Artificial intelligence once again proves that it is not supported solely by GPUs. Each cluster requires switches, optics, cables, network cards, cooling systems, rack integration, and a manufacturing chain capable of scaling all these components. Cisco has made this clear in its fiscal Q3 2026 results: demand for AI infrastructure is strongly powering its networking business and providing a boost to Taiwanese suppliers involved in the global supply chain of networking equipment and data centers.
The company closed the quarter with record revenues of $15.8 billion, up 12% year-over-year, with a 35% increase in product orders. Even excluding hyperscalers, orders rose by 19%. The most noteworthy figure for the sector is that networking product orders grew by over 50%, while data center switching orders increased by more than 40%. Cisco attributes much of this momentum to AI infrastructure and the renewal cycle of enterprise and campus networks.
Another key figure summarizes the shift: Cisco has accumulated $5.3 billion in AI infrastructure orders from webscale customers so far this fiscal year and has raised its full-year forecast to $9 billion, up from the previous estimate of $5 billion. It has also increased its AI revenue forecast for the year to $4 billion. For a company historically associated with routing, enterprise switching, and network security, the message is clear: AI is restoring prominence to the connectivity layer.
Network returns to the core of the data center
During the initial phase of the AI boom, the focus was almost entirely on accelerators. NVIDIA, HBM, CoWoS, and training capacity dominated the conversation. But large clusters do not operate as isolated machines. Thousands or tens of thousands of GPUs need to communicate with each other with low latency, high bandwidth, and minimal packet loss. This necessity is elevating the strategic importance of Ethernet, advanced optics, high-capacity switches, and architectures designed for east-west traffic within data centers.
Cisco aims to occupy this space with its Silicon One family and a portfolio of AI networking systems competing with Broadcom, NVIDIA, and Arista. In February, the company introduced the Silicon One G300, a 102.4 Tbps switching chip designed for large-scale AI deployments, fabricated using TSMC’s 3nm process (according to Reuters), intended to power new Cisco 8000 and N9000 platforms with high-density liquid cooling and optics.
The message for Taiwan is straightforward: although Cisco designs much of its technology, the manufacturing and assembly of network infrastructure depend on a broad supply chain on the island and within its industrial ecosystem—advanced semiconductors, optical modules, boards, server integrators, ODM, EMS, testing, connectors, cabling, and thermal solutions. When Cisco sells more AI infrastructure, the effect extends beyond its accounts to suppliers that produce or assemble parts of that infrastructure.
DigiTimes precisely highlighted this point: Cisco’s results have injected new momentum into the networking industry, with AI-related order growth benefiting the Taiwanese ecosystem of suppliers. The publication notes a trend already observed: AI spending is shifting from main chips to transmission layers, switching, and system assembly.
Taiwan captures value beyond AI servers
Taiwan has already been a central player in AI servers. Foxconn, Quanta, Wistron, Wiwynn, and Inventec have gained importance as system and rack integrators for hyperscalers and chip suppliers. Reuters reported in 2025 that Taiwan accounts for approximately 80% of global server shipments and over 90% of AI servers, a position that explains why every increase in AI capex often reflects in its tech supply chain.
Foxconn is the most visible example. In its latest results, the company linked its growth again to demand for AI hardware. The Wall Street Journal reported that cloud and AI-related networking products made up nearly half of its quarterly revenue, with consumer electronics accounting for about a third. Reuters also noted that Foxconn expects AI rack shipments to more than double this year, driven by advanced infrastructure clients.
But the opportunity goes beyond final server assembly. The network is becoming a bottleneck. The transition from 400G to 800G, and eventually to 1.6T, requires upgrading switches, transceivers, connectors, cables, boards, cooling, and testing systems. DigiTimes had already reported early this year that Taiwanese network equipment suppliers are increasing investments due to the move toward 800G and 1.6T, driven by demand from cloud and AI data centers.
In this context, companies like Accton, Sercomm, Wistron NeWeb, Delta Networks, Foxconn, Quanta, and Wiwynn appear across different layers of Taiwan’s industrial map, though not all have the same exposure or relationship with Cisco. Some are closer to networking equipment; others to server integration, power supplies, thermal systems, or electronics manufacturing. The common point is that AI infrastructure is broadening the market beyond the core accelerator.
Cisco is also restructuring for the AI era
The strong results have not prevented an uncomfortable decision. Cisco announced a workforce reduction of nearly 4,000 employees, less than 5% of its staff, as part of a restructuring aimed at reallocating resources toward AI, silicon, optics, and cybersecurity. Reuters explained that the company anticipates charges up to $1 billion for this process, with approximately $450 million recognized in Q4.
The paradox plays out at many large tech firms: strong results alongside workforce adjustments. Not necessarily because the business is performing poorly, but because AI is shifting investment priorities. Cisco aims to concentrate resources on areas expected to grow most: AI data center networks, own silicon, high-density optics, security, and software platforms.
For its suppliers, this means a mix of opportunity and challenge. The opportunity lies in increased order volumes. The challenge is in meeting tighter specifications: more speed, higher density, better energy efficiency, superior cooling, and tight delivery timelines. AI not only demands more units; it requires equipment that operates in racks with significantly higher power consumption, faster links, and reduced fault tolerances.
Cisco’s growth also reinforces a rising industry trend: Ethernet is gaining ground as an open and scalable alternative for AI clusters. NVIDIA continues to dominate some of the ecosystem with InfiniBand and its proprietary networking stack, but many hyperscalers prefer Ethernet solutions for flexibility, cost, multiple vendors, and compatibility with existing infrastructure. Cisco is seeking to leverage this opportunity with its switching portfolio, Silicon One, and optics.
For Taiwan, the message is positive but carries risks. If the AI cycle continues, networking suppliers and system providers can benefit from several years of investment. However, if spending among hyperscalers pauses, inventories pile up, or deployment delays occur, the impact could be swift. Taiwan’s supply chain is already familiar with such volatility: during 2023 and 2024, many suppliers experienced inventory corrections in consumer networks and broadband equipment before AI revived demand in higher-value segments.
The key difference now is that the focus has shifted. Networking is no longer a secondary part of the data center. In large-scale AI, the network determines how effectively GPUs are utilized, the occurrence of bottlenecks, and the actual performance achieved per dollar invested. Improvements in switching, optics, or rack integration can translate into more efficient clusters.
Cisco’s results demonstrate that this market is starting to be reflected in actual orders, not just in technical presentations. And Taiwan, with its mix of electronics manufacturing, server, networking, optics, and systems suppliers, is once again well-positioned in the supply chain. The AI boom began with accelerators, but it is now spreading through the entire connecting infrastructure.
Frequently Asked Questions
What has Cisco announced in its latest results?
Cisco reported record revenues of $15.8 billion in fiscal Q3 2026, with product orders up 35% and networking orders growing over 50% year-over-year.
Why is AI driving the networking business?
Because AI clusters need to connect thousands of GPUs with low latency and high bandwidth. This increases demand for switches, optics, cables, network cards, and integrated systems.
How does Taiwan benefit?
Taiwan participates in many layers of the infrastructure: AI servers, rack integration, electronics manufacturing, networking equipment, modules, testing, thermal components, and supply for data centers.
What role does Cisco Silicon One play?
Silicon One is Cisco’s family of network chips. The G300, introduced in 2026, offers 102.4 Tbps and is targeted at large AI data centers requiring high-capacity switching.