentage points to fourth-quarter revenue growth, and will contribute another 4 to 5 points in FY2027 (Cisco Q4 FY2026 earnings call, 2026-08-12). Strip pricing out of the 18% and unit-shipment growth is running somewhere around 13%. Still very good. Just a step down from what the headline "18%" makes you feel.

The arithmetic behind the full-year guide is worth walking through. The FY2027 revenue midpoint of $72.8B is $9.5B more than FY2026's $63.3B. The first-quarter midpoint of $18.1B is effectively locked, which leaves $54.7B to be produced across the remaining three quarters, or an average of $18.23B each. The benchmark for that is the record $17.3B just delivered, meaning the next three consecutive quarters each have to come in 5.4% above a record quarter, with no misses at all. Of that $9.5B increment, management's own AI infrastructure revenue path, from roughly $4.0 billion in FY2026 to $7.5 billion in FY2027, supplies $3.5B; pricing at 4 to 5 points supplies $2.5B to $3.2B. Together those two eat two-thirds of the increment, leaving a bit over $3B for genuine volume growth in traditional enterprise networking, which works out to about 5% growth. The single number to watch in the next quarterly report is non-GAAP gross margin. The company put first-quarter guidance at 65% to 66% (Cisco Q4 earnings call, 2026-08-12), another notch below the 66.3% just reported. Revenue has a backlog underneath it and is unlikely to miss badly; whether the $5.08 EPS midpoint gets delivered comes down to whether price increases keep pace with costs.

The evidence on the demand side is genuinely hard. Fourth-quarter product orders rose 35% year over year, and still rose 25% excluding hyperscale cloud providers, with Networking product orders up 40% (Cisco 8-K and earnings call, 2026-08-12). AI infrastructure orders were $4.0 billion in the quarter and $9.3 billion for the full year, roughly 4.5 times the prior year, split roughly 60% systems built on the in-house Silicon One switching chip and 40% optical interconnect; optical subsidiary Acacia alone booked more than $1 billion of orders in the quarter. AI orders from neoclouds, sovereign clouds and enterprise customers topped $400 million in the quarter and $1 billion for the year, including one large US bank that placed a single order for 1,000 data center smart switches while ripping out a competitor's switching and firewall gear (Cisco Q4 FY2026 earnings call, 2026-08-12). CEO Chuck Robbins put it this way: "we believe the accelerating adoption of agentic AI is driving a networking super-cycle," and disclosed that the company won scale-across architecture design slots at three different hyperscale cloud providers this quarter (same source).

Start with a pair of numbers that do not sit comfortably together. FY2026 GAAP net income was $13.3B, up 30%, while full-year operating cash flow was $14.2B, printing flat year over year (Cisco 8-K, 2026-08-12). Profit up 30%, cash flow standing still. Where did the difference go?

The answer is in inventory. On the balance sheet, inventory rose from $3.16B at the end of FY2025 to $5.7B at the end of FY2026, roughly 80% in a single year (Cisco 8-K balance sheet, 2026-08-12; prior-year figure from 10-Q/10-K disclosure). Net income also contains close to $4B of non-cash items that should be added back. For a company earning $13.3B, amortizing Splunk acquisition intangibles and booking stock-based compensation, normal operating cash flow ought to run above $17B. It came in at $14.2B instead, and the gap was almost entirely consumed by working capital, above all by stockpiling components.

How should you read that stockpile? In an environment where memory contract prices nearly doubled in a single quarter, locking in material early is a rational hedge, and it lines up with the $9.3 billion of AI orders and $46.7B of remaining performance obligations. There are orders in hand, so the inventory has somewhere to go. But that same inventory is the first line item to get hurt if demand air-pockets, especially when what you stockpiled was memory bought near the highs. This one gets watched quarter by quarter.

How big is the GAAP-to-adjusted gap? FY2026 adjusted EPS of $4.33 against GAAP EPS of $3.33 is a full dollar, roughly $4 billion (Cisco 8-K, 2026-08-12). In the FY2027 guide the gap is $5.08 against $4.03, still $1.05. The company attributes it in its reconciliation to stock-based compensation, acquisition intangible amortization and restructuring; I have not tied out each line item individually to public filings. The direction is clear enough, though. Acquisition intangible amortization is historical bookkeeping left over from the Splunk deal, and it neither consumes cash nor dilutes shareholders. Stock-based compensation is real dilution, and the slice it takes from existing shareholders every year is precisely what buybacks have to offset.

The capital return line is run cleanly. FY2026 returned $12.7B to shareholders, split $6.1B of buybacks and $6.6B of dividends, together equal to 99% of free cash flow. The dividend has now been raised for a 15th consecutive year, with the fourth quarter at $0.42 per share. The fourth quarter repurchased 13 million shares for $1.5 billion, leaving $8.1 billion of authorization (Cisco 8-K and earnings call, 2026-08-12). Free cash flow, taking operating cash flow of $14.2B less full-year capital expenditures of $1.41B, is roughly $12.8B, essentially one-for-one with what was returned.

The cost of that is a balance sheet that has stopped thickening. Cash and investments slipped from $16.1B at the end of FY2025 to $15.9B; short-term debt is $10.2B and long-term debt $19.4B, roughly $29.6B combined, for net debt of about $13.7B. Goodwill of $59.5B is 46% of total assets of $129.6B, against shareholders' equity of $50.3B (Cisco 8-K balance sheet, 2026-08-12). Goodwill larger than equity means that if the cash flows from a large acquisition like Splunk fail to arrive as expected, a writedown lands directly on book net assets. That is not a near-term risk, but it is a structural feature you have to accept before holding this for the long run.

The cost pressure running through networking equipment this cycle does not originate in copper or in optics. It originates in memory. The industry readings are extreme: DRAM contract prices rose 93% to 98% quarter over quarter in the first quarter of 2026, the largest single-quarter jump on record, with a further 58% to 63% quarter-over-quarter increase forecast for the second quarter. Samsung took a 32GB DDR5 module from $149 to $239, a 60% increase in one step. SK Hynix confirmed on its earnings call that HBM, DRAM and NAND capacity is already sold out through the end of 2026 (source: TrendForce and other research-house forecasts as relayed by industry outlets including Network World, TechRadar and exIT Technologies, 2025-10 through the first half of 2026). Translated into routers, memory's share of manufacturing cost on mid- and low-end models has gone from roughly 3% to more than 20% (modemguides industry survey, 2026).

Cisco's response runs on two tracks. One is passing the cost out: pricing contributed 5 points of revenue growth in the fourth quarter, with another 4 to 5 points planned for FY2027. The other is designing the material out. The CFO said on the call that more than 30 internal projects are working on exactly that, and that Wi-Fi 7 products cut memory content in half within 90 days (Cisco Q4 FY2026 earnings call, 2026-08-12). Both tracks are running, and gross margin still fell 210 basis points. That is where this supply chain's bargaining position actually sits right now.

The silicon link is, if anything, a Cisco strength. Sixty percent of AI orders are systems built on the in-house Silicon One switching chip. The company does not own fabs and relies on outside foundries; the specific foundry is not disclosed in public filings and I could not confirm it. The other 40%, optical interconnect, is carried by Acacia, acquired in 2021, which crossed $1 billion of orders in the quarter. Readings elsewhere on the same chain corroborate each other: Lumentum issued demand-driven strong guidance on 2026-08-11; AAOI's 2026-08-06 guidance carried both a volume tag and a supply-constrained tag; Airgain on 2026-08-05 likewise described surging demand layered on constrained supply; and Viavi posted quarterly revenue growth of 52.5% (company 8-Ks and guidance disclosures on the dates shown). What the optical interconnect chain is short of right now is capacity, not orders.

The downstream concentration is the passage in this piece that most needs a cool head. Hyperscale cloud providers placed $9.3 billion of orders for the year, but only about $4.0 billion was actually recognized as revenue, roughly 6% of FY2026 total revenue, against under 2% the prior year (Cisco Q4 earnings call, 2026-08-12). At 6%, cloud buyers cannot yet move Cisco's core business. But they contributed the overwhelming majority of the growth, so their marginal leverage is far greater than 6%. The company has not disclosed which three cloud providers these are, and single-customer concentration is not something I could verify.

Where is the soft spot in the contract structure? Orders are not revenue, and that is the number most in need of a haircut this quarter. Against $9.3 billion of orders sits $4.0 billion of revenue, and management set FY2027 AI revenue at $7.5 billion while volunteering the word "prudent," on the grounds that converting large orders into revenue is a non-linear process (Cisco Q4 earnings call, 2026-08-12). The more revealing figure is remaining performance obligations: $46.7B at the end of FY2026, up 7%, with the product portion up 9%, against revenue growth of 12% over the same period (Cisco 8-K, 2026-08-12). Contracted backlog growing slower than recognized revenue means the year was spent drawing down a reserve built up previously rather than adding to it. If that inversion persists for two or three quarters, the visibility of the growth has to be discounted.

The most fragile link is memory supply and pricing. It sits outside Cisco's control, yet it simultaneously sets the floor under gross margin and the pace of customer ordering. The more expensive memory gets, the tighter cloud providers' server budgets get, and the harder it becomes to protect the share left over for networking gear.

The most direct rival in the group is Arista, which meets Cisco head-on in data center switching, and their numbers tell two stages of the same story:

Three things come out of that table. Arista is growing more than twice as fast as Cisco and carries an operating margin more than ten points higher, so on pure growth Cisco is still the one being chased. Yet Arista's gross margin is nearly 3 percentage points below Cisco's, is also falling, and is guided at 62% to 64% for the full year. The entire group is trading gross margin for volume, which means Cisco's 210 basis points of compression is not a management error unique to Cisco. The third point is that order direction is highly consistent: Cisco's 40%, Harmonic's 71%, Extreme's 30% bookings growth and Motorola's 11% order growth all point the same way. This upturn in networking equipment demand is cross-validated across peer readings rather than resting on a single data point.

To measure Cisco and Arista with the same ruler you also have to remember the structural difference. Cisco has $46.7B of remaining performance obligations and $29.8B of deferred revenue underneath it, which makes its results curve far smoother than Arista's. Arista is bet almost entirely on one leg, cloud and AI data centers, which makes it both more elastic and more brittle.

The traditional price-to-earnings anchor has not broken down on this stock. It has stable earnings, positive cash flow and full-year EPS guidance, so a forward multiple works fine. The only complication is whether to use GAAP or adjusted numbers, and those are a full dollar apart.

Using the market capitalization of $447.2B as of 2026-08-13:

What does a 22x forward multiple mean here? For a company guiding to 15% revenue growth and 17% adjusted EPS growth, the PEG works out to roughly 1.3, which is neither cheap nor a bubble, and is genuinely a notch below the thirty- and forty-times multiples routinely attached to comparable names in this AI cycle. But the 6.3x EV/Sales and 2.86% free cash flow yield say something else: the market has already prepaid for that FY2027 guide, the one where three straight quarters cannot miss, as though it were settled fact.

Prepaid for what specifically? For AI infrastructure revenue going from $4.0 billion to $7.5 billion. For 4 to 5 points of price increases being absorbed by customers without argument. For gross margin bottoming at 65% and going no lower. Discount any one of those three and the $5.08 gets revised down, at which point the 22x multiple itself loses the basis for being "a notch cheap." The 8.4% drop on the day of the report was the market selling that prepaid confidence, not the report itself.

The commodity cycle actually running here is memory. Networking equipment is simply the downstream absorber. Three readings help locate it.

Prices are still heading up. After DRAM contract prices rose 93% to 98% quarter over quarter in the first quarter of 2026, a further 58% to 63% increase is forecast for the second quarter, with industry forecasts placing the peak window in the third to fourth quarter of 2026 (TrendForce and other research-house forecasts as relayed by Network World and TechRadar, 2026). Cisco has already passed through 5 points of pricing and plans another 4 to 5 points in FY2027, which means downstream buyers are still absorbing it. Customers still accepting price increases is itself evidence the cycle has not peaked.

Capacity is tight to the point of being sold out. SK Hynix confirmed HBM, DRAM and NAND capacity is sold out through the end of 2026. On the optical interconnect side, both AAOI and Airgain flagged supply constraints in their early-August 2026 guidance, Lumentum issued demand-driven strong guidance on 2026-08-11, and Viavi posted quarterly revenue growth of 52.5% (respective company disclosures). The entire upstream chain is in shortage.

Inventory and conversion cycles are lengthening. Cisco's inventory grew roughly 80% in a year to $5.7B, while remaining performance obligations rose only 7% against revenue growth of 12% (Cisco 8-K, 2026-08-12). More material being stockpiled, contracted backlog growing more slowly: that is the classic signature of the back half of an upswing.

My read is that this cycle is in the middle-to-late portion of its upswing, not the early portion. The basis for that placement is how much weight pricing carries within the growth. Early-cycle looks like volume growth without price increases; late-cycle looks like price increases carrying the numbers by themselves. Right now 5 of the 18 percentage points came from pricing, and 4 to 5 of the planned 15 points in FY2027 are meant to come from pricing, so price is shouldering roughly three-tenths of the growth. That puts us past the middle.

Two trackable leading signals for a peak. The first is the quarter-over-quarter increase in memory contract prices narrowing and eventually turning negative. It will show up before revenue rolls over, and it cuts both ways: easing cost pressure would rescue gross margin, but the same event also means the pace of AI server buildout is cooling, and that side usually does the greater damage. The second is the absolute dollar amount of quarterly AI orders from hyperscale cloud providers. The fourth quarter's $4.0 billion is the high-water mark; if that steps down materially for two consecutive quarters while remaining performance obligation growth continues to lag revenue growth, the contracted-backlog reservoir will be the first thing to run dry.

This chapter is a weekly-horizon signal governing entry and exit timing over the next few weeks. Readers planning to hold for a year or more can skip straight to the next chapter.

In the sector money flow taxonomy, Cisco sits in the data center infrastructure branch under the industrials chain. As of 2026-08-13 that sub-sector has been in the middle tier for 9 consecutive days and sits in the leading quadrant of the relative rotation graph, with the parent industrials sector also in the middle tier. Across sectors, financials occupies the strongest tier alone, while technology, consumer staples, real estate and utilities fall into the weakest tier (data as of 2026-08-13). The classification itself is a reminder: CSCO is riding the data center glow, while the technology sector it naturally belongs to is sitting in the weakest position.

Money genuinely is moving toward this branch, but look closely at how. As of 2026-08-13 the sub-sector has posted 11 consecutive days of net inflow with a money-flow strength reading of 30. The last 5 trading days total +$3.2B of net flow, while the last 20 trading days total only +$2.5B. The entire 20-day net inflow is concentrated in the last 5 days; the prior 15 days were negative in aggregate. Over the same stretch the branch is up 6.7% over 5 days. Money and price are sprinting together, with neither one running ahead of the other, which is the textbook chase structure: the capital coming in got no discount at all, and if the move gives back, the cushion of paper gains is painfully thin.

Zoom out to the broader tape and the 2026-08-13 reading is a strong bull market with a market strength reading of 55, with the S&P ETF at 777.88, 3.09% above its 20-day moving average. The tape itself is not the problem. What deserves the most caution is the rotation pair signal: money is moving out of technology and into financials, with a strength reading of 2.00 (2026-08-13). The top three inflows are gold at 69, China large cap at 50 and the quality factor at 50; the top three outflows are retail REITs, industrial REITs and short-term volatility. The leadership is defensive and safety-seeking, not growth. The macro reading the same day is net liquidity of $5.79T, down 0.8% week over week and therefore contracting, with the 10-year Treasury at 4.68%; liquidity overall is on the cautious side. For a stock trading at 22x on forward adjusted numbers and 28x on GAAP, contracting liquidity plus a 4.68% long-end rate is pure headwind.

The historical statistics pour cold water too. This sub-sector moved up from the weakest tier to the middle tier on 2026-08-03. Between 2021-05 and 2026-07 there were 5,887 comparable upgrades, and the median return over the following 20 trading days was +0.0%, with a 56% probability of being higher. The upgrade itself delivers no excess return. It merely moves a name from "don't touch" to "worth watching."

And there is a divergence sitting between this stock and the branch it belongs to: the sub-sector has 11 straight days of net inflow and is up 6.7% over 5 days, while CSCO fell 8.4% in a single day (2026-08-13). The sector's money did not catch this stock, which tells you this is a repricing at the single-stock level rather than a retreat at the sector level. That distinction determines how you wait, which is the subject of what follows.

Long term (quarters to years): worth following, provided you accept the multiple will not expand further.

The reasoning is straightforward. A 22.3x forward P/E against 15% revenue growth and 17% adjusted EPS growth gives a PEG of about 1.3. The visibility created by $46.7B of remaining performance obligations plus $29.8B of deferred revenue is close to unmatched across the communications equipment group. Of $9.3 billion in AI orders only $4.0 billion has been recognized, and the remainder is future revenue already signed on paper. And 99% of free cash flow goes back to shareholders each year, with $8.1 billion of buyback authorization still available. Added together, that is a set-up with a good chance of delivering low-to-mid teens annualized returns.

There are three costs to accept, and none of them can be dodged. First, gross margin is already pointed down. Memory cost is an exogenous variable, and the company can only chase it with price increases and design savings, with no say over whether it catches up; guidance has already stepped margin down from 66.3% to a 65% to 66% range. Second, this growth contains dilution of quality. Strip out pricing and AI order conversion and real volume growth in traditional enterprise networking is only around 5%, while the recurring revenue line is visibly failing to keep up: full-year ARR of $32.1B grew just 3%, services revenue was flat, and subscription revenue of $6.2B grew 11% and accounted for roughly 36% of the quarter's revenue (Cisco Q4 earnings call, 2026-08-12). Hardware is sprinting while software subscriptions stroll, which weakens the "AI turns Cisco into a growth stock" narrative. Third, $59.5B of goodwill sits on top of $50.3B of shareholders' equity, and if the Splunk account ever has to be written down, book net assets take the hit.

A verification-driven way to hold it: split the position in two, one part tracking gross margin and one part tracking the absolute dollar amount of AI orders. If margin holds 65% and orders do not step down, hold through and let EPS growth and buybacks do the work. If either one breaks, immediately redo the math on the premise that "22x isn't expensive," because at that point the answer to whether it is expensive will have changed.

Short term (days to weeks): don't chase. Wait for two things.

The reasoning is equally direct. The price structure broke first: after an 8.4% single-day drop the stock closed below its 50-day moving average, with a gap overhead and the trendline already broken. The money-flow structure is on the other side too: the sub-sector's entire 20-day net inflow is crammed into the last 5 days, over which the branch has already risen 6.7% (2026-08-13), so entering now means taking the other side with no safety cushion. Meanwhile broad-market money is rotating from technology into financials, macro net liquidity is contracting at -0.8% week over week, and the 10-year Treasury sits at 4.68%. All three are headwinds. The historical statistics offer no assist either: the median return over the 20 trading days following a comparable tier upgrade is +0.0%.

So what are we waiting for? For the price to hold above a support zone with an actual basis, and for money flow to turn from a pulse of inflow into sustained supply. Both of those have quantifiable readings, and the specific thresholds plus three reference levels come in the second half.

Entry condition one (money flow, trackable daily): net inflow into the sub-sector does not break, and the 5-day net flow is no longer larger than the 20-day total. As of 2026-08-13 the reading is 11 consecutive days of net inflow with the last 5 days at +$3.2B, above the last 20 days at +$2.5B, which tells you the inflow is a one-off pulse. When the 20-day cumulative clearly exceeds the 5-day cumulative, meaning the inflow has stretched out into sustained supply, and the consecutive net-inflow day count has not reset to zero, this condition is met. Conversely, if the consecutive net-inflow streak breaks and the money-flow strength reading falls from 30 to below 0, this condition is permanently void and there is no point waiting on the price.

Entry condition two (tied to the next quarterly report): Q1 FY2027 non-GAAP gross margin lands in the upper half of the 65% to 66% guidance range, meaning no lower than 65.5%, while revenue lands within or above the $18.0–18.2B guidance range. Why those two numbers: strip the first quarter's $18.1B out of the $72.8B full-year midpoint and the remaining three quarters have to average $18.23B, which is 5.4% above the record $17.3B just posted, so the revenue line cannot give an inch. And gross margin is the only reading that can prove the 4 to 5 points of price increases actually caught up with memory costs. Break below 65% and the $5.08 EPS midpoint has to be revised down, at which point the 22.3x forward P/E calculated earlier automatically becomes expensive. Both have to hold for this to count as fundamental confirmation.

Three reference levels (each anchored to publicly verifiable data):

On sequencing: if condition one is met before condition two, a starter watch position is reasonable, with the balance added once both are met. If condition two fails but the price has already reclaimed $118.01, that is a purely flow-driven bounce and the position should be one size smaller than normal.

Two things to check daily: whether net inflow into the sub-sector continues (as of 2026-08-13 it is 11 consecutive days with a strength reading of 30), with any break voiding entry condition one; and where the price sits relative to $118.01 and $111.5, since breaking below $111.5 means the company's own buyback range did not hold and the watch position has to go back to waiting.

Three things to check weekly: the quarter-over-quarter direction of DDR5 and server memory contract prices, remembering both sides of it, since a narrowing increase helps gross margin but is also a leading reading on AI buildout cooling; the order and lead-time language from chain peers Lumentum, AAOI and Viavi, because once the supply-constrained phrasing disappears from guidance, the shortage is over; and whether the rotation pair signal from technology into financials is still running (strength 2.00 on 2026-08-13), because only when that turns do high-multiple technology names get room to breathe.

Five things to check each quarter (at earnings): non-GAAP gross margin against the 65% to 66% guidance, with a break below 65% triggering a re-evaluation of the whole thesis; the absolute quarterly dollar amount of AI infrastructure orders, which was $4.0 billion in the fourth quarter, with two consecutive quarters below $2 billion to be treated as an order-cycle peak; remaining performance obligation growth compared against revenue growth, which in FY2026 was an inversion of +7% against +12%, with a widening inversion indicating the reservoir is draining; whether ARR (FY2026 at $32.1B, up 3%) lifts off that 3%, the only evidence that "hardware sprinting, subscriptions strolling" can be repaired; and whether inventory (FY2026 year-end $5.7B, from $3.16B at the start) keeps swelling, since it is the first line item to get hurt when demand softens.

Circumstances that trigger a follow-up update: gross margin below 65%, quarterly AI orders below $2 billion, sub-sector net inflow breaking and turning to net outflow, or the share price falling below $101 — any one of these four.

Catalyst calendar:

Every item on the tracking list above is something we watch in the data every day, whether it is the count of consecutive net-inflow days for the sub-sector, the share price relative to the two moving averages, or the gross margin and absolute AI order figures in the next quarterly report. If any one of them triggers, we will publish a follow-up piece alongside that day's daily review, so readers do not have to sit refreshing the data pages themselves. Every money-flow and sector reading in this piece carries its data date and can be verified item by item on the corresponding data pages on the site. The company-side figures all come from Cisco's 2026-08-12 announcement and the earnings call held that day, plus each peer company's own public disclosures, with sources given in parentheses. The upturn in networking equipment demand is real, and the pressure on the cost side is real too. Both will be delivered over the next two to three quarters, and telling which arrives first is exactly why the list exists.

This article was produced on the NextPick research workbench; every money-flow, tier and valuation figure it cites can be verified on the corresponding data surfaces of the site, and the moment any tracking condition listed in the report triggers, an addendum will go out together with that day's daily review.