ESG March 18, 2025 · BMF Services Editorial Team

Sustainable Cloud Architecture: Building for ESG Commitments

Cloud computing accounts for an estimated 2–4% of global greenhouse gas emissions, a figure growing at 9% annually as enterprise workloads continue migrating off-premises. For organizations with ESG (Environmental, Social, and Governance) commitments — whether voluntary or regulatory — IT carbon footprint is no longer a sustainability team's concern. It is an architecture concern.

Why IT Carbon Footprint Matters for ESG

ESG reporting frameworks increasingly require disclosure of Scope 3 emissions — indirect emissions from the value chain, which includes cloud infrastructure. The EU's Corporate Sustainability Reporting Directive (CSRD), effective in phases from 2024, mandates detailed sustainability reporting for approximately 50,000 companies operating in the EU. Similar requirements are emerging in the UK, Japan, and California.

For enterprises under these regimes, cloud provider emissions are part of the compliance picture. But beyond compliance, there is a genuine business case: the architecture decisions that reduce carbon emissions also reduce cloud spend. Sustainability and cost optimization are aligned, not opposed.

Measuring Your Cloud Carbon

You cannot reduce what you do not measure. All three major cloud providers now offer carbon measurement tools:

AWS Customer Carbon Footprint Tool. Provides estimated Scope 1 and Scope 2 emissions for your AWS usage, broken down by service and region. Data is available via console and API, enabling integration into dashboards and reporting workflows.
Microsoft Emissions Impact Dashboard (Azure Sustainability Calculator). Estimates carbon emissions from Azure usage with breakdowns by subscription, resource group, and service. Includes forward-looking projections based on current usage trends.
Google Cloud Carbon Footprint. Reports gross and net carbon emissions (accounting for Google's renewable energy matching) with regional granularity and hourly-level detail via BigQuery export.

For multi-cloud environments, third-party tools like Cloud Carbon Footprint (open source) and Finch provide cross-platform aggregation. The numbers are estimates, not audits, but they are accurate enough to identify hotspots and track trends.

Green Architecture Patterns

Right-Sizing and Resource Efficiency

The single largest source of cloud carbon waste is over-provisioned compute. VMs running at 10–15% CPU utilization are burning energy for work they are not doing. Implement auto-scaling policies that scale to zero for non-critical workloads, use rightsizing recommendations from your cloud provider, and enforce resource quotas per team or project. Container orchestration with proper resource requests and limits prevents the silent sprawl that inflates both cost and emissions.

Region Selection for Renewable Energy Grids

Not all cloud regions are created equal from a carbon perspective. AWS us-east-1 (Northern Virginia) runs on a different energy mix than AWS eu-west-1 (Ireland) or AWS ap-southeast-1 (Singapore). Google Cloud and Microsoft Azure publish region-level carbon intensity data. Choosing a region with a higher renewable energy percentage can reduce your workload's carbon footprint by 30–50% without changing a line of code.

Latency requirements sometimes constrain region choice, but for batch processing, data warehousing, ML training, and backup workloads, there is rarely a technical reason to avoid low-carbon regions.

Serverless Adoption

Serverless architectures (AWS Lambda, Azure Functions, GCP Cloud Run) reduce carbon through improved utilization. In a serverless model, you share underlying infrastructure with other tenants, and the cloud provider optimizes packing density across all customers. The result is higher aggregate utilization and less idle compute. For event-driven workloads with variable traffic, serverless consistently produces lower emissions than always-on VMs or containers.

Data Lifecycle Policies That Reduce Compute Waste

Data storage has a compute footprint. Every byte stored is replicated, backed up, indexed, and scanned by security tools. Aggressive data lifecycle management reduces this footprint:

Tiering. Move data that is rarely accessed to cold storage (S3 Glacier, Azure Archive, GCP Coldline). These tiers use less energy-intensive storage media and often run on fewer active servers.
Retention policies. Define and enforce data retention schedules. Deleting data that has passed its legal or business retention period is the most effective way to reduce storage-related emissions.
Log management. Structured logging at DEBUG level in production generates terabytes of data that nobody reads. Implement log level policies, aggregate and sample, and set aggressive retention windows for operational logs.
Query optimization. Inefficient queries that scan full tables instead of partitions consume more compute cycles and therefore more energy. This is particularly impactful in big data platforms where a single poorly partitioned query can process petabytes unnecessarily.

CSRD and Regulatory Reporting

The CSRD requires companies to report on how their operations affect the environment (outside-in) and how sustainability issues affect the company (inside-out). IT departments contribute to both:

Energy consumption and greenhouse gas emissions from data centers (Scope 3 for the reporting company, Scope 1/2 for the cloud provider)
Climate-related risks to IT operations (physical risks to data centers, transition risks from carbon pricing)
Technology choices that enable sustainability in other business functions (smart building systems, supply chain optimization, remote work infrastructure)

Organizations subject to CSRD need to begin collecting and auditing IT emissions data now. The first reporting periods are already active for the largest companies, with mid-size enterprises following in 2025–2026.

The Business Case: Sustainability and Cost Go Together

The most important message for engineering leaders is this: green architecture is good architecture. The practices that reduce carbon emissions — right-sizing, auto-scaling, serverless adoption, data lifecycle management, query optimization — are the same practices that reduce cloud costs, improve system reliability, and simplify operations. You are not choosing between sustainability and performance. You are choosing between wasteful and efficient systems.

The Bottom Line

Sustainable cloud architecture is not a separate discipline. It is cloud architecture done well, with carbon emissions added to the optimization function alongside cost, latency, and availability. Start by measuring your baseline. Identify your highest-emission workloads. Apply the same engineering rigor to carbon reduction that you apply to cost optimization. And report your progress — not just for compliance, but because transparency drives accountability and accountability drives results.

Need help applying these patterns? Contact us for a free consultation →