Network hardware lead times stretch and prices rise in recurring cycles driven by component supply, demand spikes, and end-of-life refreshes. That is why buyers search for Cisco and Juniper switch lead times, though the sourced record shows these pressures are industry-wide, not vendor-specific. Network as a Service (NaaS) reframes the problem: the provider owns, procures, and refreshes the equipment fleet and delivers it as a monthly subscription, moving procurement, lead-time, inventory, and obsolescence risk from your balance sheet toward the provider's.
Supply for switches, routers, and access points tightens and loosens in cycles. The drivers below are structural, so lead-time and cost risk never fully goes away, whichever vendor you buy.
A finished switch depends on hundreds of parts, and a single scarce component can stall the whole build. Analysts have called this the golden-screw problem: a vendor can have 99 of 100 parts on hand and still be unable to ship, a description Dell'Oro Group used during the 2021-2022 shortage. Memory, optics, and printed circuit boards have each been named as pinch points in different cycles, so the constraint often sits upstream in the component layer that every vendor draws from, rather than in the switch chassis itself.
When demand jumps faster than manufacturers can add capacity, lead times lengthen across vendors. The current driver is AI infrastructure. In coverage reported May 2026, Gartner said vendors are pivoting engineering talent and physical inputs like memory and lasers toward more profitable AI data-center infrastructure, and projected that AI network fabric spending will surpass general-purpose data-center networking in 2026. General-purpose enterprise switches compete for the same components and factory time.
Enterprises typically refresh network switches on a roughly 5-to-7-year cycle, driven by performance needs, security updates, and the end of vendor support rather than physical failure, even though the hardware can often run 10+ years. Cisco's published lifecycle framework, for example, issues an End-of-Life announcement typically about a year before End-of-Sale, then generally provides roughly five more years to the Last Date of Support, though individual products vary. That is a largely vendor-set clock, and when it strikes you have to buy, whatever the market looks like that quarter.
The refresh clock and a supply-constrained window can overlap. If a product hits End-of-Sale or Last Date of Support at the same moment component lead times are stretching, an obligatory refresh can run into delivery delays. The 2021-2023 cycle illustrated this collision. You do not get to choose when a support deadline and a shortage line up, which is what makes owned-fleet timing risk hard to plan around.
Trade policy and component costs can flow into hardware prices. In the 2021-2022 cycle Cisco raised base hardware prices about 7% effective November 2021, citing component cost and availability, and by February 2022 experts estimated Cisco and its peers had raised prices roughly 10% across the board. More recently, on its May 2025 call Cisco described a mitigate-first approach to 2024-2025 tariff and trade-policy cost pressure, absorbing where possible before raising prices, with tariff impacts embedded in guidance. Cost pressure that lands on a vendor tends to reach the buyer next.
Buying and owning ties up capital in depreciating assets and, in tight markets, pushes buyers toward defensive habits. After the 2020-2023 disruptions an October 2022 SAP survey found 64% of companies shifting from just-in-time to just-in-case supply chains, holding buffer inventory and placing lifetime-buy advance orders to cover long lead times. Buffer stock protects against delay, but it means paying early for gear that sits unused, adding cost and warehouse risk on top of the purchase.
The 2021 to 2023 shortage was resolved by late 2023. A separate, AI-demand-driven tightening of components and data-center switches is the 2025 to 2026 story. Both are dated below so nothing here reads as a permanent, present-tense shortage.
| Period | What happened | Status |
|---|---|---|
| 2021-2023 | Pandemic-era component shortage drove record networking backlogs. Cisco's backlog reached roughly $14 billion in early 2022 (reported February 2022), up more than 150% year over year. Juniper exited 2022 with a backlog above $2 billion (stated January 2023). The crunch was industry-wide, with Arista citing semiconductor lead times of 40-60 weeks in mid-2021 and 50-70 weeks in early 2022. | Resolved Nov 2023. On its fiscal Q1 2024 earnings call (November 16, 2023) Cisco said ship times, lead times, and backlog had largely returned to normal. |
| Late 2023 | With lead times normalized, the constraint shifted from manufacturing to installation. Cisco described customers holding one to two quarters of shipped-but-uninstalled gear, which slowed new orders rather than any supply shortage. | Supply normalized as of late 2023; a deployment backlog, not a shortage. |
| 2024-2025 | Tariff-driven cost pressure. The U.S. finalized Section 301 tariff modifications with 2024 increases effective September 27, 2024 (semiconductors were left unchanged), and 2025 trade-policy changes followed. On its May 14, 2025 call Cisco described a mitigate-first approach, absorbing where possible before raising prices, with tariff impacts embedded in guidance. Vendors reported little customer stockpiling as of May 2025. | Cost input as of 2025, largely embedded in vendor guidance and partially passed through. |
| November 2025 | On its fiscal Q1 2026 call (November 13, 2025), Cisco described tariffs as a stable, embedded input and flagged a distinct component-cost pressure: tightening supply across memory, PCBs, and optics, with significant price increases on memory, factored into forward guidance. | Vendor commentary as of Nov 2025; component-cost pressure named as a forward-looking input. |
| Q1 2026 | An AI-driven memory crunch intensified, distinct from the 2020-2023 shortage. Per Counterpoint Research, DRAM prices rose roughly 80-90% in a single quarter as makers prioritized high-bandwidth memory for AI (reported February 2026). As of March 2026, Accuris tracking put overall semiconductor lead times near 40 weeks, with memory and fiber-optic components the most constrained, an upstream component constraint rather than a finished-switch famine. | As of early 2026, pressure sits in the component layer, per analyst tracking data. |
| 2026-2027 | Gartner forecast (reported May 6, 2026) projects enterprise data-center switch lead times stretching to 3-9 months, up from one to two months in mid-2025, alongside switch price increases of 15-40%, attributed to vendors pivoting resources toward AI data-center infrastructure. | Forecast. Gartner expects constraints to ease around mid-2027 but does not expect prices to fall back. |
Figures are as reported by the sources listed at the foot of this page. Constraints in each cycle were industry-wide across major vendors, not specific to any one.
Under Network as a Service, the provider owns and operates the equipment and you pay a monthly subscription. That moves procurement, lead-time, inventory, and refresh responsibility to the provider. See what is typically included in a NaaS subscription.
Under NaaS the provider owns, installs, and operationally manages the network equipment, and you pay a monthly subscription for turnkey service. IDC frames NaaS as a cloud-enabled, usage-based model that lets you acquire network capabilities without owning, building, or maintaining the infrastructure. Because the provider owns the fleet, the procurement and lead-time exposure that otherwise falls on a hardware buyer shifts toward the provider. How any specific provider manages that exposure is positioning to verify with them directly, not a guarantee.
Buying and owning ties up capital in depreciating hardware and, in tight markets, in just-in-case buffer inventory. A consumption model converts that into a predictable operating expense, so you are not pre-funding gear that may sit in a warehouse waiting on a missing component or an install crew. This is a cash-flow and balance-sheet shift, not automatically a lower total cost, a tradeoff covered below.
The end-of-life clock, End-of-Sale and Last Date of Support, is a vendor-set forcing function behind re-buys. In a provider-owned model the refresh clock sits with the provider. NaaS is also documented to facilitate more frequent refresh and access to newer infrastructure, such as higher-speed Ethernet, rather than leaving you locked into an owned and depreciated fleet whose support window is running out.
Documented 2024-2026 cost pressure arrived in two waves: tariffs first, then an AI-driven component crunch in memory, PCBs, and optics that Cisco named explicitly in late 2025. In a provider-owned model, exposure to both waves concentrates with the provider rather than hitting your capital budget quote by quote. That is a reallocation of risk, and pricing and terms still need to be compared, since providers price that risk into the subscription.
| On this dimension | Buying and owning hardware | Network as a Service |
|---|---|---|
| Who carries lead-time risk | You do. If lead times stretch, your project waits, and you absorb the delay directly. | The provider does, as fleet owner. Your service commitment is with them, not with a shipment date. |
| Capital exposure | Upfront CapEx in depreciating hardware, often plus just-in-case buffer inventory in tight markets. | Predictable monthly OpEx; no upfront hardware outlay, though total cost over time must be compared. |
| Refresh and end-of-life clock | You track End-of-Sale and Last Date of Support and fund each refresh on the vendor's schedule. | The provider owns the refresh clock and, per analysts, can enable more frequent upgrades to newer gear. |
| Price-increase exposure | You feel tariff and component cost pass-through at each purchase, whatever the market that quarter. | Exposure concentrates with the provider, who prices it into the subscription and contract terms. |
| Buffer / just-in-case inventory | You may hold buffer stock and place lifetime-buy orders, paying early for gear that sits unused. | The provider manages fleet inventory; you consume service rather than warehousing spare hardware. |
| Time to turn up a new site | Gated by your own procurement, delivery, and install cycle for each new location. | Analysts cite faster time-to-value for new sites as a core NaaS driver, subject to provider capacity. |
Which model wins depends on your footprint, cash position, and how much lead-time risk you can absorb. A vendor-neutral evaluation framework walks through the tradeoffs.
NaaS transfers supply and lifecycle risk, but it is not automatically cheaper or automatically the right fit, and treating it that way would be misleading. An honest comparison weighs the risk transfer against real, documented tradeoffs before you decide.
Weigh these against the supply-risk relief. The NaaS pricing guide covers how OpEx and CapEx actually compare over a contract term.
Before you sign a refresh purchase order, run it through these questions. They tell you whether buying, waiting, or switching to a subscription is the better move right now.
Want the full version? See the NaaS evaluation checklist.
If you are staring at a hardware quote with a long lead time or a steep price increase, that is the moment to check whether a provider-owned model fits better than buying and owning. As an independent, vendor-neutral advisory, we help you compare providers, estimate pricing, and pressure-test the tradeoffs against your own refresh horizon, so the decision is measured rather than forced by the quarter's supply market.
Every dated figure on this page traces to public reporting or vendor disclosures. Supply conditions change, so treat forecasts as forecasts and check the original source for the latest.
An independent advisor competes NaaS providers that own the hardware, procurement, and refresh on your exact footprint, at no cost to you.