The Green Cloud Paradox

The Green Cloud Paradox: Why 44% of companies can’t accurately measure their cloud emissions

The tech industry often advertises major savings potential through the cloud, for example with model calculations that speak of many millions of tons of CO₂. However, these figures are based on assumptions about energy efficiency and the share of green electricity that are not always met in practice and are not independently validated by science. At the same time, a recent Capgemini study shows that 44% of European companies cannot correctly capture or measure their actual cloud emissions (source: Capgemini). What at first glance looks like a gap in data collection is increasingly becoming a serious compliance risk for companies in view of growing regulatory requirements in 2024.

The Corporate Sustainability Reporting Directive (CSRD) makes granular CO₂ accounting a legal obligation – with fines of up to 10 million euros for companies that do not correctly record their Scope 3 emissions. Independent studies and experts criticize the fact that leading cloud providers' emissions calculation tools do not fully capture a significant share of actual CO₂ emissions, especially indirect emissions.

This article examines the complexities of the “Green Cloud Paradox” and shows how companies can navigate between promised savings and hidden emission shifts without taking on regulatory risks.

The Anatomy of the Green Cloud Paradox

Efficiency gains vs. systemic impacts

Modern cloud data centers do in fact achieve impressive efficiency gains. Hyperscale providers (e.g. Microsoft) demonstrate 93% higher energy efficiency than traditional enterprise infrastructure through three key innovations:

Dynamic workload distribution drastically reduces idle capacity, while liquid-immersion cooling systems lower PUE values to 1.1 and mPUE to 1.03 (source: Submer) – compared with the industry average of 1.55 (as of 2022).

However, these are best-case scenarios in highly optimized data centers. According to its own sustainability report, Microsoft's average global power usage effectiveness (PUE) value is 1.18 (Microsoft, 2023). In many cases, standard colocation data centers with higher PUE values > 1.3 are also used.

AI-powered predictive maintenance extends server lifespans and can reduce downtime by 15%.

These technical advances are real and measurable. Microsoft's lifecycle analysis shows that Azure services can reduce CO₂ emissions per user by up to 98% – in regions with renewable energy integration.

The hidden emissions

But this is where the complexity begins. The EU product environmental footprint methodology exposes critical gaps in conventional cloud emissions accounting:

Embodied carbon from server manufacturing contributes 20–50% of total emissions (source: Circular Ecology). While companies traditionally use servers for 5–7 years, hyperscale operators replace hardware every 3–6 years in order to maintain technological advantages. A typical data center requires significant amounts of raw materials over the course of its operation: for example, German data centers already contained a total of around 7,000 tons of aluminum and 17,000 tons of copper in 2008, with aluminum being the most commonly used material (source: Umweltbundesamt). In addition, rare raw materials and precious metals such as gold, silver, and palladium are also used in the production of servers and infrastructure, further increasing the ecological footprint.

Rebound effects intensify the problem: despite rising efficiency, data center electricity consumption is expected to grow by 75% from 2022 to 2026, according to the IEA (source: BMWK). One reason is the enormous energy appetite of modern AI applications: training large models such as GPT-3 already requires as much electricity as a medium-sized nuclear power plant in one hour, and individual AI queries are also significantly more energy-intensive than traditional search queries at 3 to 9 watt-hours. If in the future an increasing share of the roughly 9 billion daily search queries is answered by AI, the energy demand for search queries could increase thirtyfold. In this way, increased efficiency leads not to less consumption, but to significantly more (source: elektronikpraxis.de).

The measurability crisis

The measurability crisis in digital sustainability reveals a central challenge: the lack of transparency along digital supply chains – especially with regard to server infrastructure, data centers, and cloud services. Companies can often only capture the ecological footprint of their IT in fragments because reliable, comparable data is missing. Especially when outsourcing to the cloud, it often remains unclear how high the actual energy consumption or CO₂ emissions per application or user are – because providers often disclose only average values or incomplete information. The physical infrastructures behind digital services – i.e. data centers, networks, and servers – thus remain invisible. This leads to an accountability gap: without standardized metrics, granular data, and open interfaces, digital sustainability remains difficult to assess, preventing companies from making informed decisions to optimize their IT. The result: Green IT becomes a buzzword without being reliably provable or comparable.

Regulatory reality: CSRD means business

The new rules

The Corporate Sustainability Reporting Directive fundamentally changes the landscape. Listed companies must now report cloud-related emissions under “Purchased goods and services” (NFRS E1-6).

The CSRD requires companies to comprehensively disclose their greenhouse gas emissions according to Scope 1, 2, and 3 categories. This classification includes direct emissions (Scope 1), indirect emissions from purchased energy (Scope 2), and the usually more extensive Scope 3 emissions from the entire upstream and downstream value chain. Under the European Sustainability Reporting Standards (ESRS) – in particular under ESRS E1 (“Climate change”) and there under E1-6 (“Gross Scopes 1, 2, 3 GHG emissions”) – companies are also required to disclose emissions from purchased cloud services as part of the category “Purchased goods and services” (source: TPA).

A distinction must be made between market-based (PPA-backed) and location-based emission factors. The ESRS E1 standards require companies to state and explain both market-based and location-based emission factors for Scope 2 emissions. This includes electricity procurement in particular, for example via power purchase agreements (PPAs) or guarantees of origin (source: EFRAG).

EU Taxonomy requirements

Within the framework of EU Taxonomy reporting, Beiersdorf is required to specify the extent to which revenue, capital expenditures, and operating expenses are associated with environmentally sustainable economic activities. For digital infrastructure, activity 8.1 (“Data processing, hosting and related activities”) is particularly relevant, because Beiersdorf operates its own data center and additionally uses cloud computing services from third-party providers. In the context of the taxonomy, these activities are classified as potentially sustainable if they meet certain technical screening criteria. This means that both self-operated data centers and purchased cloud services are subject to review regarding their environmental impact – especially with regard to their CO₂ emissions and energy efficiency. Classification follows category C (“low-carbon measures”) and aims to reduce greenhouse gases. This focus on digital infrastructure underlines the growing importance of data centers and cloud use as part of sustainability-relevant corporate activities in the spirit of the EU Green Deal (source: Beiersdorf).

Solutions: From the Green Cloud Paradox to measurable sustainability

The Green Cloud Paradox makes it clear: the gap between ambitious sustainability goals and actual emissions measurement is significant – especially in the cloud. While regulatory frameworks such as the CSRD and ESRS require companies to report emissions holistically, robust data is lacking in many places, especially for Scope 3 emissions from purchased cloud services. Hyperscalers do provide their own “green” tools, but studies show that up to 44% of real cloud emissions are not captured correctly as a result. This is especially critical for companies that must report their cloud and data center infrastructure as taxonomy-eligible activities – and therefore depend on transparent, reliable data.

Practical implementation and best practices

• Target transparency gaps deliberately: Companies should not record cloud emissions solely on the basis of hyperscaler dashboards, but should supplement them with independent tools and standards (e.g. Cloud Carbon Footprint, SCI – Software Carbon Intensity). Comparing them with electricity consumption at the VM or container level also creates additional validity.

• Establish Scope 3 readiness: Cloud-related emissions must be systematically captured as part of the category “purchased goods and services” and embedded in the overarching Scope 3 strategy – including infrastructure, licensing model, and user behavior.

• Design cloud architecture sustainably: Sustainability should be incorporated as a design principle into architecture decisions – for example through efficient resource allocation, scaling-aware workloads, region optimization (green data centers), and targeted shutdown of unused instances (“zombie VMs”).

• Demand supplier transparency: Companies can and should require sustainability data contractually – for example through binding KPIs for emissions reporting in SLAs or by selecting cloud providers that support Science-Based Targets.

• Develop a taxonomy-compliant digital strategy: Activities such as “Data processing, hosting and related activities” (EU Taxonomy activity 8.1) offer strategic leverage when they demonstrably contribute to emissions reduction. Requirement: a structured ESG data architecture that maps cloud usage at a granular level.

Conclusion

The green cloud does not exist as a marketing label – only through measurable, transparent, and controllable sustainability. Companies that consistently develop their digital infrastructure in light of the CSRD, ESRS, and EU Taxonomy turn regulatory pressure into real decarbonization levers. Anyone who makes cloud emissions visible creates not only trust, but also resilience over the long term.

The Green Cloud Paradox is not a technical problem – it is a transparency and accountability problem. While hyperscalers rightly highlight their efficiency gains, incomplete accounting frameworks obscure the true environmental impact.

The solution is not a return to inefficient on-premise infrastructure, but a demand for full transparency and scientifically grounded measurement methods. Companies that act now can turn the paradox into a competitive advantage.

Three key questions for your organization

1. Can you prove your cloud emissions at server level?

2. Are your provider data CSRD-audit-ready?

3. Do you have a strategy for dealing with hidden Scope 3 emissions?

The answers to these questions determine whether your cloud strategy becomes a sustainability engine or a compliance risk.

The future belongs not to companies with the lowest emissions, but to those with the most transparent measurements. Real progress begins with honest accounting – everything else is just marketing.