Here is a question most operations managers never think to ask until it is too late:
When was your UPS battery bank last tested?
Not visually inspected. Not switched on. Tested, meaning someone measured the resistivity, checked the float voltages, and gave you a verified answer on how long that battery bank will actually hold your load if the power goes out tonight.
If you cannot answer that question with a date and a name, you have a problem. And the problem is not the UPS. The problem is that you are running a critical infrastructure facility on an assumption.
When was your UPS battery bank last tested?
Nathan Chirwa, Director at Boron Facilities, has been maintaining UPS systems in financial services and telecoms environments for over a decade. He describes battery failure in terms that are worth understanding precisely.
When batteries age, the electrolyte inside them dries out. As it dries, the cells become conductive in ways they were never designed to be. They begin shorting against each other. The hydrogen gas that builds up inside the casing pressurises it. And then the casing bursts.
That is not a metaphor. Batteries explode. Inside your UPS cabinet. And if you have a cabinet full of them and one goes, the cascade effect is not theoretical; it is a question of when, not if.
But most facilities never reach that point. What they reach first is something quieter and more dangerous: a battery bank that appears to be functioning, but has silently lost a third of its autonomy.
A UPS specified to carry your IT load for 60 minutes may only give you 20. The discharge curve is not linear; it is exponential. When a battery is approaching end-of-life, it holds voltage reasonably well until it doesn’t, and then it drops fast. Your IT team expects 60 minutes to execute a controlled shutdown. They get 20 minutes of warning and then a blackout.
“Batteries are your last line of defence in your IT systems. That’s why you have to take care of them at all times.”— Nathan Chirwa, Director, Boron Facilities
In a financial services environment, the losses from a UPS failure are not measured in inconvenience.
A telecoms network with five million active users goes down for 10 minutes. Every transaction, every call, every data packet that should have been processed in those 10 minutes is simply gone.
The penalties in enterprise contracts, owed to downstream clients who had their own SLA obligations, can run to R200,000 per hour, per client. Multiply that across 15 enterprise tenants, and the number becomes untenable.
For investment firms, the exposure is different but equally severe. A 10-minute window where a server room is dark is a 10-minute window where shares are moving, and trades are not being executed. Those are not recoverable losses. The market does not pause while your UPS is repaired.
A Tier 3 data centre is required to maintain 99.982% uptime.
Translated: the total allowable downtime across an entire year is approximately 1.6 hours. Not per event. Total. Annually.
To achieve that standard, you do not rely on a battery bank that is overdue for service. You run dual redundant systems, two UPS units, each capable of carrying the full load independently, each maintained on a three-month service cycle without exception.
At Boron, the three-month cycle is not a recommendation. It is the standard they have held across MTN, Old Mutual, Metropolitan, and every other critical infrastructure client in their portfolio
Beyond the obvious, keeping your servers on when the power fails, a well-maintained UPS performs a function most people overlook. Eskom supply is not always clean. It dips. It swells. Frequencies drift. Your generator takes between 8 and 30 seconds to come online, and during that time, it runs rough. Your sensitive IT equipment does not want any of that on its power rail.
The UPS, and specifically the battery bank at its heart, sits between the misbehaving supply and your load, converting whatever chaotic input it receives into a clean, stable output. That is what batteries do. They are the armour. And most companies have no idea whether their armour is still intact.
For investment firms, the exposure is different but equally severe. A 10-minute window where a server room is dark is a 10-minute window where shares are moving, and trades are not being executed. Those are not recoverable losses.
The market does not pause while your UPS is repaired.
Not “is the UPS working?” It is working until it isn’t.
The question is: when was the battery bank last independently tested, by a qualified contractor, with measured results on paper?
If the answer is more than three months ago, or if you do not know the answer, the time to find out is before load-shedding returns, not during it.