Is Cloud Repatriation a Green Excuse?
The Flexera State of the Cloud Report found 57% of organisations either have or plan sustainability initiatives, including cloud carbon footprint tracking.
The same report found 21% of cloud workloads have already been repatriated, while 31% of organisations now prioritise cost optimisation and sustainability equally. Environmental concerns are influencing infrastructure decisions. Cloud repatriation is accelerating. The question is whether these are two separate trends, or a convenient alignment.
When actual energy consumption data from AWS, Azure, and Google Cloud is compared with on-premises infrastructure, the environmental case for cloud repatriation becomes more complex.
The sustainability claim behind cloud repatriation
Moving workloads to more energy-efficient on-premise or colocation facilities sounds responsible. Choosing cloud providers based on environmental commitments sounds sensible too. The real question is whether the numbers support it.
Microsoft’s research shows Azure Compute delivers a 52–79% reduction in power consumption compared with on-premises environments, depending on deployment maturity. That includes 79% against dedicated physical servers and 52% where on-premises servers are already virtualised.
Those are not marketing estimates. They are measured against real on-premise configurations.
Google Cloud regions publish hourly carbon-free energy (CFE) scores. Finland’s data centre reports 97% CFE. Montreal reports 98% CFE. That means 97–98% of the time, the electricity flowing into those facilities comes from carbon-free sources rather than certificates purchased to offset annual use.
So when organisations cite environmental reasons for cloud repatriation, the comparison matters. Which on-premise setup is being measured, and against which cloud region?
The data centres few organisations are building today
On-premise infrastructure in 2026 usually means one of three things: older data centres using legacy cooling and power systems, colocation facilities sharing modern infrastructure across multiple tenants, or a newly built private cloud designed to hyperscale standards.
That third category remains uncommon.
Google’s data centres operate around 1.1 PUE. The average on-premises data centre sits closer to 1.58, according to Uptime Institute data, while older facilities often range between 1.5 and 1.8. PUE measures the ratio of total facility energy to the energy of computing equipment. Every kilowatt-hour of computing in a legacy facility may require 0.5–0.8 kWh of overhead for cooling and power delivery. Google’s equivalent overhead is closer to 0.1 kWh.
That efficiency gap grows at scale. Microsoft, Google, and AWS continue investing billions in cooling innovation, custom server design, and AI-led power management. Many legacy on-premise environments do not have those advantages.
The environmental case for cloud repatriation often depends on one of three assumptions: that an organisation can match hyperscale efficiency, use colocation facilities that outperform the cloud, or accept higher energy use while claiming sustainability benefits.
What the carbon calculators show about cloud repatriation
AWS, Azure, and Google Cloud all offer tools for tracking emissions. The challenge is that they do not always measure the same things, and methodology matters as much as the tool itself.
AWS and Azure use market-based accounting, which can allow near-zero emissions claims in fossil-fuel-heavy regions if enough renewable energy certificates are purchased. Carbone 4’s methodology analysis noted that some calculators can “give near-zero emissions as a result, creating an illusion of an impact-free service.”
Google publishes actual grid carbon intensity on an hourly basis. That allows organisations to verify whether workloads in Finland are genuinely running on hydro, nuclear, and wind power 97% of the time, or whether certificates are masking fossil-fuel consumption elsewhere.
For organisations using environmental reasoning to justify cloud repatriation, the methodology question is critical. Market-based accounting and location-based accounting can produce very different conclusions.
The cloud repatriation maths that needs scrutiny
Companies repatriating to modern colocation facilities with high utilisation and strong PUE can credibly argue environmental benefits versus some cloud regions with fossil-heavy grids. Companies repatriating to legacy on-premise environments built between 2015 and 2020 need to account for the PUE gap, utilisation gap, and grid carbon intensity where cloud regions already source renewables.
| Infrastructure Scenario | Energy Efficiency (PUE) | Carbon Intensity | Total Impact |
|---|---|---|---|
| Google Cloud Finland | 1.1 | 97% carbon-free hourly | Baseline for comparison |
| AWS Ohio (us-east-2) | 1.2 | 28% renewable (market-based) | Higher carbon, similar efficiency |
| Modern colocation | 1.3–1.4 | Depends on facility location | Lower utilisation may reduce gains |
| Legacy on-premise 2015–2020 | 1.5–1.8 | Depends on procurement mix | Up to 3–4x carbon per workload vs Google Finland |
| New private cloud | 1.1–1.2 | Depends on procurement | High capital cost, multi-year build |
Most cloud repatriation projects are not building new hyperscale private infrastructure. They are moving workloads back into existing environments that became underused after cloud migration.
What equal prioritisation often means
Flexera found that 31% of organisations now prioritise cost optimisation and sustainability equally. Not sustainability first. Equal priority.
Andreessen Horowitz’s analysis of 50 top public software companies found that cloud infrastructure costs can erode gross margins by 50% or more. Dropbox reportedly saved around $75 million over two years by repatriating workloads to colocation facilities. 37signals publicly documented leaving AWS, with CTO David Heinemeier Hansson projecting savings of more than $10 million over five years.
These are cost-led decisions with clear financial outcomes. Environmental benefits may be secondary.
When organisations say they equally prioritise cost and sustainability, it often means moving workloads for cost reasons first, then evaluating the sustainability case afterwards.
What procurement teams should actually check
If a vendor says cloud repatriation is sustainability-led, ask for:
- Current cloud region(s) and their location-based carbon intensity
- Destination data centre PUE and grid carbon intensity
- Server utilisation rates today versus projected after migration
- Three-year total energy consumption forecast with methodology disclosed
Without that data, an environmental justification may simply be a cost justification framed through sustainability messaging.
If a vendor proposes staying in the cloud for environmental reasons, verify:
- Which regions they use
- Whether accounting is market-based or location-based
- Whether workloads could move to higher-CFE regions without latency or compliance issues
Cloud platforms are often more energy efficient than legacy on-premise systems. But not all cloud regions are equal, and not all on-premise environments are outdated.
Distilled
Microsoft data shows Azure Compute can reduce power use by 52–79% compared with on-premises systems. Google Cloud Finland runs at 97% carbon-free hourly energy. Those are measurable outcomes.
Cloud repatriation is frequently driven by cost, from Dropbox’s reported $75 million savings to 37signals’ projected $10 million reduction over five years. Environmental claims require closer scrutiny. Legacy on-premise infrastructure with a PUE of 1.5–1.8 compared with high-efficiency cloud regions at 1.1 is not automatically a greener move.
For IT procurement teams, the environmental case for cloud repatriation requires location-based carbon data, PUE comparisons, utilisation assumptions, and grid-intensity verification. Without that level of detail, sustainability claims may hide a simpler cost story.
