Rethinking Cloud Computing Through a Sustainability Lens

IA FORUM MEMBER INSIGHTS: ARTICLE

By Sudipta Dey, Cloud Infrastructure Modernization Lead, ELSEVIER

Sustainability and business efficiency are no longer separate goals - in modern cloud computing, they go hand in hand.

Green Cloud Computing: Sustainability by Design

The cloud may seem borderless and invisible - but in my experience, that could not be further from the truth. Every enterprise workload sits on top of a very physical reality: millions of servers crammed into massive data centers, running around the clock and burning through staggering amounts of power.

Data centers today account for nearly 3 - 4% of global electricity consumption, and that number is only heading in one direction. Cloud workloads keep expanding, and with AI inference set to significantly drive-up energy demand at major cloud providers by 2027, I think it’s clear that sustainability can no longer be treated as an afterthought - it should be a core business priority.

What I have come to realize is that true sustainability is not something you can just buy your way by shifting workloads to a renewable-powered cloud provider and calling it a day. If the underlying architecture is inefficient, you are largely just moving the problem around. The way I see it, sustainability must be approached holistically spanning infrastructure, applications, operations, and the design decisions that shape all of them.

Why Sustainability Matters in Technology

We all care about nature and the environment, and we all share a responsibility to preserve it for future generations. But how do we contribute when we spend eight or more hours every day working on laptops, applications, and cloud systems? The answer lies in the choices we make while designing and operating technology systems.

The way I see it, sustainability in cloud computing is not just about environmental responsibility - it is equally about building systems that are efficient, scalable, and cost-effective. Every infrastructure decision, architecture pattern, and deployment strategy either adds waste or cuts it.

What I have observed is that teams design applications with one goal: get to the cloud. Operational costs, energy consumption, and long-term environmental impact rarely enter the conversation early enough, and sustainability gets quietly pushed aside during planning.

By the time teams wake up to it, they’re already dealing with:

• Rising cloud operating costs

• Excessive server utilization

• Energy inefficiencies

• Increased carbon emissions

• Rework and redesign efforts

• Budget constraints and shifting priorities

Better architectural decisions made at the start could have significantly reduced both the environmental footprint and operational costs. Instead, teams end up paying for shortcuts they did not even realize they were taking.

Pillars of Green Cloud Computing

In my opinion, a mature green cloud strategy rests on interconnected pillars and organizations that address only one while ignoring the others rarely achieve anything meaningful on the environmental front.

• Energy Efficiency: For me, this starts at the foundation maximizing useful computation per watt through hardware efficiency, workload consolidation, virtualization density, and intelligent power management across chip, server, and rack levels.

• Renewable Energy Sourcing: Aligning data center electricity consumption with credible renewable generation, whether through direct power purchase agreements, on-site solar or wind installations, or hourly matched energy attribute certificates rather than annual averages, in my view, that can obscure underlying fossil fuel usage.

• Carbon-Aware Workload Scheduling: One practical shift I advocate for is moving flexible workloads to times and regions where the grid is genuinely cleaner, guided by real-time carbon intensity signals, not just assumptions.

• Efficient Software Architecture: I firmly believe the most sustainable resource is the one never consumed. That means designing efficiency from day one cutting over-provisioning, maximizing caching, and eliminating technical debt that drives unnecessary computation.

• Sustainable Hardware & Resource Stewardship: This may get overlooked, but in my experience, it matters extending server lifespans, responsibly recycling retired hardware, and choosing facilities with strong environmental and operational governance standards.

Sustainability Must Be Part of the Roadmap

In some technology discussions I have been part of, business vision, technical requirements, and problem statements receive thorough attention. Sustainability and environmental governance, however, tend to be missing from the conversation entirely, and I think that is worth addressing.

This gap persists despite clear industry direction. Major cloud providers have already embedded sustainability as a core pillar within their well-architected frameworks, yet some organizations still treat it as optional during solution design, revisiting it only when costs rise or compliance requires it.

In my view, that approach carries a real cost. Sustainability decisions made late in the process are significantly more expensive to implement than those considered from the start. Embedding sustainable thinking into architecture planning and operational governance early on is both more effective and more efficient.

What I find genuinely encouraging, though, is that sustainability and business performance reinforce each other. Efficient architectures consume less energy, reduce cloud spending, and free up budget for innovation. In my experience, what is good for the environment is also good for business.

Benefits of Sustainable Cloud Computing

From my experience, adopting sustainable cloud practices is not a tradeoff; it delivers meaningful benefits on both fronts.

On the environmental side, what I consistently see is a lower carbon footprint, reduced energy consumption, improved efficiency, and more responsible use of computing resources overall.

The business case, in my view, is equally compelling. Organizations that embrace sustainable practices tend to benefit from lower cloud operating costs, better infrastructure utilization, long-term operational efficiency, and freeing up a budget that can be redirected toward innovation and future initiatives.

What I find particularly valuable is the compounding effect: infrastructure optimization drives cost savings, and those savings create the financial headroom to fund modernization efforts and new investments. It is a cycle that rewards getting it right early.

Designing for Sustainability

In my view, sustainability should be one of the core goals of infrastructure and application design and not an afterthought. Here are a few practical approaches that organizations can consider:

• Choose Low-Carbon Data Centers: Where feasible, I would encourage deploying workloads in regions or data centers powered by renewable energy or with lower carbon intensity.

• Design Green Cloud Architectures: Build architectures that minimize unnecessary resource consumption and prioritize operational efficiency from the outset.

• Adopt Serverless & Event-Driven Architectures: These consume resources only when needed, which in my experience meaningfully reduces idle infrastructure and energy waste.

• Rightsized Infrastructure: Overprovisioning is one of the most common inefficiencies I see. Selecting VM sizes based on actual workload requirements goes a long way.

• Use Spot Instances Where Appropriate: For non-critical workloads, spot instances are a practical way to reduce both costs and unused capacity.

• Enable Auto Scaling & Load Balancing: Dynamic scaling ensures resources are allocated only when demand genuinely exists, cutting idle compute consumption.

• Implement Tiered Storage: Keep frequently accessed data in high-performance storage and move archival data to low-cost, energy-efficient tiers. It is a simple practice that adds up.

• Use Energy-Efficient Compute Options: I would particularly highlight options like AWS Graviton-based instances and similar offerings from other cloud providers as they deliver strong performance at lower energy consumption.

• Align Virtual Machines Selection with Use Cases: Not every workload needs high-performance infrastructure. Matching the right virtual machines type to each workload meaningfully improves utilization efficiency.

• Periodically Revisit Architectures: Sustainability is not a one-time exercise. Regular architecture reviews, in my experience, consistently surface optimization opportunities that would otherwise go unnoticed.

• Stop Unused Resources: Virtual machines and environments that are not needed after hours or over weekends should be shut down automatically. It is one of the simplest wins available, and it is often overlooked.

The Role of AI in Sustainable Operations

When used responsibly, I believe AI can be a meaningful contributor to sustainability. AI-powered developer productivity tools accelerate software delivery, eliminate repetitive tasks, and reduce the operational effort that tends to spill into evenings and weekends. Shorter delivery cycles mean development and testing environments no longer need to run continuously outside business hours thereby creating natural opportunities to shut down non-production infrastructure during idle periods, reduce unnecessary compute consumption, and lower operational costs.

AI workloads are energy-intensive, and the productivity gains they enable only translate into genuine sustainability wins when paired with deliberate, efficient infrastructure planning.

The goal, as I see it, is not simply to adopt AI. It is to adopt it in a way where the efficiency it creates meaningfully outweighs the energy it consumes.

How AI is Accelerating Green Cloud

What I find genuinely fascinating is the irony at the heart of the AI sustainability story: the very technology driving data center growth is also one of the most powerful tools we have for reducing it.

• Intelligent Cooling: A compelling example, to me, is Google DeepMind’s reinforcement learning system, which reduced data center cooling energy by roughly 30% by dynamically adjusting fan speeds, water set points, and airflow in real time. It identified non-obvious interactions across dozens of variables simultaneously - optimizations that human operators simply couldn’t achieve at that scale.

• Workload Prediction: AI models can forecast compute demand hours in advance, enabling workloads to be scheduled during periods of lower grid carbon intensity. In my view, this is particularly valuable as it smooths demand curves that would otherwise require carbon-intensive standby capacity to remain on call.

• Efficient Model Design: Techniques like knowledge distillation can reduce inference energy consumption by an order of magnitude with minimal impact on accuracy. As energy cost becomes a primary engineering constraint, I have noticed that the field of Efficient AI is maturing rapidly and making sustainability not just an environmental consideration, but a core design criterion on its own right.

Balancing Sustainability & User Experience

The most energy-efficient, low-carbon data centers are not always located where users are - and routing traffic across continents to reach them can introduce latency that meaningfully degrades the user experience.

In my view, the answer is not to choose between sustainability and performance. With thoughtful design, that trade-off rarely needs to be made at all. Approaches like edge delivery, intelligent workload placement, caching strategies, and distributed architectures allow organizations to serve users responsively while still directing compute toward cleaner, more efficient infrastructure.

Where Can Organizations Choose to Start?

In my experience, sustainability doesn’t require a perfect starting point. Teams can begin at any stage - whether launching a new solution, modernizing existing applications, or running infrastructure and cost optimization reviews. Even mature, well-established environments hold untapped opportunities.

That said, I have seen the same barriers come up repeatedly: difficulty quantifying environmental impact, organizational inertia, and a lack of carbon-aware tooling in everyday engineering workflows. A structured, phased approach, in my view, is the most practical way to move forward without requiring everything to change at once.

• Phase 1 - Measure: I will always start here. Deploy cloud provider-native carbon tools to establish a baseline and identify the highest-emission workloads. Optimization without measurement is guesswork - knowing where emissions originate is the essential first step.

• Phase 2 - Optimize: With a baseline in place, focus on high-impact, low-effort changes: rightsizing instances, enabling auto-scaling with scale-to-zero, applying storage lifecycle policies, and shifting eligible workloads to ARM-based processors. In my experience, these actions alone can reduce emissions by 20 - 40% and build momentum quickly.

• Phase 3 - Shift:  Once the foundations are solid, go further. Integrate carbon intensity APIs into batch schedulers to route flexible workloads toward cleaner grid regions and evaluate serverless architectures for stateless services to eliminate idle resource consumption at its source.

But I would also add that technical change alone is not enough; cultural change matters just as much. What I have consistently observed is that teams treating carbon as a first-class engineering metric, alongside latency and cost, meaningfully outperform those approaching sustainability as a compliance checkbox. The mindset, in my experience, shapes the outcome.

Conclusion

In my view, green cloud computing is about more than reducing environmental impact. At its core, it is about building technology systems that are smarter, more efficient, and financially sustainable over the long term.

What I firmly believe is that sustainability belongs at the beginning of every architecture conversation and not at the end. The decisions made early in a project about how infrastructure is sized, where workloads run, how resources are managed quietly determine energy consumption, cloud costs, and operational efficiency for years to come. Getting those decisions right from the start is far less costly, in my experience, than revisiting them later.

Organizations that embed sustainable practices into their cloud strategy, I have observed, stand to gain across multiple dimensions: lower emissions, reduced operational costs, better infrastructure utilization, and greater financial flexibility to invest in innovation.

Author Disclaimer: The views and opinions expressed herein are those of the Author alone and are shared in a personal capacity, in accordance with the Chatham House Rule. They do not reflect the official views or positions of the Author’s employer, organization, or any affiliated entity.