Sustainability in Data Centres: It Starts With Us ♻️


When sustainability comes up in tech, the spotlight usually swings straight to the SP's, hyperscalers and co-los: megawatt hungry halls, cooling towers, power purchase agreements, the whole lot. And yes—cloud providers and service operators absolutely must drive efficiency and decarbonisation. But there’s an awkward truth we often swerve: consumer behaviour multiplies demand, duplication, and waste. In other words, the cloud isn’t some distant, abstract weather system. It forms because we keep boiling the kettle.

A confession to set the scene

Here’s a very real (and slightly embarrassing) example from my own routine. I take a photo on my iPhone. It saves to iCloud. At home, my setup syncs that photo to a local NAS. The NAS then shuttles a copy to Google Photos, another to Amazon Photos, and another to OneDrive. One picture; many journeys; multiple providers; overlapping storage; repeated compute and network hops. Every hop means more power, more cooling, more embodied carbon wrapped up in disks, servers, and switches. It’s convenient, yes. It’s also duplication in a trench coat calling itself “belt and braces.”

Streaming sprawl (video & audio)

Want to watch a particular series tonight? Best of luck. Between Netflix, Disney+, Apple TV+, Discovery+, Paramount+, Sky (and friends), you’re likely juggling half a dozen subscriptions just to see the content you care about. Each platform runs its own encoding pipelines, catalogues, recommendation engines, DRM, analytics, and content delivery—often on separate cloud stacks. None of that is “free” just because it’s invisible on your telly.

Now zoom out to a typical evening at home: one screen on Disney+, another on Netflix, someone else on Amazon Music or Spotify. That’s multiple concurrent CDNs, caches, storage systems, logs, and analytics services firing in parallel. Architecturally, much of the workload is similar across providers, but competition and exclusivity mean parallel infrastructure rather than shared rails. The result? Overlapping capacity, duplicated compute, and higher aggregate energy use to deliver three shows and a playlist. Prestige TV, yes—prestige efficiency, not so much.

Smart home chatter (IoT and always-on)

A “smart” home often means a chorus of tiny devices constantly phoning home: thermostats, lights, sensors, cameras, doorbells, voice assistants—each chatting to its own cloud. Telemetry here, a video clip there, daily firmware checks, usage analytics, and push notifications sprinkled on top. Individually the data is small; in aggregate—scaled across millions of households—it’s a permanent background hum of compute, storage, and networking.

Add multi-vendor reality and you’ve got parallel clouds doing near-identical jobs. Some of this could be trimmed by sensible defaults (edge processing, local retention, selective sync), but the default experience tends to be “capture everything, upload everything, keep everything.” Convenient for features; costly for the planet.

Collaboration copy-and-paste (work & personal files)

We’ve all done it. A deck starts in Google Drive, gets exported to PowerPoint, copied into Dropbox, shared via Slack, emailed as an attachment (twice), and finally parked in OneDrive “for posterity.” Meanwhile, versioning makes offspring like rabbits. Each platform stores its own rendition, generates previews, indexes content, and maintains activity logs. If you also run local backups plus an off-site backup service, you’ve now multiplied the same 120 MB file across half the internet.

Just a few examples, I am sure you have many of your own examples of similar behaviour.

This isn’t malicious; it’s muscle memory. But every duplicate is a small IOU written against someone’s data centre. The bill arrives as more racks, more disks, more cooling, and—unless fully decarbonised—more emissions.

Why the overlap matters

  • Energy stacks up: Redundant uploads, mirrored libraries, and parallel platforms consume power—even when each provider is “efficient.”
  • Embodied carbon counts: More demand means more hardware manufactured, shipped, and installed. You can’t “offset” a server into existence.
  • Over-provisioning is rational but costly: Providers plan for peaks and resilience. If our habits inflate the baseline, the whole estate grows to match.


“But the providers should fix it” (yes, and…)

Absolutely, providers should keep pushing: renewable energy, heat reuse, advanced cooling, high utilisation, clever scheduling, and transparent reporting. Many are. But sustainability isn’t a spectator sport. If we continue to behave like every photo needs four clouds, every home needs ten perma-connected gadgets, and every document needs a residency in every SaaS under the sun, we’ll simply push even the greenest data centres to do more.

Practical ways consumers can help (without moving into a yurt)

  • Rationalise storage: Pick one primary cloud for photos and documents. Use the NAS or a single secondary cloud for true backup—no need for a conga line.
  • Rotate streaming services: Keep two active, park the rest. Binge, cancel, rotate. Your wallet and the grid will both breathe easier.
  • Tune your IoT: Disable unnecessary cloud features, prefer local processing where available, set shorter retention for motion clips and logs.
  • Stop attachment inflation: Share links with permissions instead of lobbing multi-MB files into five different platforms.
  • Choose greener vendors: Where you can, favour providers with credible renewable commitments and efficiency roadmaps—and who expose settings that let you minimise duplication.
  • Declutter periodically: Old backups and auto-uploads you’ll never use? Archive or delete. Digital minimalism is still minimalism.


Sustainability starts with the person in the mirror

Data-centre sustainability is not solely a supplier problem to be solved with bigger PPAs and clever evaporative cooling. It’s a shared responsibility. Our collective behaviour—how many services we subscribe to, how many clouds we duplicate into, how “always-on” we insist our homes and workflows must be—sets the demand curve that operators then have to meet.

If we want a truly sustainable digital ecosystem, we should start by trimming our own digital hedges: fewer redundant copies, fewer overlapping subscriptions, more thoughtful defaults, and a dash of restraint. Providers must keep cleaning up their house; consumers must stop tracking mud through the hallway. Sustainability doesn’t just live in the data hall—it begins with our decisions, one photo, one stream, and one shared file at a time.

[by: Grant Marais]