Cisco Calls It a Networking Super Cycle — And Raises Its AI Infrastructure Target to $9 Billion
J.P. Morgan 54th Annual Global Technology Conference, May 18, 2026
Cisco Chairman and CEO Chuck Robbins sat down with J.P. Morgan analyst Samik Chatterjee at the firm's annual technology conference in what turned out to be one of the more substantive executive appearances of the conference season. Robbins covered a wide range of topics — from Silicon One's expanding role in hyperscaler architectures to the security implications of Anthropic's Mythos model — and offered several data points that investors had not previously heard.
The $9 Billion AI Infrastructure Target Is the Headline Number
The most significant disclosure was Robbins confirming that Cisco has raised its AI infrastructure revenue target for the fiscal year to $9 billion, up from the prior $5 billion target. That is an 80% upward revision, and Robbins was unambiguous about what is driving it: Silicon One. "If we didn't have our own silicon, the $9 billion that we announced would probably be close to zero, because we would simply be a sheet metal distributor of merchant silicon," he said. The comment is pointed. Without proprietary silicon, Cisco's value proposition in hyperscaler networks — where customers increasingly demand either custom software or custom chips — would be severely diminished. The decision to acquire Leaba in 2016 for its chip design capabilities, which Robbins acknowledged was prescient even if the full payoff was not obvious at the time, is now the foundation of a roadmap that will have Silicon One powering all high-end systems across the entire Cisco portfolio by fiscal year 2029, with the ramp accelerating through fiscal 2026.
Robbins also highlighted five new hyperscaler design wins in the most recent quarter alone — two in optics and three in systems — and disclosed that a third hyperscaler has been awarded a scale-out architecture win in just the first weeks of the current quarter. The scale-out category, addressed by the P200 chip, is purpose-built for connecting multiple AI data centers together as AI clusters outgrow single-facility deployments. This inter-data-center connectivity use case is relatively new and appears to be gaining commercial traction faster than the market had anticipated.
Mythos: Cisco Had Access for Seven Weeks Before It Went Public
Robbins' discussion of Anthropic's Mythos model was unusually candid. Cisco had the model, including a version with no guardrails, for seven weeks before it became public knowledge, and the company restricted access internally to its security team only. The team used it to run Cisco's own code through the model to identify vulnerabilities that human reviewers had missed — and also confirmed that the model is highly capable of generating exploits to take advantage of those same vulnerabilities. "These models are as bad as they're going to be right now, and they're pretty good," Robbins said, signaling that the offensive cybersecurity threat from such models will only intensify.
In response, Cisco open-sourced a framework it calls a "harness" — a structured methodology for running software through these models to uncover vulnerabilities — and published a white paper called Shields Up with defensive recommendations for enterprise customers. It also announced Cisco IQ, a tool customers can deploy into their own infrastructure to inventory all Cisco equipment, flag what is past last day of support, and identify what can no longer be patched. This last item connects directly to a significant commercial opportunity: Robbins estimated that end-of-life equipment in customer environments represents "at minimum tens of billions and perhaps $100 billion to $200 billion" in potential refresh spend. He was careful to note that nothing from this opportunity is currently baked into guidance, and that the refresh cycle is still in its earliest stages — "top of the first" in his words, with mid-single-digit percentages of the addressable installed base having been replaced so far.
Security Business: Progress, But Slower Than Expected
Robbins was notably measured on the security business. "It has been a part of the portfolio that we've been hoping would improve, and it's beginning to show that everything we believe is going to happen, it's just happening at a slower pace than what we had hoped, to be honest." That is a meaningful admission. He did note five consecutive quarters of high firewall win rates against competitors following a portfolio refresh, and significant firewall growth in the most recent quarter. The longer-term thesis — that agentic AI will require security to be embedded directly in the network rather than routed through external appliances — remains intact, and Robbins emphasized that no networking competitor has a meaningful security business and no security competitor has a meaningful networking business. But the pace of monetization continues to lag the strategic narrative.
On the smart switch product line specifically — switches that incorporate CPUs or DPUs to run security services in-line at line speed — Robbins estimated that only about 5% of what customers have been buying from Cisco has been smart switches to date. He expects that to escalate, particularly as agentic AI deployments at the edge, in manufacturing environments and robotics, create latency constraints that make appliance-based security architectures untenable.
Optics: A $1 Billion-Plus Quarter and a Structural TAM Doubler
Robbins confirmed that Acacia delivered over $1 billion in revenue in the most recent quarter from long-haul optics, and articulated why owning both optical and systems capabilities is strategically important: pluggable optics work with any vendor's equipment, meaning Cisco captures optics revenue even in deployments where a competitor wins the switching business. "The optics is at least equal to the size of the networking spend of the switching spend. So you really double your addressable market by having both." As co-packaged optics — where the optical components are integrated directly onto switching ASICs — approaches commercial deployment, Cisco's ownership of both sides of that equation positions it to benefit from what could be a significant architectural transition.
Supply Chain as Competitive Moat
Robbins described Cisco's supply chain posture in terms that go well beyond operational management. The company has made equity investments into supply chain partners to help them diversify geographically, and made a specific investment in memory supplier Nanya to help secure memory supply, which he described as "a big issue right now." CFO Mark Herren is scheduled to present at the same conference the following day and is expected to provide more detail on the magnitude of advance purchase commitments and inventory levels the company has been accumulating. Robbins framed the supply chain buildout as a competitive differentiator — one that is enabled by Silicon One giving Cisco greater control over its component requirements — rather than a defensive necessity.
AI's Threat to Software: Largely Dismissed, With One Honest Carve-Out
On the broader question of AI disrupting enterprise software businesses, Robbins was relatively sanguine about Cisco's own exposure, while acknowledging that single-use-case SaaS companies face genuine displacement risk. Cisco's software assets — spanning security, observability, network infrastructure monitoring — are "not the single use case SaaS products, which I think are the early low-hanging fruit for AI to disrupt." He added that some modules from larger platform vendors can now be bypassed through AI tooling, but said Cisco's software is more tightly integrated with network infrastructure in ways that are difficult to replicate. Whether that assessment holds as AI models become more capable of understanding and managing network state is a question that will likely be revisited.
On AI-driven productivity internally, Robbins noted that Cisco has 27,000 daily users of its proprietary AI interface called Circuit, which automatically routes queries to the most effective model. He referenced an external estimate suggesting 75% of Cisco's code could be written by AI by the end of 2027, and flagged token economics — the cost of inference at scale — as an emerging operational consideration that will likely become a more prominent topic across the industry.
The Macro Framing: Bigger Than the Dot-Com Era, With Better-Funded Customers
Robbins was direct in situating the current environment relative to the late 1990s buildout: "It's bigger than anything we've seen, and it's happening faster." His key differentiation from the dot-com era is the financial profile of the companies doing the spending. "The companies that are spending most of the money or a large percentage of the money are very well funded, very successful companies" with strong balance sheets and demonstrated ability to navigate prior technology transitions. He acknowledged that misplaced capital will occur and some companies will fail, but argued the structural demand drivers — hyperscalers, enterprise AI readiness, sovereign clouds, neo-clouds, and public sector defense spending — are all moving simultaneously in a way that creates what he called a networking super cycle. On quantum, Robbins noted Cisco began investing in 2020 and has built a prototype networking chip for quantum interconnect, but placed practical timelines in the 20-to-30-year range, consistent with most technical consensus.
Cisco Systems Deep Dive
The Engine Room: Business Model and Economics
Cisco operates a hybrid business model that has fundamentally transitioned from its legacy as a pure-play hardware vendor into a software- and services-centric infrastructure provider. The defining catalyst for this transformation was the $28 billion acquisition of Splunk, which was completed in early 2024 and deeply integrated into the corporate architecture by 2026. Today, software and services account for more than half of the company's total revenue, structurally elevating margin profiles and providing immense revenue visibility. The core economic engine relies on selling foundational enterprise networking equipment, such as campus switches, data center routing platforms, and wireless access points, and layering these physical deployments with high-margin, recurring software subscriptions.
The company monetizes its ecosystem by acting as the unified connectivity and security backbone for modern digital enterprises. Cisco generates robust cash flows by cross-selling its proprietary networking hardware with software-defined networking overlays, cybersecurity tools like extended detection and response, and unified observability platforms. In this architecture, Splunk functions as the ultimate data operating system, breaking down silos by correlating telemetry across network performance, application health, and security domains. By integrating these previously disparate functions, Cisco commands a premium pricing model, leveraging its end-to-end stack to drive sticky, multi-year enterprise license agreements that smooth out the volatility traditionally associated with hardware upgrade cycles.
The Value Chain: Customers, Competitors, and Ecosystem
Cisco's customer base is exceptionally diversified, spanning small-to-medium businesses, the world's largest telecom service providers, governments, and cloud hyperscalers. The enterprise and public sector segments provide a steady, albeit cyclical, baseline of revenue. These organizations rely on Cisco for their everyday campus networking and security needs. In contrast, the hyperscaler segment, which includes technology giants like Meta, Microsoft, and Amazon Web Services, drives extreme volume but demands relentless performance and cost efficiency to run complex artificial intelligence and machine learning training clusters.
The competitive landscape is fiercely contested across all segments. In the core campus and enterprise networking arena, Cisco faces a newly strengthened Hewlett Packard Enterprise, which finalized its $14 billion acquisition of Juniper Networks in mid-2025 to bolster its AI-driven Mist networking platform. In the high-speed data center and artificial intelligence networking tier, Arista Networks serves as a formidable rival, utilizing a software-first approach that has captured significant market share among cloud builders. Nvidia has also aggressively entered the Ethernet switching space with its Spectrum-X platform, leveraging its graphics processing unit dominance to sell vertically integrated computing fabrics. On the supply side, Cisco competes against merchant silicon giant Broadcom, whose Tomahawk architecture powers many white-box and competitor switches. To counter this, Cisco designs its own chips but relies on third-party foundries, notably TSMC for its 3-nanometer processes, to manufacture its proprietary Silicon One architecture.
Market Position and Share Dynamics
Cisco retains an iron grip on the non-datacenter enterprise network, consistently capturing roughly 60 percent to 65 percent of the campus switching market. This dominance is driven by an ongoing multi-year campus refresh cycle and the deeply entrenched nature of its operating systems in corporate IT departments. However, the market share dynamics shift radically inside the data center. By early 2026, the overall global Ethernet switch market saw Arista and Cisco operating neck-and-neck at roughly 28 percent market share each, but the underlying composition of those figures tells a divergent story.
In the high-speed data center segment, where artificial intelligence workloads dictate capital expenditure, Nvidia and Arista have seized the leadership mantle. Nvidia's Spectrum-X platform captured an estimated 25.9 percent of the data center Ethernet switching segment in recent quarters, capitalizing on the tight coupling of its networking fabric with its computing clusters. Cisco's data center market share, hovering near 9 percent, reflects its historical struggles to penetrate the hyperscaler core. However, recent momentum indicates a shift. Following record artificial intelligence infrastructure orders of $1.9 billion in the third fiscal quarter of 2026, and a raised $9 billion fiscal 2026 artificial intelligence order target, Cisco is beginning to aggressively claw back relevance in the hyperscale data center, driven largely by its optical interconnects and custom silicon deployments.
The Competitive Moat: Silicon and Scale
Cisco's most potent competitive advantage lies in its sheer scale and incumbency. With over one million enterprise customers and an embedded base across 82,000 government organizations, the financial cost and operational risk of ripping out Cisco infrastructure is a massive deterrent for enterprises. This incumbency advantage is supported by an unparalleled global distribution and channel partner network that acts as a virtually insurmountable barrier to entry for smaller networking startups attempting to target the broad enterprise market.
Technologically, Cisco's moat is increasingly anchored by its in-house semiconductor design, specifically the Silicon One architecture. By designing its own networking silicon, Cisco bypasses reliance on Broadcom's merchant chips, giving it tighter control over supply chains, superior gross margins in the mid-60 percent range, and the ability to optimize power and latency characteristics for specific workloads. This vertical integration allows Cisco to offer a unified architecture that scales from top-of-rack data center switches down to campus routing, significantly reducing the operational complexity for network engineers. Furthermore, the integration of Splunk creates a unique structural advantage that hardware-only competitors lack, enabling unified observability and automated remediation across the entire technology stack.
Industry Dynamics: Opportunities and Threats
The transition from proprietary InfiniBand networks to open Ethernet standards in artificial intelligence data centers presents a massive structural opportunity. As the Ultra Ethernet Consortium pushes for standardized, low-latency Ethernet to connect computing clusters, Cisco is perfectly positioned to capitalize on hyperscalers seeking vendor diversity and open ecosystems. Concurrently, an aging post-pandemic networking infrastructure has triggered a multi-year campus refresh cycle, driving robust, above-trend demand for Cisco's traditional enterprise hardware.
However, the industry dynamics also present severe threats. Hyperscaler capital expenditures are highly concentrated and notoriously lumpy. A pause in infrastructure spending from just two major cloud providers could drastically impact Cisco's revenue trajectory in its highest-growth segment. Additionally, the rise of white-box networking, where hyperscalers purchase commoditized hardware and run their own custom operating systems, threatens to relegate legacy vendors to the role of mere component suppliers. There is also the persistent macroeconomic threat of component cost inflation, particularly in memory and optics, which can rapidly compress hardware gross margins if pricing power wanes in a competitive bidding environment.
Growth Catalysts: Next-Generation Innovations
Cisco's future growth hinges on its ability to execute within the high-speed networking fabric required for generative artificial intelligence. The recent launch of the Silicon One G300 chip, a 102.4 terabits-per-second Ethernet switch silicon, is a critical catalyst. Built to directly challenge Broadcom's Tomahawk 6, the G300 is designed to power gigawatt-scale data centers, optimizing job completion times by minimizing packet drops and latency in heavy computing environments. The ability to route data in milliseconds while delivering 33 percent better link utilization makes this a compelling alternative to proprietary fabrics.
Beyond the silicon layer, Cisco is pushing boundaries in cybersecurity and next-generation connectivity. The deployment of Hypershield, an AI-native, distributed security architecture, embeds security directly into the network data path rather than relying on perimeter firewalls. Furthermore, Cisco's unexpected foray into quantum networking, marked by the May 2026 launch of the Universal Quantum Switch, opens a speculative but potentially massive total addressable market. Engineered to route quantum information at room temperature over existing telecom infrastructure, this hardware positions Cisco at the forefront of post-quantum cryptography and next-generation secure communications.
Disruptive Forces and New Entrants
The networking industry is facing a wave of disruption not from scrappy startups, but from adjacent technology titans. Nvidia is the most consequential new entrant in the Ethernet switching space. Historically dominant in computing and InfiniBand, Nvidia's aggressive push into Ethernet with Spectrum-X represents a vertically integrated threat that bypasses traditional networking procurement entirely. By selling the processor, the data processing unit, and the network switch as a single optimized factory, Nvidia structurally locks out third-party networking vendors from the most lucrative tier of the market.
Additionally, the continued advancement of merchant silicon empowers original design manufacturers to build highly capable, low-cost switches. This commoditization force constantly pressures Cisco to prove that its proprietary hardware and software integration justifies a premium price tag. As hyperscalers increasingly opt to write their own networking software to run on these commodity switches, Cisco is forced to pivot its monetization strategy toward optical interconnects and standalone silicon sales to remain relevant in the largest cloud environments.
Leadership and Capital Allocation
Under the tenure of Chief Executive Officer Chuck Robbins, management has executed a methodical and often painful transformation from a traditional box-seller to a recurring revenue engine. The acquisition of Splunk was a high-stakes capital allocation decision that has definitively paid off, structurally elevating the company's gross margins and creating a software business of formidable scale. Management has consistently demonstrated the operational discipline required to navigate intense supply chain volatility while sustaining non-GAAP operating margins near 34 percent.
The leadership team has also demonstrated a willingness to make difficult operational decisions, evidenced by the mid-2026 restructuring that eliminated approximately 5 percent of the global workforce. This $1 billion reallocation of capital was clinically targeted to fund aggressive investments in silicon, optics, and artificial intelligence, rather than functioning as a mere cost-saving exercise. The track record reflects a disciplined executive team that commits over $9 billion annually to research and development, all while systematically returning the vast majority of its free cash flow to shareholders through consistent share repurchases and dividends.
The Scorecard
Cisco Systems stands at a critical juncture, having successfully navigated its transition from a legacy hardware vendor into a software-augmented infrastructure powerhouse. The integration of Splunk has definitively shifted the financial profile, delivering superior margin resilience and cementing the company's software and services revenues at over half of the total mix. While the enterprise campus segment remains a highly reliable cash engine driven by a robust corporate refresh cycle, the true inflection point for the thesis lies in the data center. Cisco's aggressive internal semiconductor development, highlighted by the 102.4 terabits-per-second Silicon One G300 architecture, demonstrates a clear, competitive capability to capture hyperscaler infrastructure dollars, evidenced by a steeply accelerating $9 billion order book for fiscal 2026.
Despite these technological strides, the competitive landscape is more hostile than ever before. Nvidia's vertical integration strategy with Spectrum-X and Arista's software-first agility present structural headwinds in the high-margin networking fabric, capping Cisco's immediate data center market share gains. Furthermore, the integration of Juniper Networks into Hewlett Packard Enterprise fortifies competition in Cisco's foundational campus territory. Ultimately, Cisco offers a highly durable, cash-generative business model with a credible call option on artificial intelligence infrastructure growth, making it a stabilizing anchor in a volatile infrastructure cycle rather than an undisputed apex predator.