Power blips happen every day — most of them so brief you never notice. But your server notices. Here's what a UPS actually does, what it costs, and why it's one of the best bets in IT for the money.
A few years ago we got a call from a manufacturing company on a Monday morning. The weekend shift had experienced a brief power blip — maybe two seconds, probably from a passing thunderstorm. Nobody thought much of it. When the IT person came in Monday and checked the server, the database that ran their entire order management system had a corrupted index. It wouldn't start.
They spent three days recovering. A consultant was brought in. Some data had to be manually re-entered. Total damage, between downtime, the consultant, and lost productivity: somewhere north of $40,000. The server had been plugged directly into the wall.
The fix that would have prevented all of it? A UPS — an uninterruptible power supply. Cost: about $300.
I tell this story not to scare you, but because it's one of the most avoidable disasters we see. After 30 years of managing small business infrastructure, power protection is still the item most often missing from server rooms. People spend tens of thousands on a server and then plug it into a $15 power strip. It doesn't make sense, but it's incredibly common.
So let's fix that. Here's everything you need to know about UPS devices — what they do, why they matter, and how to size one for your setup.
Most people think of a UPS as a backup battery for when the power goes out. That's part of it — but honestly, the battery is almost the least important feature for most small businesses.
What a UPS does on a minute-to-minute basis is much more valuable:
Your wall outlet isn't delivering perfect, clean electricity. Voltage sags, voltage spikes, and electrical noise from motors, HVAC systems, and other equipment all travel through your building's wiring. Over time, this "dirty power" degrades electronics. Hard drives are especially sensitive — voltage spikes can cause read/write errors, bad sectors, and premature failure. A UPS conditions the power before it reaches your equipment, delivering clean, stable electricity regardless of what the building's wiring is doing.
A surge protector will protect against a large spike, but most of them do nothing about voltage sags — brief dips where the voltage drops but power doesn't cut out entirely. These are actually more common and more damaging than full outages. A server during a sag is like a car engine being starved of fuel: it keeps running, but something inside is straining. UPS devices regulate voltage in real time so your equipment never sees those dips.
When power does go out, the UPS switches to battery in milliseconds — faster than your equipment can detect the interruption. For a server, this means no sudden shutoff, no file system corruption, no half-written database transaction left dangling. You get minutes (sometimes 10-30 depending on the model and load) to either restore power or shut down cleanly.
Better UPS units connect to your server via USB or network and run software that monitors battery level. If power hasn't come back after a few minutes, the software triggers a clean, graceful shutdown — without anyone needing to be in the building. This is huge for businesses without 24/7 staff. Your server shuts down safely at 2am during a storm, and when power returns, it boots back up normally.
Let's be specific, because this is where the $50,000 number comes from.
Modern databases write data in transactions. A transaction might involve updating three tables in a specific sequence. If power dies halfway through that sequence, the database is left in an inconsistent state — some records updated, some not. Recovery can range from "restart and let it roll back automatically" to "hire a database expert and spend days manually reconciling data." SQL Server, QuickBooks databases, point-of-sale systems, ERP software — all of these are vulnerable. This is the most common and costly failure mode we see.
Windows and Linux both maintain metadata about files and the file system. An abrupt shutdown can leave this metadata in an inconsistent state. Usually the OS detects this on the next boot and runs a file system check. Sometimes it fixes itself. Sometimes the server won't boot at all without intervention. Best case, it costs you an hour. Worst case, you're rebuilding a server from scratch.
Traditional spinning hard drives have read/write heads that "park" safely when powered down normally. An abrupt power cut can leave those heads in a position where they contact the disk platter — called a head crash. With SSDs this is less of a mechanical concern, but sudden power loss during a write can still cause controller or firmware corruption. Either way, you're potentially looking at data recovery costs that start around $1,000 and can go much higher for RAID arrays.
Even if each individual event doesn't cause visible damage, repeated voltage sags and dirty power cycles shorten the lifespan of your equipment. We've seen servers fail years before they should have because they were plugged directly into a building with industrial equipment running on the same electrical circuit. A UPS essentially extends the working life of everything connected to it.
Walk into any electronics store and you'll see UPS units from $50 to $5,000. Here's how to make sense of it.
These are your basic units — the ones that look like a fancy power strip with a battery. They pass power through directly and only switch to battery when they detect an outage. The switch takes 20-30 milliseconds, which is long enough that some older equipment notices. They offer minimal power conditioning. These are fine for desktop computers and home setups, but we don't recommend them for servers.
This is the sweet spot for most small businesses. Line-interactive units actively regulate voltage in real time, handling sags and surges without switching to battery. The switchover time on a full outage is very fast (typically 2-4 milliseconds, which no modern server will notice). They're efficient, they last, and they give you real power conditioning. APC and CyberPower both make solid units in this range. For a single server, this is almost always what we recommend.
These run your equipment off battery power all the time, constantly recharging from the wall. There is literally zero switchover time because you're never running off wall power directly. Complete power isolation. These are overkill for most small businesses but worth considering if you have sensitive medical or industrial equipment, or if you're in an area with chronically bad power quality. If your lights flicker regularly, take a closer look at this tier.
For the vast majority of small businesses with one or two servers, a line-interactive UPS in the $200-$400 range is exactly right.
UPS capacity is rated in VA (volt-amperes) and watts. The rule of thumb: you need a UPS rated for at least 1.5 times the total wattage of everything connected to it, to give the batteries room to breathe.
Here's a quick reference for common equipment:
Add up what you're going to plug in, multiply by 1.5, and that's your minimum wattage. Then find a UPS at that wattage (or above) in the line-interactive category.
For a typical small business server room with one server, a NAS, a switch, and a firewall — you're probably looking at 600-900 watts of actual load. A 1500VA / 900W unit gives you comfortable headroom and usually provides 10-20 minutes of runtime at that load. That's plenty of time for the software to initiate a graceful shutdown, or for someone to restart a generator.
One important note: connect only your server, NAS, and network gear to the battery-protected outlets on the UPS. Don't plug monitors, printers, or other non-critical equipment into the battery side — they drain the runtime and don't need it. Most UPS units have a mix of battery-protected and surge-only outlets for exactly this reason.
If your firewall and switch go down with the server, you lose remote access — meaning you can't VPN in to check on things or trigger the shutdown remotely. Protecting your core network equipment on the same UPS as the server is worth the extra outlet. It also means your VoIP phones and Wi-Fi stay up during a brief outage, which matters if your team is in the middle of something.
Buying the right UPS is step one. Setting it up correctly is step two.
Most quality UPS units ship with or offer free software (APC's PowerChute, CyberPower's PowerPanel, and NUT for Linux are the common options). Install it on your server, connect the UPS via USB, and configure it to shut down the server automatically if power hasn't returned within a set time (usually 5-10 minutes). This is the piece most people skip, and it's arguably the most important. Without it, when power goes out at 2am, nobody is there to shut the server down, and it runs until the battery dies — which is the same as having no UPS at all.
Once everything is set up, test it. Plug the UPS in, confirm the server is running, and briefly pull the power cord from the wall. Watch the server continue to run. Verify the software sees the battery status and would trigger a shutdown if needed. Then plug the wall power back in. This takes ten minutes and tells you definitively whether your protection is actually working. For more on testing your critical systems, see our post on backup testing — the same philosophy applies here.
UPS batteries last 3-5 years. After that, they hold less and less charge until they're effectively useless — but the UPS will still look like it's working. Most units have an indicator light for battery health, and the monitoring software will also flag aging batteries. When the battery is due, either replace it (replacement batteries are usually $30-$80) or replace the whole unit if it's already old. A UPS with a dead battery is just a fancy power strip — no protection.
If you add equipment over time — another drive, a new server, a larger switch — recalculate your total load and make sure the UPS can still handle it. It's easy to accidentally overload a unit you sized correctly three years ago.
A decent line-interactive UPS for a small business server setup: $200-$400. Installation: an hour. Battery replacement every 3-4 years: $50-$80.
Total cost over five years: maybe $500.
Now consider the other side. A single incident of database corruption serious enough to require professional recovery: $5,000-$25,000 in consultant fees, data recovery, and lost time. Hardware damage from a bad surge: $2,000-$10,000 for a server replacement. A full business disruption from a server that won't boot Monday morning: hard to put an exact number on it, but we've seen businesses lose entire client relationships over reliability issues that started with a power event.
The ROI on a UPS is almost comically good. It's one of those rare cases in IT where the cheapest option is also the right option.
And here's the thing: if you look at your server room honestly and find a server plugged into a bare wall outlet or a consumer power strip, the UPS isn't the only problem. It might be worth scheduling a broader assessment of what else might be missing. We cover the most common oversights in our small business IT audit checklist.
But start with the UPS. It's the fastest win with the clearest payoff, and it's something you can have in place this week for under $300.
We're not going to get into brand-level specifics in a blog post — what's right depends on your equipment, your load, and your budget — but here's our general guidance:
For a small business with one server and standard network gear: A line-interactive UPS in the 1500VA / 900W range from APC (Back-UPS Pro series) or CyberPower (CP1500PFCLCD or similar). Budget $200-$350. Connect the server, NAS, firewall, and core switch. Install the management software. Done.
For a server rack: Consider a rackmount UPS so everything stays tidy. APC Smart-UPS series and CyberPower's OR series both offer rackmount options with good software support. Expect to spend $400-$700 for proper rack-mount protection in the 2200VA range.
If you're in a location with chronically bad power: Talk to us. There are situations where the right answer involves a generator, a transfer switch, or a double-conversion UPS — and making the wrong call can be expensive. We're happy to look at your specific setup and give you a straight recommendation. See our services page for how we approach infrastructure assessments.
One more thing: if you're not sure what's currently in your server room or whether you already have UPS protection, check. Walk back there right now if you haven't recently. Look at what's plugged into what. It's a five-minute walk that could save you a five-figure bill.