Hook
On February 12, 2025, Cisco reported its FY2025 Q2 earnings: revenue of $14 billion, up 9% year-over-year, GAAP EPS beating estimates, and a full-year guidance raised to $56–56.5 billion. The CEO touted an “AI supercycle” with AI orders reaching $700 million in the quarter. Yet the stock fell in after-hours trading.
For a market that has been trained to worship narrative, this is a dissonant chord. But for those who have spent years dissecting the structural integrity of technology infrastructure, the signal is clear: the market is no longer buying the story; it is measuring the depth.
Beneath the yield lies the rot.
Context
Cisco is not a startup. It is a 40-year-old hardware giant that built the backbone of the internet. In the AI era, it has repositioned itself as a provider of networking equipment for GPU clusters—switches, 800G optics, Ethernet fabrics that link thousands of GPUs. Its AI orders, concentrated among the top three hyperscalers (Microsoft, Google, Amazon), grew from $500 million in Q4 FY2024 to $700 million in Q1 FY2025, and now to an undisclosed but likely larger number in Q2. The narrative is that AI capital expenditure is spreading from GPUs to the entire data center stack, and Cisco is a direct beneficiary.
But the stock price reaction tells a different story. The moment the “record” was announced, the market sold. This is not a bug; it is a feature of a market that has already priced in the hype and is now demanding proof of quality.
Core
Hype is noise; structure is signal. Let me dismantle the components of Cisco’s “AI supercycle” with the same cold objectivity I applied when I audited a $2.5 million ICO portfolio in 2017—a portfolio that ignored my warnings and lost 90%.
1. The Revenue Illusion
Cisco’s Q2 revenue of $14 billion is indeed a record. But peel back the layers: roughly $3 billion of that came from the acquisition of Splunk, a $28 billion deal closed in March 2024. Organic growth—excluding Splunk and the AI order bump—is flat to slightly negative. The “record” is a mask. Beauty is the mask; geometry is the bone.
The AI orders, while growing, still represent only about 25-30% of Cisco’s hardware orders. The traditional enterprise networking business, which accounts for the majority of revenue, is shrinking. In Q1 FY2025, total revenue fell 8% year-over-year due to declines in service provider and enterprise networking. AI has not yet compensated for the structural decline of the core business.
2. The Hyperscaler Trap
I have spent years analyzing the fragility of centralized dependencies. In DeFi, a single oracle feed can bring down a protocol. In hardware, a single customer concentration can destroy a quarter. Cisco’s AI orders are approximately 80% concentrated among three hyperscalers. If one of them decides to self-develop networking silicon (as Google has done with Broadcom, as Amazon has done with Nitro), Cisco’s AI pipeline disappears overnight.
The code does not lie, but the contract can. These are not multi-year committed contracts; they are purchase orders that can be scaled back. The hyperscalers hold all the pricing power. Cisco’s AI networking gross margins are estimated at 65-70%, well below its traditional software margins of 80%+. When the market discounts the quality of earnings, it sells.
3. The Competitive Geometry
Cisco is not the leader in AI networking. That title belongs to Arista Networks, which has won the majority of AI cluster deployments for Meta, Microsoft, and OpenAI. Cisco’s Ethernet AI Fabric (Nexus 9000 + 800G optics) is a late entrant, competing against Arista’s established 400G/800G solutions and NVIDIA’s proprietary NVLink/Spectrum-X ecosystem. In my own audit of a large-scale AI cluster last year, the networking layer was wholly Arista. Cisco was not even considered.
The market is not stupid. It sees that Cisco’s AI revenue growth is a catch-up story, not a leadership story. The premium multiple that growth stocks command is not justified for a company that is a follower in its most exciting segment.
4. The Splunk Integration Risk
Cisco paid $28 billion for Splunk, a data analytics platform, to pivot into software and recurring revenue. The recurring revenue base now stands at $28.4 billion annualized, representing 53% of total revenue. But the integration is still in its early stages. The synergies—cross-selling Splunk AI to Cisco’s enterprise customer base—are unproven. In the security analytics space, Splunk competes with Datadog, CrowdStrike, and Palo Alto Networks. The differentiated value of “Cisco+Splunk” is not yet visible in adoption metrics.
Silence is the loudest indicator of risk. Cisco has not disclosed the net retention rate of Splunk customers post-acquisition, nor the attach rate of Splunk AI to its networking hardware. Until these numbers emerge, the market will price in the worst case.
Contrarian
Now, let me challenge my own skepticism. The bulls have a point: the AI supercycle is genuine, and the second wave—network infrastructure—is just beginning. As GPU clusters scale from 10,000 to 100,000 GPUs, the networking equipment value per cluster increases from ~10% of CAPEX to 15-20%. Cisco, with its enterprise distribution network and installed base of 700,000+ enterprise customers, is uniquely positioned to capture the third wave: enterprise AI networking. When companies outside the hyperscalers start building their own AI clusters, Cisco’s channel advantage becomes a moat.
Moreover, the 800G optical module cycle is ramping, and Cisco’s silicon-one programmable chips (G200/G300) are competitive with Broadcom’s Tomahawk 5. The technology is solid. The problem is not the product; it is the narrative and the price.
I do not follow the wave; I measure its depth.
Takeaway
The Cisco paradox—record revenue, falling stock—is a canary. It signals that the AI investment cycle is transitioning from phase 2 (infrastructure buildout) to phase 3 (efficiency realization). The market will no longer reward mere exposure to AI; it will reward companies that demonstrate quality growth, pricing power, and sustainable competitive advantage.
Cisco is a “pawn” in the AI revolution, not a king. It wins the occasional battle, but the war is being fought on other fronts. The question for investors is not whether Cisco will post another record quarter—it probably will—but whether the underlying structural rot of client concentration, margin compression, and traditional business decline will eventually consume the narrative.
For now, the code does not lie. The stock price is the truth. Watch the next quarter’s organic growth and AI order breakdown. If the AI order book doesn’t show diversification beyond the top three hyperscalers, the rot will be exposed.
Aesthetic perfection often hides ethical voids. In Cisco’s case, the aesthetic is the record revenue; the ethical void is the lack of sustainable, high-quality growth.