Cisco Takes AI Hyperscaler Orders to $9.3 Billion, but Margin Guidance Drops Shares 4.6% After Earnings

Cisco closed the fourth quarter of fiscal year 2026 with revenue of $17.3 billion, an 18% year-over-year increase, and non-GAAP earnings per share of $1.22, a 23% rise. Both figures exceeded the upper range of the company’s guidance. Net income soared 51% to $3.9 billion, while the Street consensus of $16.82 billion was missed.

Nevertheless, shares fell 4.6% in after-hours trading on August 12, to $118.18. The reason did not lie with the top line. It was in the projected gross margin, which management indicated was pressured by a heavier hardware mix.

The Number That Changed the Narrative

The data the company wanted to imprint on analysts was the AI orders. Cisco received $4 billion in infrastructure AI orders from hyperscaler clients in just the fourth quarter, bringing the total for the fiscal year to $9.3 billion. This is a jump of 4.5 times compared to the approximately $2 billion from the previous fiscal year.

CEO Chuck Robbins emphasized that the company is reducing positions in some areas and heavily investing in its own silicon, optics, security, and internal AI tools. This is a continuation of the restructuring announced in May, when Cisco cut 'under 4,000 positions' to redirect budget towards AI.

2027 Guidance Above Consensus, but with an Asterisk

The company projected revenue of $72.2 billion to $73.4 billion for fiscal year 2027, with earnings per share of $5.05 to $5.11. Both figures exceeded the Street consensus ($69.1 billion and $4.83). AI revenue for fiscal year 2027 is expected to rise 87.5% to $7.5 billion, and AI accounts for 37% of the anticipated growth in combined revenue.

The issue lies in the mix. When the hyperscaler order comes in heavily for customized hardware (AI-optimized switches, optical transceivers, proprietary silicon), the gross margin declines relative to the historical dominance of software and subscription revenue. Cisco did not hide the trade-off. Robbins described it as a strategic choice to capture the wave of infrastructure.

Market Reading

In the United States, the results double the pressure on AI switching competitors like Arista, Marvell, and Broadcom. If the hypothesis that AI network demand is growing faster than pure compute holds true, a thesis that Robbins himself reinforced publicly during the call, 2027 will become a central battleground among network fabric providers. This is where the competition for CIOs shifts from classic switches to the appropriate topology for training LLM.

In Europe, Cisco’s results come alongside the expansion of AI infrastructure under sovereign branding. Telecom operators and European governments have recently announced their own AI datacenters, with Cisco being the incumbent supplier in a significant portion of these contracts. The tightened margins indicate that the vendor is accepting relative lower gains per volume, a rare move for a company that has spent decades defending pricing power.

In India, where TCS, Infosys, Wipro, and HCL are setting up 'AI factories' for global clients, Cisco is the incumbent supplier. The increase in hyperscaler orders may create allocation tension: if a Microsoft datacenter in Virginia requests priority, an Infosys AI factory in Bangalore waits. This is the type of bottleneck that appeared in previous supply restructuring cycles and now reappears with a more explicit hierarchy.

In Brazil, the effect is sector-specific. Banks and telecoms that heavily purchased AI-optimized switches in their own LLM pilot programs are now entering a window where Cisco has less flexibility for aggressive discounts. The $9.3 billion hyperscaler portfolio consumed a significant portion of the vendor's production capacity in fiscal 2026, and the delivery schedule for Brazilian clients becomes a direct function of the global queue.

What the market has yet to digest is the long-term implication. Cisco is shifting its brand from recurring software vendor to critical infrastructure provider. Both generate high revenue, but attract very different enterprise value multiples. If the CFO can stabilize the gross margin mid-curve, the market will continue to pay for the hardware. If not, the company enters 'utility' mode and its multiples will decline.