Uptime Institute, LLC

This is why some facilities provide alternative routes for power and cooling, allowing equipment to be maintained while the computers continue operating.

Uptime Institute describes this capability as being concurrently maintainable.

Not every data center is built to the same level, and additional equipment alone does not guarantee uninterrupted service.

Procedures matter too, which switch is opened, how a repair is checked, and what the next person needs to know.

Uh, the, really the, the main point I was gonna add is that a lot of these, uh, redundancy, uh, requirements have been, you know, part of the industry, um, for a long time from a data center perspective.

So, uh, there's been organizations like the Uptime Institute as well as the, uh, Telecommunications Industry Association.

Uh, they established a tiering program, so there's, there's four tiers, tier one being just a basic, what we call basic infrastructure.

So to the critical path that, you know, David, you know, talked about, if you have your electrical primary path to the critical load, a basic infrastructure means no redundant components built in.

This is about between, it's about Q4, really, or between now and year end, or even just the next 90 days, because by December, it's done, right? So what's really actually genuinely coming in and what's genuinely going out and where are the gaps and where are the assumptions and where are the reserves? Where have been the major hiccups and where did you say that you were going to make it up? And can you still? Q4 always has surprises, right? And the question is whether you know where your actual buffer is right now versus where you thought it maybe was in January.

And the reason why this matters is that more than half of the respondents to the 2024 Uptime Institute survey said that their most recent significant outage costed them more than $100,000.

And one in five said it cost them more than a million.

And 90% of mid-sized and large enterprises lose upwards of $300,000 per hour of downtime And for 41% of the enterprises, that figure reaches 1 to 5 million per hour.

And I think that's something I'd love to see happen.

The Uptime Institute has it.

And what was interesting, the outage data this year, not only did they declare what a fire is, this year I think they said 6% of outages.

Don't hold me to that number.

The industry-wide average sits at 1.

52 in Uptime Institute's 2026 survey, recent builds routinely post 1.3 orbiter, and the biggest hyperscale fleets report numbers close to 1.1.

But PUE measures how efficiently you reject heat.

It has no term for whether the heat did anything useful after it left the rack.

In a media release, CEO Craig Scroggie said M4 would have direct-to-chip closed-loop cooling supporting rack densities of 1,000 kilowatts using recycled non-potable water.

M4 has all the top-tier global safety certificates, including the prestigious Tier 111 and Tier 4 certifications from the Uptime Institute, which sets the global standard for evaluating data centre reliability, availability and performance.

NextDC did not return calls for comments on the fire risk.

Ironically, the elaborate cooling systems used to cool the computers in data centres can cause fires.

As the climate warms and facilities attempt to minimise cooling for efficiency reasons, we could see a trend for cooling system failures, wrote Peter Judge in Data Centre Dynamics magazine.

Judge is now a senior research analyst with the Uptime Institute.

He told Michael West Media he still stood by these quotes, reporting that a cooling system caused a fire in China in October 2022.

A backup cooling tower caught fire in the block that houses the data center at China's Suzhou Supercomputing Center in the Suzhou Industrial Park, a technology and business flagship for the country.