Cisco just did something Cisco doesn’t do. It moved.
Shares of CSCO popped 17 percent in extended trading Wednesday after the networking giant posted fiscal Q3 revenue of $15.84 billion versus the $15.56 billion analysts had penciled in, and adjusted EPS of $1.06 against a $1.04 consensus. The kicker, though, was not the beat. It was the number Chuck Robbins put on the AI infrastructure scorecard: a full-year hyperscaler order target of $9 billion, nearly double the $5 billion the company was guiding to one quarter ago.
For a stock that has spent most of the last decade synonymous with “value trap,” that is a regime change. The old Cisco trade was a 3 percent dividend, a single-digit growth rate, and a slowly eroding switch business that Wall Street priced like a utility. The new Cisco trade is the one the after-hours tape just printed: a credible second-derivative play on AI capex, with $5.3 billion in cumulative AI infrastructure orders booked through three quarters and the largest single-day move in the stock since the late 1990s.
The Numbers Behind The Pop
Total revenue grew 12 percent year over year, the fastest organic growth rate Cisco has posted since the post-pandemic networking surge. Robbins disclosed that networking product orders climbed more than 50 percent year over year, and data center switch orders jumped more than 40 percent. Those are not “AI-adjacent” numbers. Those are AI numbers.
The company also raised FY26 revenue guidance to a range of $62.8 to $63.0 billion, up from the prior $61.2 to $61.7 billion band. Free cash flow guidance moved in lockstep. According to the company’s Q3 FY26 earnings 8-K filing, management lifted EPS guidance to the upper half of its previous range and reaffirmed the $14 billion FY26 share repurchase commitment.
The AI infrastructure order book is the line item Wall Street has been chasing for two quarters. Cisco’s silicon photonics, optical interconnects, and high-performance Ethernet switching slot into the same hyperscaler buildouts driving the Nvidia-Corning $3.2 billion optical onshoring deal we covered last week. The big customers are the same names. The capex cycle is the same cycle. Cisco is finally getting paid for participating.
Splunk Is Quietly Working
Two years ago, the $28 billion Splunk acquisition was, depending on which sell-side desk you asked, either a strategic masterstroke or a defensive money pit. The bear case was that Cisco had bought a slowing observability business at a frothy multiple to plug a hole in its growth story.
The May 13 results suggest the deal is finding its feet. Splunk added 500 new logos in the first half of FY26 and is on track for 1,000 by year end. The cloud subscription transition remains a near-term revenue drag, which is what migrations always look like before they become tailwinds. But on the call, management framed the Splunk security and observability data plane as the front door to the AI infrastructure stack, not an attached product.
For investors who watched Cisco overpay for Scientific Atlanta and lose at the cable headend, the bar for “Splunk is working” is high. This was the first quarter where the answer was unambiguously yes.
The Layoff That Was Not A Retreat
Robbins also confirmed Cisco would cut fewer than 4,000 jobs this quarter, roughly 5 percent of total headcount. In a different earnings season, that headline would have dominated the after-hours coverage. Wednesday it barely registered. Why? Because Cisco is doing the same move every hyperscaler-adjacent company is doing right now: trimming the bottom and pouring the savings into AI talent and silicon.
This is the same playbook that drove the AMD Q1 2026 record on data center revenue, the Fidelity tech reorg, and the Cloudflare “agentic AI” workforce reset. The capital is moving from low-margin services and back-office engineering toward the parts of the stack that touch AI revenue. Robbins simply did it without the dramatic press release.
The market read it correctly. CSCO did not trade like a layoff story. It traded like a margin story.
What This Means For The Networking Trade
The clearest tell on whether the Cisco quarter was a one-off or a sector signal is what happens in pre-market Thursday across the rest of the networking complex. Arista Networks, the hyperscaler darling that has historically been the long when Cisco was the short, faces a different question now. If Cisco is real in the AI buildout, Arista’s premium multiple compresses. If Cisco is fluke, Arista holds.
HPE-Juniper, the third leg of the trade, also has to answer for itself. The combined entity has been pitching the Mist AIOps story for two quarters with limited traction. A Cisco that is winning the AI switching budget makes that pitch harder, not easier.
Then there is the bigger question: how much of the AI capex dollar actually lands in networking versus compute? For three years, the answer has been “almost none.” Roughly 80 cents of every AI capex dollar has gone to Nvidia and accelerator suppliers. The Cisco quarter is the first hard data point suggesting that ratio is rebalancing as hyperscalers build out the optical, switching, and interconnect plumbing that the GPUs sit on top of.
According to CNBC’s coverage of the report, Robbins specifically called out scale-out switching for AI clusters and 800G optical as the fastest-growing product lines. That is the networking layer of the AI infrastructure stack, and it has been the most underestimated part of the buildout.
The Trade From Here
CSCO closed Wednesday’s regular session near $76 and printed indications above $89 in extended trading. At those levels, the stock is no longer cheap on the multiple it traded at six months ago, but it is also no longer the value trap the market priced it as. The forward P/E shift implied by the new guidance puts Cisco at roughly 19 times next year’s earnings, in line with the broader S&P and well below the AI infrastructure cohort.
The risk is straightforward. If the $9 billion AI infrastructure target turns out to be front-loaded with one or two hyperscaler commitments rather than a durable book of orders, Q4 numbers will tell on the company fast. Cisco has been here before with its 2000-era backlog math, and Wall Street remembers.
But on a single-quarter read, this is the strongest Cisco tape since the dot-com peak. The market spent twenty years pricing Cisco as the wreckage of the last AI buildout. The May 13 numbers are the first real argument that Cisco is, finally, a beneficiary of the next one.
If you owned CSCO at 14 times earnings, the trade just got rerated for you. If you did not, the question is whether the next 17 percent comes from multiple expansion or further upside on the order book. The answer is probably both.