Cisco set a excessive bar for itself within the new 12 months after a prolific 2025.
Final 12 months, Cisco refreshed its core switching and router portfolios, including new software program capabilities for the info middle in 2025. It additionally made a variety of AI networking strikes, together with the introduction of the Cisco 8223, a high-end, 51.2 Tbps router and chip designed for distributed AI workloads.
As well as, Cisco expanded its Nvidia relationship with a variety of joint choices, together with the Cisco N9100 sequence knowledge middle change. The 64-port, 800Gb Ethernet change is powered by Nvidia’s Spectrum-4 ASIC, which is designed for AI, high-performance computing and cloud knowledge middle networks that require ultra-low latency, congestion management, and predictable efficiency.
Cisco additionally teamed with Nvidia and VAST Knowledge to supply prospects a pre-integrated AI infrastructure bundle, together with compute, community, storage and knowledge, to assist AI workload knowledge materials in addition to large-scale coaching and inference.
The demand for AI infrastructure has already begun to impression Cisco’s earnings, and the seller expects continued development.
On the shut of its 2025 fiscal 12 months in late July, Cisco reported income of $14.7 billion for the fiscal fourth quarter, which is up 8% in comparison with the year-earlier quarter. Annual 2025 income got here in at $56.7 billion, which is up 5% from the earlier 12 months.
“Enterprise product orders had been up 5% year-over-year in This fall [while] networking product orders grew double digits in This fall, marking the fourth consecutive quarter of double-digit development,” CEO Chuck Robbins stated on the time. “There may be robust curiosity from prospects within the new household of Cisco Catalyst 9000 sensible switches, together with a totally refreshed lineup of extremely safe routers, wi-fi entry factors, and industrial IoT units, that are purpose-built for the AI-ready campus and department.”
Count on Cisco to proceed including to and solidifying these merchandise within the coming 12 months, business watchers say, as the seller goes after what it sees as a multi-year, multi-billion-dollar refresh alternative within the enterprise networking enviornment.
It’s an Ethernet world
Efforts to increase Ethernet’s dominance are gaining traction, which may gain advantage Cisco.
In a observe about 2026 data center networking predictions, analyst Sameh Boujelbene said that scale-up is rising as a brand new battlefield for Ethernet:
“After securing a number one place within the scale-out phase of AI back-end networks, Ethernet is now increasing into scale-up, the place NVLink has traditionally dominated. On this house, Ethernet will compete not solely with NVLink but additionally with UALink, one other different to NVLink,” wrote Boujelbene, vice chairman of the Dell’Oro Group. “We anticipate 2026 can be a 12 months filled with vendor bulletins concentrating on each Ethernet and UALink alternatives in scale-up. Scale-up represents what could possibly be the most important complete addressable market enlargement the business has ever seen.”
In 2025, Ethernet overtook InfiniBand in AI back-end networking — “a decisive turning level,” Boujelbene wrote. “This shift is especially placing on condition that simply two years in the past, InfiniBand accounted for practically 80% of the info middle change gross sales in AI back-end networks.”
The speedy adoption of Ethernet in AI back-end deployments propelled complete Ethernet knowledge middle change gross sales to an all-time excessive in 2025, practically doubling annual revenues in contrast with 2022 ranges, Boujelbene said. “800 Gbps effectively surpassed 20 M port inside simply three years of shipments: As a degree of reference, it took 400 Gbps six to seven years to attain the identical milestone.”
Including to the Ethernet effort are teams such the Extremely Ethernet Consortium and the Ethernet for Scale-Up Networking, that are working to undertake Ethernet for AI and high-performance computing. Cisco is a contributor to each efforts.
“Cisco is a big incumbent, which signifies that they win by default so long as they don’t mess up too badly,” stated Tom Nolle, principal analyst at Andover Intel and a Community World columnist. “Thus, they should kiss the standard networking infants, which means AI and safety. The important thing for them is to be significant in these areas so a competitor doesn’t steal prospects/market share.”
AI infrastructure momentum
Demand for AI infrastructure is rising at an exponential charge, which stands to learn Cisco and its main rivals, together with Arista, HPE/Juniper, and Excessive.
Whole international knowledge middle IT gear spend — together with servers, accelerators (XPUs), networking, optics, and storage — between 2025 and 2030 is anticipated to succeed in as a lot as $4.7 trillion {dollars}, in accordance with a McKinsey & Company report cited in a blog post by Jake Katz, Cisco’s vice chairman, AI product administration.
Hyperscalers are answerable for over 60% of this infrastructure funding, Katz famous. Neoclouds are answerable for about 17%, and their share is anticipated to develop to over 30% over the following ten years, Katz said.
Whereas a lot of the demand for AI knowledge middle capability immediately comes from hyperscalers and neocloud suppliers, that may change as enterprise prospects delve extra into the AI networking world.
“The opposite ecosystem members and enterprises themselves have gotten answerable for an growing proportion of the AI infrastructure buildout as inferencing and agentic AI, sovereign cloud, and edge AI turn out to be extra mainstream,” Katz wrote.
Extra enterprises will transfer to host AI on premises by way of the introduction of AI brokers which can be designed to inject clever perception into purposes and assist enhance operations. That’s the place the AI impression on enterprise community visitors will seem, suggests Nolle.
“Enterprises have to host AI to create AI community impression. Simply accessing it doesn’t do a lot to visitors. Having cloud brokers entry native knowledge middle sources (RAG and many others.) creates a governance challenge for many company knowledge, in order that received’t go too far both,” Nolle stated.
“Enterprises are taking a look at AI brokers, not the way in which hyperscalers tout agentic AI, however brokers operating on small fashions, usually open-source, and are regionally hosted. That is the place actual AI visitors will develop, and Cisco could possibly be susceptible in the event that they don’t perceive this level and no less than increase it in dialogs the place AI internet hosting comes up,” Nolle stated. “I don’t count on they’d go too far, as a result of the true marketplace for enterprise AI networking might be a pair years out.”
In the meantime, observers count on Cisco to proceed bolstering AI networking capabilities for enterprise department, campus and knowledge facilities in addition to hyperscalers, together with by means of optical assist and different gear.
Infrastructure modernization fueled by Cisco-Nvidia pairing
Know-how options supplier World Huge Know-how lately printed its IT infrastructure Modernization Priorities for 2026 report, which discovered that many organizations are nonetheless reliant on outdated, legacy infrastructure that’s ill-equipped to deal with fashionable workload necessities. The rise of AI has heightened the significance of IT modernization, World Huge Know-how stated.
Cisco’s wide-ranging partnership with Nvidia reinforces its plans to advertise that sort of modernization.
For enterprise prospects, probably the most attention-grabbing components of the Cisco-Nvidia collaboration is the tie-in between Cisco’s Silicon One-based switches and Cisco N9100 Collection Switches that incorporate Nvidia’s Spectrum-X Ethernet change silicon and are managed from the Cisco Nexus dashboard.
“Embedding Spectrum X expertise right into a Nexus setting, the place prospects already probably have the ability units, they know the way to handle and run that setting, is an excellent transfer,” Neil Anderson, vice chairman and CTO of cloud, infra, and AI options for World Huge Know-how, instructed Community World. “Should you’re a financial institution buyer for instance, you could have already certified on Nexus. Now all you must do is incorporate the Nvidia expertise — it simply accelerates the power to undertake it and assist it.”
One of many causes Nvidia is partnering with Cisco is as a result of they know that this distributed phenomenon goes to occur, Anderson stated. “And so they want someone that’s acquired deep community expertise to increase a cluster and make it distributed.”
One other factor of the Nvidia partnership that’s anticipated to affect Cisco’s 2026 efforts is the Cisco Safe AI Manufacturing facility with Nvidia, which brings collectively Cisco safety and networking expertise, Nvidia DPUs, and storage choices.
“The notion of a safe AI manufacturing facility and what they’re doing with Nvidia there’s distinctive available in the market to me, and it resonates with our purchasers. As a result of should you’re going to run your mental property on that distributed cluster, you’ve acquired to have safety within the stack, and that’s what this addresses,” Anderson stated. “I’m encouraging them to maintain doubling down on that one in 2026.”
Not associated to the Nvidia partnership however solidly targeted on distributed AI is Cisco’s new Unified Edge providing, in accordance with Anderson. The Unified Edge, introduced in November, is an built-in bundle of networking, compute and storage aimed toward serving to enterprise prospects extra effectively deal with knowledge from AI and different workloads on the edge, the place the info is often created.
In the end, prospects are transferring from centralized genAI clusters within the knowledge middle to extremely distributed AI that lives nearer to the sting the place knowledge is created, Anderson stated. “This bundle gives the distributed efficiency and ease we predict can be an vital, helpful providing for purchasers.”
