We wrote about why Wi-Fi 6E and 7 quietly become a building project — the cable, the switch, the UPS that have to change. This is the math that made us write that article. The closet that powered Wi-Fi 5 with 30 watts per access point is now being asked to deliver 60 watts per AP, on more APs, through a switch that draws three times the power it used to. We've watched this play out enough times in real buildings now that the pattern is consistent: the wireless upgrade ships, and the UPS shuts down on the second brownout because nobody resized it. This article walks through the numbers — and the dual-PSU question that determines whether your network stays up when something fails.
The closet, before Wi-Fi 6E
Let's take a real, common configuration. A 24-port access switch sitting in a small IDF closet, supporting a single floor of an office or school building. The original deployment, sized for Wi-Fi 5 (802.11ac):
Total closet load at the receptacle: roughly 285 watts. Against a 1.5 kVA / 1,050 W rack UPS, that's about 27% load. Plenty of headroom. Battery runtime at 285 W on a typical 1.5 kVA unit is in the 25–35 minute range — comfortable for ride-through during a utility blip, plenty of time for a controlled shutdown if the outage extends.
This worked. For ten years in some buildings, it worked.
The closet, after the Wi-Fi 6E (or Wi-Fi 7) upgrade
Now the same closet, same square footage, same building, after the wireless team rolls out a Wi-Fi 6E or Wi-Fi 7 deployment. Here's what changes — and almost every line is a real change, not a marketing exaggeration:
Per-AP power numbers reflect Wi-Fi 6E/7 access points running 802.3bt (Type 3, up to 60 W PSE). A typical Wi-Fi 7 enterprise AP — Cisco Catalyst 9166, Meraki CW9166, Aruba AP-655, Ruckus R770 — pulls between 30 W and 55 W at sustained operation, with peak draws toward the upper end of the 802.3bt budget. [1][2] The 50 W average we used above is conservative — the real number can run higher in dense deployments.
The picture, as a load chart:
The 1.5 kVA UPS that comfortably carried the original deployment is now running at 86% load before redundancy is even considered, and goes into overload condition — meaning shutdown — the moment a second power supply is added to the switch. A dual-PSU configuration on the same UPS pushes you past 100% rated capacity, which the unit will not tolerate. This is the most common version of the failure we walk into: "the UPS shut down during a brownout last week and we don't know why."
The right-sized UPS for the new closet
The honest answer for a Wi-Fi 6E/7 IDF closet supporting 12 APs plus dual-PSU switch is somewhere in the 3–5 kVA range, depending on runtime requirements. Here's the math:
Why size to 60% rather than 80%? Three reasons. First, you want headroom for transient draws during AP startup, switch failover, and PoE negotiation events — these can pull 1.5×–2× steady state momentarily. Second, you want capacity for the next platform refresh (Wi-Fi 8 won't change the power profile much, but the next switch generation might add ports or features that draw more). Third, UPS efficiency is highest in the 30–70% load band; running at 80%+ pushes you into a less efficient operating region and accelerates battery wear.
This is the part that almost always surprises customers. The wireless RFP shows up with a $40K AP refresh. The cable upgrade adds $25K. The switch is another $20K. And then KAT-5 shows up and says "by the way, your $2,000 closet UPS needs to become a $5,500 closet UPS." It's the smallest line item in the project and the one people most want to skip — and it's also the one that, if you skip it, will take the entire deployment offline within months of go-live.
And then the heat. Where does it go?
Power consumed becomes heat dissipated. 1 watt of electrical draw equals about 3.41 BTU per hour of cooling load. That's not a KAT-5 number — it's a physical conversion factor. [3] The closet that was dissipating ~285 W = ~970 BTU/hr is now dissipating ~900 W = ~3,070 BTU/hr. That's a 3.2× increase in heat load in the same physical space.
For a 60-square-foot IDF closet, a 3,000 BTU/hr heat load is roughly the output of a small space heater running continuously. Closets that previously stayed within thermal limits through passive ventilation — an open door, a louvered panel, building HVAC drift — are now reliably running hot. Manufacturer specs for enterprise switches typically rate operating temperature ceiling around 113°F (45°C); above that, lifespan degrades quickly and thermal throttling kicks in. [4]
The fix is usually a small ductless mini-split or a closet-rated cooling unit — equipment in the 0.5–1 ton range, sized for the new heat load with margin. This is often a budget surprise for facilities teams, because nobody expected a wireless upgrade to require a cooling project.
Now the dual power supply question — and why it matters
Walk over to the back of any enterprise access switch — a Cisco Catalyst 9300, an Aruba CX 6300, a Juniper EX4400 — and you'll see two power supply slots. Sometimes both are populated. Sometimes only one. The reasons each switch has two slots, and the reasons your deployment should populate both, deserve clear explanation because almost nobody outside the network engineering discipline gets this right.
Reason 1: PoE budget, not just redundancy
For modern 802.3bt switches, two PSUs are often required to deliver full PoE budget, not just to provide redundancy. Cisco's own engineering documentation states it directly: a Catalyst 9300-48U with a single 1100W PSU delivers 560W of PoE budget. With a second 1100W PSU in combined mode, the budget rises to 1,440W. [5] That's a 2.6× increase in available power for downstream devices.
For a 48-port 802.3bt switch fully populated with Wi-Fi 6E or 7 access points (which want 60 W per port), the math demands it: 48 × 60 W = 2,880 W of PoE just for APs, plus the switch itself. Even two 1100W PSUs in combined mode can't deliver that — which is why Cisco StackPower and high-density UPOE+ deployments often need stacking power or dedicated power injection. [6] But for typical 24-port and lightly-populated 48-port deployments, dual PSUs in the same chassis are the right answer.
Reason 2: Failover redundancy
The other reason — and the one most people think of first — is that if one PSU fails, the other carries the load. PSUs do fail; they have moving parts (fans), capacitors that age, and they tend to be the first thing in a switch to die. A switch with two PSUs in active-active or active-standby mode survives a PSU failure without dropping a packet. A switch with one PSU goes dark.
For any network where downtime has business cost — and at this point that's nearly every enterprise network — dual PSUs with a real redundancy strategy are the right answer. The only exception is small branch offices or pilot deployments where the cost of downtime is genuinely low. For a school, a hospital wing, a manufacturing floor, or a busy office, the answer is dual PSUs every time.
Reason 3: A+B power feed
Here's the design that nearly nobody outside dedicated data center work gets right: the two PSUs in the switch should be fed from two independent power sources. Not two outlets on the same UPS. Not two UPS systems on the same circuit. Two genuinely independent paths — different UPS units, different circuits, different breakers, ideally even different power feeds into the building.
This is called A+B power (or "A-side / B-side"). It's the standard for any infrastructure where the network is critical, and it solves a class of failures that single-feed dual-PSU configurations don't:
- UPS failure. A UPS that fails — battery shorted, electronics fried, fan death — takes down everything plugged into it. If both PSUs are on the same UPS, both PSUs lose power simultaneously. The switch goes dark even though it has two PSUs.
- Circuit trip. A breaker that trips on the closet's primary circuit takes down the UPS feeding both PSUs. Same outcome.
- Single-receptacle damage. A failed PDU outlet, a damaged power cable, even a power strip going bad — any single physical point of failure removes both PSUs from service if they share that point.
With A+B power, each PSU is fed from a different UPS on a different circuit. Either one can fail completely and the switch keeps running. The cost of going from one UPS to two is real — you're roughly doubling the closet's UPS investment — but for any network where downtime has a measurable business cost, it's the right design.
How critical is the network, really?
This is the question we ask every customer before we recommend dual-UPS A+B for their closet, because the right answer isn't always yes. Network criticality lives on a spectrum, and the spend should match the spectrum:
- Mission-critical (always-on): Hospitals, public safety, manufacturing lines with continuous-process machinery, financial services, 911 dispatch, schools during instructional hours. Network down means people get hurt, money gets lost, or operations halt. Dual-UPS A+B is the standard. No exceptions worth making.
- Business-critical (operating-hours availability): Most office buildings, retail stores during business hours, professional services, distribution warehouses. Network down means productivity loss, missed transactions, customer impact. Dual-UPS A+B is recommended for the main MDF; single-UPS dual-PSU is acceptable for closets serving non-critical floors. The math on lost productivity vs. UPS spend usually favors A+B for the buildings where you'd notice.
- Convenience (best-effort availability): Small branch offices, pilot deployments, satellite spaces where work can shift to laptops or LTE if the wired network drops. Single-UPS, single-PSU acceptable. A few minutes of downtime is annoying but not consequential.
The honest test: if the network drops for 30 minutes during a normal business day, what does it cost? If the answer involves real dollars — patient delay, missed sales, idled staff, regulatory issues — your network is at least business-critical, and the closet should be designed for it. If the answer is "people would be annoyed", you're in convenience territory and don't need to overspend on redundancy.
One more piece of context before the criticality answer settles. Networks today carry more than they used to, and the trend is clear at industry scale. Cisco's State of Wireless 2026, surveying 6,098 wireless professionals across 30 markets, identified IoT and connected-device growth as the #1 driver of rising wireless dependence (63%), ahead of mobility, bandwidth-intensive applications, and AI workloads. [7] Sensors, cameras, badge readers, building controls, IP phones, smart-display panels — every one of them sits on a switch port and pulls PoE from the same closet now being asked to feed the new APs. The "is this network mission-critical" question gets harder to answer with a no the more of those devices end up downstream of a single switch.
The full closet redesign — what it looks like
Putting all of this together: here's what a properly-designed Wi-Fi 6E/7 closet looks like for a business-critical deployment. Compare to the original "1.5 kVA UPS, single switch, single PSU" configuration:
The capital delta — roughly $10,000 to $14,000 per closet — is real. For a building with one MDF and three IDFs, you're looking at a $40,000–$55,000 power and cooling project on top of whatever the wireless and switching upgrade cost. That number lands hard, and we don't sugarcoat it. What we do is help customers see the math behind it: the existing UPS is going to fail, and the question is whether it fails on KAT-5's schedule (planned upgrade, controlled cutover) or on the building's schedule (unplanned outage, weekend emergency, executive escalation).
The planning order — actually doing this work
If we were redesigning a Wi-Fi 6E/7 closet from scratch today, here's the sequence we'd follow. Each step depends on the one above it; doing them out of order is how projects ship broken.
- Confirm the AP count and PoE class. 802.3bt Type 3 (60W per port) is the planning baseline for Wi-Fi 6E and 7. [2] Don't undersize this.
- Calculate switch PSU requirements. Use manufacturer documentation — Cisco's StackPower white paper, the equivalent for your switch vendor — to figure out whether one PSU is sufficient or two are required for full PoE budget. [5] Don't guess.
- Sum total closet load. PoE delivered + PoE losses + switch system + non-network devices. Round up.
- Determine network criticality. Mission-critical → A+B dual UPS. Business-critical → A+B for major closets, single-UPS for minor. Convenience → single-UPS acceptable.
- Size the UPS(s). Target 60% load at steady state, including the dual-PSU configuration if applicable. Plan runtime to match what your facilities team can actually respond to.
- Verify circuit availability. A+B requires two genuinely independent circuits. If the closet only has one, the electrical scope expands.
- Calculate heat load. 3.41 BTU/hr per watt. Compare to existing cooling capacity. Plan for cooling upgrade if needed.
- Sequence the work. Electrical first (new circuits), then UPS install, then switch refresh, then AP cutover. Cooling can usually run in parallel.
- Cisco Meraki (2025). Wi-Fi 7 (802.11be) Technical Guide — power, cabling, MLO requirements. documentation.meraki.com
- Cisco Systems. Wi-Fi 7 and the Growing Future of Wireless Design Guide — explicit 802.3bt power planning recommendation. cisco.com
- 42U. Measuring Rack Cooling Requirements — BTU/hr conversion factors and rack-level thermal load methodology. 42u.com
- Cisco Systems. Catalyst 9300 Series Switches Hardware Installation Guide — operating temperature specifications and thermal management. cisco.com (HIG)
- Cisco Systems. Understanding Cisco StackPower White Paper — PSU configurations, PoE budget math, redundancy modes for the Catalyst 9300 family. cisco.com (white paper)
- Cisco Systems. Configure and Troubleshoot StackPower and XPS 2200 on Catalyst 9300 Switches — operational guidance for redundant power configurations. cisco.com (config guide)
- Cisco Systems (April 2026). State of Wireless 2026: Unlocking the Multiplier Effect — global survey of 6,098 wireless professionals across 30 markets. IoT identified as #1 driver of rising wireless dependence at 63%. cisco.com (State of Wireless 2026)