The mundane reality behind data centre announcements
Huge numbers. Massive builds. Here's what they don't tell you.

Huge numbers. Massive builds. Beyond the tantalising, breathless headlines, here's some mundane truths about the data centre industry that we don't talk about.
I've covered the data centre space for over a decade, and I've seen plenty of bold headlines or sweeping calls for the industry to "do better."
Not all of them make sense.
The truth is usually a lot less flashy, weighed down by technical constraints and pesky, real-world trade-offs.
Counting data centres
I used to keep a close eye on the number of data centres built, using it to convey a sense of scale in my stories. I've since stopped doing that, however.
Why?
Because when one facility can be 500 times bigger than another, the total count loses meaning, and can actually obscure more than it reveals.
And this is the reality today, with some 0.5MW data centres still operational, as @Roy Samuels observed, even as massive 290MW facilities are being constructed in Johor.
Tell me your MW
Fine, you say. Let's go by capacity in MW then. Surely that offers a better gauge. That sounds like a good plan, until you realise almost everyone overprojects.
Sure, they'll get there if they:
- Secure additional power.
- Sign on 2nd hyperscaler.
- Get that extra financing.
- Fill up the first 2 phases.
- Are still around 3 years later.
See where this is heading?
Or as James Rix astutely noted recently, most data centres never get to the full figure.
More sustainable data centres
I daresay all of us love our planet. I mean, it's not like we have the option to go elsewhere.
The reality is we can build very energy-efficient data centres. The problem usually isn't technology, but the financial impact of doing it.
- Capital expenditure.
- Operational cost.
- Speed to market.
- Reliability.
And even if you would pay more for your web services (don't lie), nobody's asking you, as cost is a top consideration across the ecosystem.
- Data centre operator.
- Hyperscale cloud.
- Tier 2 cloud.
- SaaS provider.
In fact, some might even move non-sensitive workloads to another geography to pare down costs.
tldr; it's complicated.
AI data centres overdose
I don't blame you if you think that everyone is building AI data centres. But it's a misguided view.
I'll classify DCs broadly into:
(a) AI data centres.
(b) Cloud data centres.
(c) Colocation data centres.
I'll write more on this - but let's just say AI data centres don't make up the majority.
Can you think of any boring or even inconvenient truths about the data centre industry?