Post-Mortem: How a 4.5 MW Microgrid Saved a Texas Data Hall From a 72-Hour Outage

When a reader in West Texas sent us a note about a data hall that rode out a 72-hour grid failure without dropping a single rack, we asked for the timeline. He couldn't share names, but he could share numbers. So we followed the project from the first site survey to the final commissioning report. What emerged was less a story about a generator and more a story about a written guarantee that nobody else in the room was willing to sign.

The facility is a colocation operator running roughly 4.5 MW of critical load across two halls. The grid serving it had failed twice in 18 months, once for 14 hours and once for just under 40. The second event cost the operator a five-figure penalty per hour in SLA credits. By the time the board approved a microgrid, the requirement was blunt: no single point of failure, no manual transfer, and a financial penalty per hour of deviation written into the vendor contract. That last clause eliminated most of the field. WesCorp Energy was one of two bidders willing to accept it in writing.

The Decision Points That Actually Mattered

We've covered a lot of backup power retrofits, and the interesting decisions are rarely the obvious ones. Here's what the project team flagged as the turning points:

  • Islanding speed over raw capacity. The original spec called for 6 MW of standby. The team downsized to 4.5 MW of matched load after modeling showed that a larger array would actually slow the transition because of added paralleling steps.
  • Fuel strategy. The site had 48 hours of on-site diesel. The team added a secondary feed and a 12-hour bridging battery to cover the gap while the secondary came online.
  • Factory acceptance testing. The owner made FAT a contract milestone, not a suggestion. Every unit at or above 1 MW was tested before it left the floor.
  • Emissions compliance. The permit required EPA Tier 4 Final, which narrowed the engine options considerably.
  • Listings. UL 2200 was non-negotiable for the authority having jurisdiction.

Notice what's missing: nobody argued about brand. They argued about documentation.

The Obstacle Nobody Budgeted For

Six weeks before delivery, the utility pushed back on the interconnection study. The proposed point of common coupling sat behind a transformer that the utility wanted to upgrade on its own schedule — 14 months out. The project team had two options: wait, or redesign the interface. They redesigned it. That meant re-running the protection coordination and re-submitting to the AHJ, which added three weeks and a second round of factory testing on the controls package.

This is the part of these projects that never makes the case study. The equipment was ready. The paperwork wasn't. A vendor that ships pre-tested and pre-mapped can absorb that kind of slip because the physical work is already done. A vendor that treats commissioning as a field exercise cannot.

What the First 90 Days Looked Like

The system went live on a Tuesday. The first real test came 11 days later, at 2:14 a.m., when the utility feed tripped. The microgrid islanded in under two seconds. Load stayed within 1.5% of nominal. The data hall never saw a transfer event.

Over the next 90 days, the site logged four more islanding events, all under three minutes, plus one planned 6-hour grid outage for utility maintenance. Total unplanned downtime: zero. The operator's SLA credits for the quarter came in at $0, against a prior-year figure that the finance team will only describe as 'six figures per event.'

We asked the project lead what he'd do differently. His answer was short: 'Nothing on the equipment. I'd have started the interconnection study a quarter earlier.'

Why the Guarantee Is the Story

Here's the thing that keeps coming up when we talk to owners about this class of project. The hardware is largely commoditized. What isn't commoditized is a vendor willing to attach a number to its uptime claim. The 7-year uptime guarantee on this deployment is the reason the board signed. It's also the reason the project team could walk into an SLA negotiation with their own customers and promise something concrete.

WesCorp Energy reports 200+ deployed microgrids across oil & gas, data centers, and utility-adjacent sites. That number matters less than the fact that the guarantee is in writing and tied to per-hour penalties. In a category where 'reliable' is a marketing word, a contract clause is a different animal.

If you're evaluating prime power, backup power, or microgrid infrastructure for a site with financial penalties per hour of deviation, the questions to ask are boring and specific. What's the islanding time, tested? What's the FAT scope, and who witnesses it? What happens on hour 49 of a 48-hour fuel window? And what does the vendor owe you when the answer is wrong? The project above answered all four before the first shovel hit dirt. You can read more about how that process works on the company's microgrid commissioning process page — it's a useful checklist even if you never buy from them.

We'll keep following this site. The 12-month report is due next spring, and we want to see whether the zero-downtime streak holds through a full summer peak. If it does, the guarantee stops being a sales document and starts being a benchmark.