Device Power Is a Cost Line and a Carbon Line. Most Organisations Measure Neither.

Editor’s note. This post was published in April 2016 as the write-up of an internal customer survey. The survey figures — an average return in 6.26 months, £56,292 saved per organisation per year, 363 tonnes of CO2 per organisation, and the NHS-wide extrapolations drawn from them — have been removed on migration because we cannot substantiate them against a published source. So has the product the original was about. The argument underneath them still holds, and it is what follows.

Most of the cost in IT is committed. You cannot stop paying for the licences people use, the storage that holds live data, or the network that carries it. Idle computing is different. A workstation drawing power overnight in an empty building, a virtual machine stopped but never deallocated, a training cluster left running after the run finished — none of these produce anything. Removing them costs nobody anything.

Idle computing is also one of the few IT costs that shows up twice: once on the invoice and once in the carbon account. For any organisation with a reporting obligation on both, that makes it unusually good ground.

The reason it rarely gets addressed is not that the saving is hard to believe. It is that the measurement is hard to produce.

You cannot schedule what you have not discovered

Every power or idle-resource programme runs into the same wall in its first week: nobody can produce a device list anybody trusts. The service desk’s list disagrees with the directory, the directory disagrees with the discovery tool, and none of them know about the Macs, the Linux boxes or the machines on the sites that were acquired rather than built.

That matters more than it sounds. A saving estimate built on a device list that double-counts is wrong upwards. One built on a list that misses whole platforms is wrong downwards. You end up defending a number instead of banking a saving.

CerteroX ITAM addresses this at the discovery layer rather than the reporting layer:

• Ten discovery methods into one schema — native agent, command-line inventory, agentless, standalone for air-gapped systems, Active Directory import, network scan, third-party ITAM import, cloud and SaaS connectors, browser monitoring and file metering. There is no reconciliation project afterwards because there is nothing to reconcile.
• Six operating system families on the same agent — Windows, macOS, Linux, IBM AIX, HP-UX and Oracle Solaris. Platforms that most tools treat as an integration problem are inventoried the same way as everything else.
• Network Discovery sweeps a class-C subnet in under five seconds across NetBIOS, SNMP and ICMP, so you find the machines before you own the agent rollout.
• Duplicate system detection, stale device archiving and non-persistent VDI support — the three things that most reliably corrupt a device count.

Get that right and the rest is arithmetic. Get it wrong and every figure you publish afterwards is contestable.

Where the idle cost has moved since 2016

In 2016 this argument was almost entirely about desktops overnight. It is not any more. The same waste pattern now runs through cloud infrastructure, where it is metered by the hour and billed automatically.

29% of cloud spend is wasted, up for the first time in five years. Certero’s own figure for the average saving achieved across cloud environments under management is 38%.

CerteroX Cloud Management attacks it with named, individually tunable checks rather than a general “savings identified” number:

• VM power schedules for automated start and stop — the direct descendant of switching desktops off at night.
• Instances in a stopped state for a long time — stopped is not deallocated, and the difference is a bill.
• Abandoned instances, images, load balancers, S3 buckets and Kinesis streams, and obsolete images, IPs, snapshots and snapshot chains.
• Short-living instances flagged as spot and preemptible candidates.
• Resource TTL with automatic lifecycle enforcement, so the thing that was meant to live for a day does not survive the quarter.
• Per-check thresholds, pool exclusions and account skips, because a check that cannot be tuned to your environment gets switched off within a month.

Twenty-six checks ship, each named and each configurable. The same engine is applied to machine-learning executors, which is where GPU time quietly becomes the most expensive idle resource an organisation owns.

Turning inventory into a cost position

Discovery tells you what exists. Two further things turn it into something a finance director will act on:

Costing rules. CerteroX ITAM carries cost tabs with manual and automatic costing rules, so hardware records hold a cost model rather than just a serial number. That is the difference between “we have 3,000 devices” and “this is what those devices cost us.”

Grouping and thresholds. Dynamic, static and custom groups can be built with a query builder or SQL, which lets you target a policy at a site, a department or a shift pattern rather than at everything at once. Trend charts, KPIs and threshold alerts then tell you when a group drifts, instead of leaving it to a quarterly report nobody opens.

Policy that is checked, not circulated. Governance Policies work as compliance-as-code over a reusable filter builder, with policy definitions that export and import as JSON. The examples that ship are security and tag hygiene checks — BitLocker enabled, Defender running, Azure VM tag hygiene — but the builder works against inventoried attributes generally, which is what makes a configuration standard enforceable rather than aspirational.

The public sector case

Certero’s published NHS work is where this argument was originally made, and those references stand:

NHS South West London ICB mitigated £100k of Microsoft compliance risk. Reece Emson, their ITAM Asset/PSL Manager:

Certero’s SAM managed service allowed us to significantly mature our license posture at a fast pace, something that would have taken 3-4 years without their involvement.

East of England Ambulance Service brought 130 sites into view. Andy Marrs, IM&T Security & Resilience Manager:

The tool has truly transformed how we work, making life a lot easier with complete visibility of assets and automation removing the need for manual intervention.

An NHS Commissioning Support Unit deployment serves over 100 NHS organisations.

Those are published outcomes with named sources. They are a better argument than an extrapolated survey, which is part of why the extrapolated survey is no longer in this post.

What to do about it

1. Fix the device list first. Every subsequent number depends on it, and a contested denominator kills the business case before it reaches finance.
2. Attach a cost model to the inventory, so the conversation is about money rather than machine counts.
3. Start where the meter runs fastest. Cloud and GPU idle time bills by the hour; desktops bill by the year. Sequence accordingly.
4. Automate the enforcement. Power schedules, TTLs and threshold alerts survive a reorganisation. A policy document does not.
5. Report the carbon alongside the cost. It is the same measurement, and it reaches an audience the cost figure does not.

Idle computing is the rare saving with no trade-off attached. The only thing standing between an organisation and it is usually an inventory nobody believes.

Book a demo, or read more about CerteroX ITAM and CerteroX Cloud Management.