Cisco Cloud Control launches — the software test under the AI boom

Today Cisco turned on Cisco Cloud Control, a new operations platform, for U.S. customers: more than 300 companies already onboarded and another 5,000-plus waiting before general availability. In plain terms, it is one console where a company's IT staff and its AI agents watch over everything Cisco sells at once — the network, the firewalls, the servers, the observability logs — from a single login, with agents allowed to diagnose and fix problems while humans approve the changes. Two weeks earlier, Cisco reported the strongest revenue year in its 40-plus-year history, and then the stock took what reports called its sharpest one-day repricing in years. The launch is Cisco's answer to the question that selloff raised: can the AI boom be turned into software revenue that recurs?

Start with the quarter that punctured the longer-term momentum. Fiscal 2026 — the year ending late July — brought revenue of $63.3 billion, up 12%, with the fourth quarter at about $17.3 billion, up 18% and above the high end of the company's own guidance. Fourth-quarter product orders rose 35% year over year. The engine is AI construction: every machine-learning cluster needs racks of servers wired together at speed, and Cisco's Silicon One chips and Acacia optics sit inside that fabric. AI-infrastructure orders for fiscal 2026 reached $9.3 billion, roughly 4.5 times the prior year, with about $5 billion of booked demand still waiting to convert to revenue. Cisco expects AI-infrastructure revenue to grow to about $7.5 billion in fiscal 2027.

The boom has a price, and it shows up where a hardware company always pays it. On a non-GAAP basis, product gross margin fell about 270 basis points in fiscal 2026 to roughly 65%, because growth skewed toward lower-margin gear and memory costs rose. Cisco's early fiscal-2027 guidance calls for gross margin around 65–66% — in other words, more compression — plus slower growth in the second half as easy comparisons end.

The record is real; it is just being set by the least profitable part of the business.

That is the uncomfortable context for Cloud Control, and why it deserves more than keynote nostalgia. The durable prize was never the switch; it is the subscription that runs on the switch. Cloud Control is that bet: tiered licenses (Essentials, Advantage, Premium) plus metered "AI token" packs, sold on top of an installed base, with the $28 billion Splunk analytics engine Cisco acquired in 2024 serving as the data layer underneath.

None of that is vapor, and part of it is genuinely new. A hundred small facts in the launch point one direction — 300 onboarded customers, a waitlist of more than 5,000 sign-ups, integrations already live with ServiceNow and Atlassian, a builder for custom agents. But the honest reading is that the core is early and much of it is repackaging: Cloud Control absorbs tools Cisco already sold separately, from Meraki and Catalyst Center to ThousandEyes and Security Cloud Control, reorganized under one agent-flavored console. Cisco's showcase stories — a supply-chain customer diagnosing a site issue in under 10 minutes instead of over 10 hours — are vendor-supplied testimonials, not audited benchmarks, and the same "agentic ops" category is being marketed simultaneously by Microsoft, Amazon, Google, Dell, and IBM. Analysts tracking the launch already flagged the pricing traps: license "tier creep" as heavy use pushes customers up, and ambiguity over how metered tokens bill when third-party agents touch Cisco data. No price list came with the announcement.

Now the number that should discipline all of that enthusiasm: annual recurring revenue. Cisco's ARR — the forward-looking gauge of its software business — stood at $32.1 billion in the fourth quarter, up just 3% from a year earlier. Even during the best hardware year in the company's history, the recurring-software engine barely grew. Software revenue itself reached $6.2 billion in the quarter, up 11%, but ARR is where durability shows, and it is not accelerating.

That is also why the market's cold feet after a record quarter make sense. The stock is the story of a massive repricing: from about $66 at its low to a $130 peak, and back to about $111 today after the selloff. At roughly 22 times the midpoint of Cisco's own fiscal-2027 non-GAAP earnings guidance, and about 25 times trailing EV/EBITDA — richer than Microsoft's roughly 19 times — investors are now paying growth-company multiples for a company whose record was built on hardware orders that will eventually lap themselves.

What would change the picture is specific and observable: ARR growth accelerating well past 3%, a gross margin that stops compressing as software takes a bigger share, and the $5 billion AI backlog converting to revenue on schedule. What would end it is equally specific: a quarter when AI order growth decelerates and Cloud Control looks like a renamed menu of existing tools — features, not new revenue. Nothing in today's launch tells you which outcome is closer, because the launch itself disclosed no revenue expectations at all. The reasonable stance is to watch the software-company numbers, not the console.