Cisco shares rocketed as much as 18% into record territory after hours Thursday as the networking giant unveiled a $1 billion restructuring, including nearly 4,000 job cuts, to double down on artificial intelligence infrastructure. The move followed a stronger-than-expected quarter and an upgraded long-term outlook, with CEO Chuck Robbins telling investors the company will reallocate aggressively to where demand is “strongest.” CSCO last traded near $102, up sharply on the day.
Cisco is cutting roughly 5% of its workforce to redirect spending toward AI data center networking, optics, security, and software tied to model training and inference. It is a blunt message: AI is the growth lane, and everything else will be sized to fit. Robbins framed the decision as a test of discipline in an arms race where capital and execution speed separate winners from the pack. Cost saves from the plan will be recycled into high-throughput Ethernet switching, 400G and 800G optics, and software layers that help large customers operate AI clusters reliably at scale.
The knife is coming out at a moment of strength. Cisco posted fiscal Q3 revenue of $15.8 billion, up 12% year over year, with adjusted net income of $4.2 billion, or $1.06 a share. Management raised its fiscal 2026 revenue forecast to a range of $62.8 billion to $63.0 billion from $61.2 billion to $61.7 billion, a clear tell that AI orders and pipeline visibility are improving. Investors who have watched hyperscaler capex break records this year are betting Cisco is now better positioned to capture a bigger slice of the networking spend that underpins training clusters, inference farms, and the security perimeters around them.
Underneath the headline, the strategic wager is on Ethernet winning share in AI data centers. InfiniBand—favored in early AI clusters—has dominated the highest performance interconnects. But hyperscalers and large enterprises are pushing for Ethernet-based architectures that are cheaper, easier to scale, and more open. Nvidia has rolled out Ethernet offerings alongside its InfiniBand portfolio, and the industry’s Ultra Ethernet Consortium is pushing a roadmap to close latency and congestion gaps. Cisco’s core strength is Ethernet. If the center of gravity in AI clusters tilts toward Ethernet at 400G and 800G, that’s a structural tailwind for the company’s switching, routing, optics, and telemetry stacks.
Arista Networks has been the market’s proof point that Ethernet can scale for AI workloads, riding demand for cloud spine-leaf fabrics and next-gen optics. Broadcom, which is critical in merchant silicon for Ethernet switches and optical engines, stands to benefit as well. Cisco’s message tonight is that it intends to fight for share across that stack with a fuller portfolio and tighter integration. The company’s observability and security tools—augmented by past acquisitions—give it a differentiated attach opportunity: sell the pipes, the visibility, and the protection together. If Cisco can compress qualification cycles with top customers and cut total cost of ownership for large training and inference networks, share gains can follow.
The stock reaction isn’t just about headlines. Cisco’s multiples and growth expectations have long reflected a legacy hardware story, with heavy exposure to enterprise refresh cycles and government spending. An AI mix shift changes that narrative. Recurring software and services tied to AI operations, optics with better pricing dynamics, and high-capacity switching can reshape gross margins and revenue durability. The 18% after-hours pop implies investors are assigning a higher probability that CSCO participates meaningfully in the AI build-out—beyond one-off wins—backed by a multiyear capex wave from cloud, enterprise, and telecom customers.
The risks are real. Job cuts on this scale can unsettle teams and slow delivery if not sequenced carefully. AI programs demand specialized talent and long qualification cycles with exacting customers; any stumble could see orders diverted to rivals. Component supply, especially for advanced optics, remains tight in parts of the chain. And while Ethernet is closing the gap, latency, congestion management, and lossless performance at hyperscale are unforgiving problems. Cisco will need to prove that its fabric, silicon, and software can keep up with the rapid cadence of AI model and system upgrades without crushing customer budgets or complexity.
Watch bookings tied to AI infrastructure specifically, not just aggregate product revenue. Look for disclosures on 400G and 800G shipments, optics availability, and early wins in training or large-scale inference deployments. Track the sales motion with top cloud providers and large enterprises building private AI clouds. Signals from hyperscaler capex guides, server GPU shipments, and data center power build-outs will triangulate with Cisco’s order trends. On margins, mix should matter: optics and software attach can cushion volatility if broader enterprise networking slows. Any acceleration in deferred revenue tied to software subscriptions would support the rerating case.
Investors are rewarding credible AI roadmaps and visible cost discipline. Cisco offered both, and the stock responded. The company now has to turn a clean headline into durable contracts and consistent execution. If Ethernet continues to win designs in AI clusters and Cisco can marry high-capacity switching with observability and security at scale, today’s spike could mark the start of a more durable phase for the stock. Miss the qualification windows or stumble on delivery, and the multiple swings back toward a legacy profile. The next two earnings prints will show whether this AI pivot is momentum or muscle.
In the meantime, the tape is clear. CSCO is trading at record levels on a plan to cut, refocus, and lean into AI infrastructure. It is the kind of decisive move markets demanded all year from legacy tech. Now Cisco has to deliver, packet by packet, on the promise investors just bid to the highs.
