Cisco’s AI Infrastructure Bet: Where Should It Focus Next?
Download case PDF (5 pages)Request teacher’s pack
A strategy case for MBA students and working professionals
Decision date: October 3, 2026
Information cutoff: October 3, 2026, 11:59 p.m. Pacific Time
Case status: Constructed teaching case. Cisco did not disclose the specific resource-allocation decision or internal deliberations posed here.
A portfolio of AI offers
In late summer 2026, Cisco broadened its AI infrastructure portfolio along two fronts. It announced a rack-scale offer that adds Supermicro compute to its Secure AI Factory with NVIDIA. A few weeks later, it announced an AI POD for Splunk that brings self-managed AI capabilities to customers running on-premises, private-cloud, and air-gapped environments. These launches sat alongside Cisco’s networking, security, observability, and software businesses.
The offers address different buyers and operating needs. A neocloud or sovereign-cloud provider may want dense compute, high-speed networking, cooling, and a validated system. A regulated enterprise may want to run AI against sensitive data without moving it to a public cloud. Another customer may already have a preferred server or accelerator stack and want Cisco networking, security, or observability to work across it.
Cisco had already announced these products and partnerships. The classroom decision is how to prioritize the next wave of commercialization and integration effort: which customer problem should receive the clearest focus, how tightly should Cisco package the system, and what evidence should guide further investment? The public record does not disclose a specific FY27 allocation, AI-system margin, or customer-level adoption results.
A strong year, with different signals inside it
Cisco reported fiscal 2026 revenue of $63.3 billion, up 12% from the prior year. Product revenue rose 16% to $48.3 billion, while services revenue was essentially flat at $15.0 billion. GAAP operating income was $15.4 billion, or 24.3% of revenue. Cash provided by operations was $14.2 billion. Total software revenue across Cisco’s product areas and services was $23.2 billion, up 4%; total subscription revenue increased 1%. These are company-wide measures, not results attributable to AI products or Splunk AI. [C1, pp. 31–33; C2]
Networking was the fastest-growing major product category. Revenue reached $34.7 billion, up 22%, an increase of $6.4 billion. Cisco attributed most of the increase to AI infrastructure—including Silicon One-based systems and optics—and also cited campus networking. Security revenue rose 2% to $8.2 billion; Collaboration rose 4% to $4.3 billion; and Observability rose 4% to $1.1 billion. Cisco does not disclose a separate operating margin for AI infrastructure or these product categories. [C1, pp. 37–38]
Exhibit 1. Cisco FY2026 company and product-category results
Fiscal year ended July 25, 2026; dollars in billions. Source: Cisco FY2026 Form 10-K. Growth rates are reported by Cisco.
| Measure | FY2026 | Change vs. FY2025 |
|---|---|---|
| Total revenue | $63.3 | +12% |
| Product revenue | $48.3 | +16% |
| Services revenue | $15.0 | Flat |
| GAAP operating income | $15.4 | +31% |
| Networking product category | $34.7 | +22% |
| Security product category | $8.2 | +2% |
| Collaboration product category | $4.3 | +4% |
| Observability product category | $1.1 | +4% |
Product categories are not reportable operating segments. They do not reveal category-level or AI-specific profit.
Orders, revenue, and expectations
At its August 12 earnings release, Cisco reported $9.3 billion of AI infrastructure orders from hyperscalers in FY2026, including $4 billion in the fourth quarter. It reported approximately $4 billion of AI infrastructure revenue for FY2026 and expected $7.5 billion for FY2027. Separately, its Form 10-K said hyperscaler AI infrastructure demand represented approximately 6% of FY2026 revenue, compared with less than 2% a year earlier. Cisco expected this demand to remain a significant driver in FY2027. [C1, p. 32; C2]
Orders, recognized revenue, and guidance describe different things. The $9.3 billion order figure is not a disclosed AI-specific backlog balance, and it should not be treated as revenue already earned. The $7.5 billion figure is management’s expectation, not an outcome. Cisco’s Q4 total product orders grew 35% year over year and 25% excluding hyperscalers; Networking product orders rose 40%. The broader order trend suggests demand beyond hyperscalers, but order growth alone does not establish delivery, margin, or persistence. [C2]
Supply decisions carry both upside and exposure
Cisco increased the supply it secured as demand and lead times grew. At July 25, 2026, inventory was $5.7 billion, compared with $3.2 billion a year earlier. Supplier purchase commitments were $17.2 billion, versus $7.6 billion. Inventory deposits and prepayments were $1.1 billion, compared with $0.8 billion. Cisco said inventory plus commitments increased 112%, primarily because of commitments related to Silicon One and other products for hyperscalers and other customers. It also cited higher memory prices and arrangements to secure supply and pricing. The figures cover more than AI alone. [C1, pp. 45–46]
Some supply agreements allow changes before firm orders; a significant portion of reported commitments is firm, noncancelable, and unconditional. Some components are specific to a customer design and have limited alternative use, while some memory can be used across Cisco’s portfolio. Cisco said its inventory and commitments were appropriate for its current and expected demand. Its filing also warned that lower or delayed customer demand, changing architectures, or excess quantities could result in lower margins, excess or obsolete inventory, and commitment losses. The same supply position can protect customer lead times if demand arrives as expected and increase exposure if forecasts change. [C1, pp. 45–46; Note 13]
Exhibit 2. Selected supply-chain balances
At fiscal year-end; dollars in billions. Source: Cisco FY2026 Form 10-K.
| Balance | July 25, 2026 | July 26, 2025 | Change |
|---|---|---|---|
| Inventory | $5.7 | $3.2 | +80% |
| Supplier purchase commitments | $17.2 | $7.6 | +126% |
| Inventory deposits and prepayments | $1.1 | $0.8 | +34% |
Cisco reported the combined increase in inventory and commitments as 112%, primarily related to Cisco Silicon One and other products for hyperscalers and other customers. Not all balances are AI-specific; deposits and prepayments are shown separately and are not included in the commitment total.
Three paths to commercial focus
The following alternatives are classroom constructs based on public announcements, not options Cisco said it had formally debated. Each could be paired with capabilities useful across more than one customer segment.
- Lead with rack-scale Secure AI Factory. Concentrate go-to-market and integration resources on neocloud and sovereign-cloud customers that want a validated system combining compute, networking, security, and services. Cisco announced Supermicro air- and liquid-cooled systems within its NVIDIA-based architecture and said it would begin offering them in October. The integrated offer may simplify purchasing and technical coordination; it also raises execution, partner-dependence, and supply exposure. The announcement’s claims about faster deployment and lower risk had not been independently measured in the public record by the cutoff. [C3]
- Lead with self-managed Splunk AI. Focus on enterprises that need AI near sensitive machine data, including on-premises, private-cloud, or air-gapped settings. Cisco announced AI POD for Splunk, agent observability and token-cost tracking, and security developments including a multi-year AWS co-development agreement. This path may connect Cisco’s software and security assets to the customer’s existing environment. Paid adoption, incremental revenue, customer cost savings, and security outcomes were not disclosed. Some deployments may use customer or partner infrastructure rather than Cisco compute. [C4]
- Lead with modular networking, security, and observability. Sell core Cisco capabilities into customer-selected compute and AI stacks, while investing in reference designs and partner interoperability. This can preserve customer choice and reduce the need to supply every system component. It may also leave Cisco with less control of deployment quality, fewer attached services, and a smaller share of an integrated-system purchase.
The competitive field supports more than one architecture. Arista reported $3.0 billion in Q2 2026 revenue, up 37.7% year over year, and announced 1.6-terabit rack-scale AI fabric products with air- and liquid-cooled configurations. Dell described its AI Factory with NVIDIA as an integrated infrastructure, services, and partner offering and said more than 5,000 customers were deploying it. Those financial, product, and customer-count statements come from the competitors themselves; they are not directly comparable to Cisco’s fiscal results, orders, or AI infrastructure revenue. [R1–R3]
The next wave
On October 3, Cisco’s strategy team must decide how to focus commercialization and integration work for FY2027. The recently announced products create several possible paths, but the information needed to compare their economics remains incomplete. Public disclosures do not show AI-specific product margins, order cancellation terms, customer-by-customer concentration, deployment effort, workloads in production, Splunk AI paid adoption, or customer return on investment.
The team can make a directional choice and define what it needs to learn. It must distinguish interest from orders, orders from revenue, and product availability from deployment outcomes. It must also decide which partners own compute, software, validation, implementation, support, and the ongoing customer relationship. A clear sequence could concentrate scarce attention; a broader approach could preserve options across segments. Either choice carries the risk of spreading resources too thin or committing ahead of customer readiness.
Discussion questions
- Which customer segment and use case should receive Cisco’s primary FY2027 commercialization focus? What value is Cisco positioned to capture there?
- Compare the rack-scale, self-managed Splunk AI, and modular paths. What must Cisco own, integrate, or govern in each, and where is partner coordination essential?
- What do Cisco’s order, revenue, and guidance figures establish? What do they leave uncertain about future demand and economics?
- How should the team treat the rise in inventory and supplier commitments? Identify conditions under which the same position is an advantage or a liability.
- Recommend a sequence of investment and a small set of measures that would support expansion, adjustment, or pause. State the assumptions that would change your recommendation.
Source note
The case uses Cisco’s FY2026 Form 10-K filed September 2, 2026 (fiscal year ended July 25, 2026), its August 12 earnings release, its August 25 and September 15 product announcements, and rival disclosures available by October 3, 2026. Full citations and locators are in the source register. The case excludes information published after the cutoff. Financial cases and Excel analysis are deferred.