Cut data center energy costs with bare metal automation

by David Beamonte on 26 June 2025

Data centers are popping up everywhere. With the rapid growth of AI, cloud services, streaming platforms, and connected devices, the demand for compute is growing, – and data centers are in the middle of it all. But while they’re essential for the digital economy, their energy consumption poses a major challenge.

Unfortunately, a large part of data centers’ energy just isn’t used efficiently. A surprising amount goes into physical machines that are powered on but idle. And it’s not just wasteful, it’s also unnecessary, given how automation and tools like Canonical MAAS (Metal as a Service) can make a real difference. In this blog, we’ll explore how to improve data center energy efficiency and reduce power waste through smart automation tools.

The energy challenge in modern data centers

It’s easy to see why data centers need so much power: servers, cooling systems, network infrastructure – it all adds up. And the more powerful the workloads are (think large-scale AI training or video processing), the more energy is needed to keep things running. In fact, some providers are going as far as building their own renewable energy plants just to keep up.

But here’s the thing: not all of those servers need to be on 24/7.

In virtualized or containerized environments, orchestrators can handle scale automatically. Need more capacity? Spin up more VMs or containers. Traffic slows down? They scale back, saving resources. But physical machines don’t work that way – or at least, not out of the box. Most orchestration tools stop at the OS level and don’t control the power state of the underlying hardware.

So in many cases, physical servers stay powered on even when they’re not doing anything. Development machines run over weekends. Infrastructure for bursty workloads (like telco or streaming services) stays fully online, just in case demand spikes.

This kind of constant uptime drains your power budget and shortens your hardware’s lifespan. But with the right automation tools, there’s no reason bare metal can’t be just as dynamic and responsive as VMs.

When can machines be powered down?

Not every workload needs to run all the time. In fact, there are plenty of scenarios where it makes perfect sense to power down machines, if you have the tools to do it cleanly and bring them back up when needed.

Here are a few common examples:

Development and test environments

Many companies maintain dedicated physical environments for development, testing, or staging. But developers usually work weekdays, 9 to 5. That means those machines often sit idle overnight and throughout the weekend. With automation, these servers can be shut down during off-hours and brought back up when the team logs in Monday morning.

Bursty or on-demand services

Some workloads only spike at specific times. For instance, think of telecom platforms during major events or streaming services during prime time. During low-traffic periods, there’s no need to keep all infrastructure running at full capacity.

Scheduled jobs and batch processing

Some systems run on a schedule. Nightly data processing, weekly builds, or monthly reporting tasks are good examples. Instead of keeping the underlying hardware online 24/7, you can power it up just in time for the job, then shut it down afterward.

Labs and temporary environments

QA labs, customer demos, or sandbox environments often run temporarily but remain online out of habit or convenience. Automating their lifecycle ensures they’re only using resources when they’re actually needed.

These are just a few examples, but the core idea is simple: if a machine isn’t doing valuable work, it probably shouldn’t be on. That’s the kind of mindset that leads to real energy savings. And it’s exactly the kind of thing automation can handle, especially with a bare metal management tool.

The opportunity: smarter infrastructure management

So, if energy waste is such a common problem in data centers, why hasn’t it been fixed already?

A big part of the answer is tooling. Virtual machines and containers have powerful orchestration layers (such as Kubernetes, Juju, and so on) that can scale workloads automatically. But bare metal has traditionally lagged behind. Managing physical infrastructure has often meant static, manual provisioning, and once a server is on, it tends to stay on.

That’s a missed opportunity – and the best way to benefit from it is to use smart automation or management tools. Such tools typically allow you to:

Reclaim unused hardware quickly
Repurpose machines on demand
Provision and decommission systems based on real usage patterns

Bare metal automation opens the door to treating physical infrastructure more like cloud infrastructure: elastic, dynamic, and efficient. With the right tools in place, you can cut waste, reduce costs, and make better use of the hardware you already have.

That’s where Canonical MAAS comes into the picture, and how it fills the orchestration gap left by VM- and container-focused tools.

How bare metal automation solves the data center energy waste problem

Canonical MAAS brings cloud-like automation to bare metal. It provides full lifecycle machine management, from discovery and commissioning, to OS deployment. The API abstracts physical machines so that the upper level applications and orchestrators can treat physical machines as if they were virtual machines.

And how can we use MAAS to solve the energy waste problem?

On-demand provisioning and deprovisioning

MAAS lets you treat physical servers like elastic resources. You can deploy a machine when it’s needed and release it when it’s not. Whether it’s powering down unused dev machines or spinning up nodes for a high-traffic window, MAAS makes it easy to manage physical infrastructure dynamically.

Power management through integrated APIs

MAAS supports a wide range of power management interfaces (IPMI, Redfish, and others), so it can turn machines on or off remotely. That means you can build automation around actual usage, powering down idle servers during off-hours and waking them up when needed.

Integration with orchestration tools

While Kubernetes handles your container workloads and Juju orchestrates applications across clouds, MAAS takes care of the underlying metal. This creates a complete stack where every layer can scale intelligently from bare metal to application. With the right orchestration tools’ configurations, machines can be powered off when unused and can be brought to life again when the resource is needed.

Scriptable and customizable

Do you want to create your own application? No problem. MAAS offers REST APIs and CLI tools so you can script exactly how your infrastructure behaves. Want to schedule weekend shutdowns for certain machine pools? Or automatically decommission hardware after a job finishes? MAAS makes it possible.

Example: Juju + MAAS in Action

Juju is widely used to deploy and manage applications across hybrid infrastructure. By integrating Juju with MAAS, you can extend this automation to physical servers’ power states.

How it works:

Deploy your MAAS and Juju controllers.
Add MAAS to Juju as a cloud provider.
Now, Juju can deploy applications to the MAAS-managed bare metal servers
Create a centralized system script that:
- Monitors machine states by polling Juju’s status or querying machines directly (e.g., via SSH or agent checks).
- Powers off idle machines via the MAAS API when they meet criteria (e.g., no Juju workloads for >1 hour).
This solution relies on MAAS + Juju auto-scaling to power machines back on when resources are needed (e.g., during juju deploy or add-unit).

Result:

Servers only consume energy when actively used.
Teams retain on-demand access without manual intervention.
Energy costs drop significantly—especially for environments with predictable downtime (e.g., nights/weekends).

This solution mirrors the elasticity of the cloud but applies it to bare metal, closing the efficiency gap.

Use MAAS to cut your data center power bill today

Data centers don’t have to be power-hungry monsters. With smart automation, you can reduce energy waste and operational costs, and make your infrastructure greener, without sacrificing performance or flexibility.

Here’s how to get started:

Try MAAS for free: Experiment with Canonical MAAS to automate your bare metal power management.
Integrate with Juju: Combine MAAS with Juju to orchestrate workloads and hardware power states seamlessly.
Assess your idle workloads: Identify the events when machines can be powered down (e.g., dev environments, batch jobs).
Automate, measure, repeat: Start small, track energy savings, and expand automation to more workloads.

The future of data centers is efficient, dynamic, and sustainable. With tools like MAAS, that future is within reach.

Ready to cut energy costs? Learn more about MAAS or contact our team for a custom solution.