Moving to the cloud serves as a driving force for business transformation. However, when such a significant change occurs, environmental consequences are too often overlooked.
Resource waste is generally linked to a lack of maturity in cloud service management but also associated with new uses such as the rise of AI services. These elements have a significant impact on cost explosion and energy consumption. Given this reality, here are some reflections and best practices to implement to minimize environmental footprint while meeting sustainability goals.
The environmental dilemma of the cloud
When we speak of cloud waste, we're essentially talking about poorly used or unused resources that consume energy without creating value. Data centres already account for nearly 4% of global carbon emissions, a figure that continues to increase with the rise of artificial intelligence. There are currently more than 250 data centres in France, and this number is expected to double by 2030.
With the growth of artificial intelligence, this issue has shifted from concerning to fundamental. The energy requirements associated with new digital uses related to artificial intelligence (Gen AI, LLM, etc.) will considerably amplify the problem, as energy consumption increases exponentially without proportional value creation.
Implementing impact measurement tools
At Sopra Steria, we rely on the OPTE model to structure our approach:
- Organization: to align our clients' governance structures with regulations to prioritize CSR improvement areas.
- Process: to integrate best ecological practices while respecting eco-design frameworks.
- Technology: to industrialize the collection of footprint measurement data.
- Eco: to support organizations in a holistic approach.
The OPTE methodology provides a structuring framework for addressing digital sobriety, but its effectiveness lies in its concrete day-to-day application. This is precisely where the strategic combination of FinOps and GreenOps comes into play.
These two approaches act as two sides of the same coin: FinOps optimizes financial costs by eliminating superfluous expenses, while GreenOps reduces ecological impact by minimizing consumed resources. This natural complementarity creates a virtuous circle: what's good for the planet also proves beneficial for financial results.
To implement this dual approach, Sopra Steria's teams rely on innovative infrastructure management platforms (on-premise, private, hybrid, or public). These tools integrate sophisticated observability solutions, automation, and infrastructure lifecycle management for FinOps and GreenOps, for a more sustainable approach. The result is tangible: optimized operational costs and a simultaneous reduction in carbon footprint.
From theory to practice: The importance of acting now
The complexity of the subject may seem discouraging, but inaction is no longer an option. Companies still hesitating to engage in this path must become aware of two unavoidable realities. Given this complexity, what concrete actions can organizations take? Expert guidance is essential, given the significant risks involved.
First, the regulatory landscape is evolving rapidly. With the implementation of the CSRD (Corporate Sustainability Reporting Directive) and the development of the European taxonomy on CSR, the requirements for transparency and environmental performance are becoming increasingly stringent. The risk of non-compliance, with its financial and reputational consequences, is very real.
Second, the market itself is constantly changing. In our calls for tender, we notice that questions related to energy consumption now represent more than 20% of our clients' requirements. This trend will only intensify in the coming years.
To navigate this new environment, having guidance becomes not a luxury, but a strategic necessity. The technical, methodological, and regulatory expertise provided by specialists often makes all the difference between a successful transformation and a missed opportunity.
However, some concrete actions can be implemented by companies without reasonable effort:
- Governance: Deploy a global strategy for reducing costs related to the company's CSR strategy applicable to the information system, including the cloud.
- Renewable Energy: Choose suppliers that use data centers powered by renewable energy.
- Eco-design: Rely on an eco-design approach from the design phase of cloud services, as part of a cloud implementation or transformation.
- Intelligent Orchestration: Promote intelligent resource allocation (IaaS/PaaS) by implementing resource lifecycle automation (Infra As Code, Terraform for example), all associated with the flexibility offered by the cloud (right sizing/auto scaling).
- Resource Lifecycle Management: Implement automatic decommissioning strategies for unused instances and adjust capacities according to demand. Deploy good "tagging" rules and quotas on services associated with the organizations that consume them for precise real-time monitoring.
- Virtualization and Containerization: Promote the deployment of virtualization and especially containerization (Docker, Kubernetes) to maximize resource utilization.
- Storage Optimization: Limit excessive data redundancy and archive efficiently to reduce used space, implement a hierarchical storage approach based on usage (tiering).
- Cloud carbon calculator: Use tools to measure and adjust carbon emissions.
- FinOps: Implement a tooled FinOps strategy allowing fine cost management.
If we focus on AI, which marks a major change in cloud usage, improvement areas mainly revolve around the following axes:
- Reduction of computational load by optimizing AI models (quantization, pruning, distillation).
- Development of lightweight models: Use more efficient algorithms such as sparse AI models to reduce electricity consumption.
- Edge AI deployment: Performing AI processing on local devices rather than in the cloud, to limit energy-intensive transfers.
- Inference at scale: Implementing inference supervision tools to manage the performance/price/energy impact ratio of scaling AI models.
It should be noted that data center providers and hyperscalers have already initiated actions to reduce their environmental impact in the following areas:
- Use of renewable energy.
- Manufacture of used infrastructures, via a limitation of suppliers.
- Recycling of hardware platforms.
- Longer maintenance of platforms.
- Reduction of energy costs (water cooling to ensure machine cooling...).
- More efficient waste management practices.
In the race for digital sustainability, success depends on precise measurement and strategic action. Companies that master both will gain a decisive competitive advantage.