Cisco reports fourth-quarter and full-year fiscal 2026 results after the close today, with the call scheduled for 4:30 p.m. Eastern, and Wall Street has spent the year repricing the company as an artificial-intelligence infrastructure story rather than a networking incumbent. The problem with that trade is arithmetic: to hit the $9 billion AI order target management raised in May, Cisco has to book roughly $3.7 billion of hyperscaler orders in a single quarter, after taking $5.3 billion across the three that came before it.
The Number That Has to Print
Cisco’s own third-quarter release is the cleanest source on where this stands. The company reported record revenue of $15.8 billion, up 12%, lifted its full-year AI order goal from $5 billion to $9 billion, and disclosed $5.3 billion of AI orders taken year to date. Total product orders grew 35%, and 19% excluding hyperscalers, which is a genuinely healthy enterprise number that gets almost no attention because the hyperscaler line is doing all the narrative work.
Consensus for tonight sits near $1.17 in non-GAAP earnings on roughly $16.8 billion of revenue. Those are not the figures that will move the stock. Three things are:
- The Q4 hyperscaler AI order number. Anything meaningfully short of $3.7 billion means the $9 billion target was set on a customer schedule Cisco does not control.
- Fiscal 2027 guidance. The order book only matters if management will say what converts, and when.
- Whether AI revenue guidance moves again. It went from $3 billion to $4 billion in May. A third revision in either direction tells you how firm the backlog really is.
Options pricing suggests the market knows this is binary. Traders are positioned for a share-price swing of more than 8%, which is a wide band for a company that spent a decade as a dividend-paying bond proxy.
Orders Are Not Revenue, and Cisco Said So Itself
Here is the gap almost nobody puts on a slide. Cisco guided to $9 billion of AI orders and about $4 billion of AI infrastructure revenue in the same fiscal year. Less than half the order book converts inside the period it was booked. The rest becomes backlog that recognizes over future quarters on a schedule set by when hyperscalers actually take delivery and stand equipment up.
Against full-year revenue guidance of $62.8 billion to $63.0 billion, that $4 billion is about 6% of the company. Six percent. The multiple expansion this year has been driven by a line item that, on management’s own numbers, is a rounding error against the installed base of switches, routers, security and the subscription software that actually pays Cisco’s bills.
That is not an accusation of hype. Backlog conversion lag is normal in infrastructure, and a quadrupling order book is a real signal about where enterprise and hyperscaler networking spend is heading. It is a statement about what the stock is currently pricing versus what the income statement will show for at least another year. Those are different clocks, and only one of them is in Cisco’s control.
The Competitive Read Nobody Is Making
Cisco is not the only company selling the plumbing between GPUs, and the AI networking market has been consolidating around exactly this pitch. Hewlett Packard Enterprise spent the last year integrating Juniper on the same argument about AI networking, and Arista continues to hold the accounts where hyperscalers buy merchant silicon and build their own stack.
The strategic question tonight is whether Cisco’s Silicon One and optics business is winning share or riding a rising tide. A $3.7 billion quarter from a handful of named hyperscalers is concentrated revenue by definition, and concentration cuts both directions: it is the fastest way to a record quarter and the fastest way to a guidance reset when one customer moves a deployment two quarters right. When we covered the $9 billion target being set in May, the open question was whether the order pace could hold. Tonight is the first hard answer.
What a Miss Would Actually Mean
Not much for the business, quite a lot for the stock. Cisco has a fortress balance sheet, a subscription base that keeps growing, and enterprise product orders up 19% without a single hyperscaler dollar. A short AI quarter dents none of that.
What it does dent is the premise. The bull case going into today, as the bull and bear framing has it, requires AI orders to be accelerating rather than merely sustaining, because the rerating already happened. Stocks that have been repriced on a story do not get graded on the quarter. They get graded on whether the story still holds, and Chuck Robbins has one call to make that case with a number attached.