Despite ambitious sustainability pledges, carbon emissions are on the rise. Can AI help fix the problem it is contributing to?
“The actual emissions from data centers can be up to seven times higher than what’s reported.”
Microsoft and Google began publishing environmental sustainability reports in 2009 and 2010, respectively. Amazon was slower and only began releasing emissions data in 2020, after immense pressure from employees, shareholders, and the general public. And yet, the actual emissions from data centers can be up to seven times higher than what’s reported.
Several critical nuances are lost in these public disclosures as hyperscalers offset their true emissions using instruments like renewable energy certificates (RECs). Their carbon accounting also lacks the required granularity – instead of disclosing their cloud division-specific or data center-level emissions, they roll up emissions data into broad, company-wide figures, blending the environmental impact of their cloud services with other major business units, like Amazon’s retail operations or Google’s AdSense. As a result, the true environmental impact of cloud infrastructure remains vague and obscure.
“Hyperscalers’ opaque practices, along with their reluctance to disclose Scope 3 emissions (supply chain-related), certainly won’t help their customers who are already under pressure to report their own environmental impact.”
Despite ambitious sustainability pledges and net-zero promises, Microsoft and Google have respectively reported a 30% and 48% rise in their carbon emissions, attributed primarily to their power-hungry AI endeavours. In contrast, Amazon has surprisingly reported a steady decline in emissions, albeit a marginal one compared to its overall carbon footprint. Even this marginal decline could easily be attributed to Amazon’s creative accounting practices instead of any meaningful sustainability achievements.
Overall, data center electricity demand is projected to reach 945 terawatt-hours by 2030, more than double today’s levels. To put that in perspective, that’s more than the entire electricity consumption of Japan. Hyperscalers’ opaque practices, along with their reluctance to disclose Scope 3 emissions (supply chain-related), certainly won’t help their customers who are already under pressure to report their own environmental impact. Even if cloud providers aren’t fully transparent and detail-oriented in their disclosures, regulations are pushing businesses themselves to be fully accountable.
The EU’s Corporate Sustainability Reporting Directive (CSRD) went into effect in January 2024. It mandates detailed sustainability reporting for large companies operating in the EU and those with significant EU presence, with the very first reports due in 2025. Similarly, the EU Energy Efficiency Directive (EED) introduces specific requirements and reports for energy savings and efficiency in data center operations.
Amid tightening regulations, one thing is clear: AI will be central to how companies curb emissions, optimize energy consumption, and meet reporting and compliance demands. Sounds ironic, but AI can actually help offset some of its own emissions by driving efficient resource consumption, energy optimization, and enabling accurate carbon accounting.
“More than 30% of organizations are already relying on GenAI to reduce their overall greenhouse emissions.”
AI may be part of the problem, but it’s also part of the solution. Broadly speaking, it can drive business and operational efficiencies and hasten research and innovation for adopting green alternatives. Within the cloud, AI can optimize data center operations like predictive maintenance and upgradation, proper resource allocation, carbon-aware workload scheduling, and dynamic cooling systems.
More than 30% of organizations are already relying on GenAI to reduce their overall greenhouse emissions. By adopting established FinOps and GreenOps practices, powered by ML algorithms and AI models, they’re laying the groundwork for smarter, more sustainable operations. For instance, AI and predictive analytics help:
These applications don’t just support environmental sustainability, they also ensure financial viability for organizations grappling with unaffordable cloud bills and struggling to unlock the full value of their cloud infrastructure. As AI models continue to mature, more organizations will adopt data-driven, carbon-aware decision-making through AI-powered cloud management and optimization platforms that will enable them to meet both ESG targets and regulatory requirements.
The emma cloud management platform is an AI-powered, cloud-agnostic platform that allows organizations to monitor, manage, and optimize their cloud usage and costs across any number of cloud providers and regions they operate in.
With complete multi-cloud visibility into resource consumption, you can ensure your workloads and cloud decisions align with internal sustainability KPIs. And thanks to emma’s intuitive dashboard, there will be less friction and more cooperation between IT and sustainability teams.
“As AI gets greener itself through efficient architectures, model pruning, federated learning, and cleaner training windows, it will find broader acceptance and integration across ESG initiatives and strategies.”
AI’s role in enabling green cloud and sustainability initiatives will continue expanding. There’s a lot that AI-powered systems can help achieve already, and there’s so much more on the horizon. As the wave gains momentum, cloud and cloud-based service providers will start baking carbon-awareness into their APIs, schedulers, and developer tools by default. AI won’t just make suggestions, it will autonomously adjust workloads to reduce carbon emissions – like autoscaling compute across cleaner regions or deferring non-critical jobs to low-emission windows or solar hours. Within data center facilities, AI will help through smart HVAC, liquid cooling optimization, battery storage management, and power distribution tuned for real-time conditions.
As AI gets greener itself through efficient architectures, model pruning, federated learning, and cleaner training windows, it will find broader acceptance and integration across ESG initiatives and strategies. Until then, it’s important to keep perspective. Carbon dioxide can linger in the atmosphere for hundreds of years. Even if AI eventually achieves net negative emissions in the future, it won't be able to erase the damage already done.
So, it's important to start adopting green cloud now instead of waiting for hyperscalers and other cloud providers to play their part in addressing climate change. Start with the tools and best practices already within reach, and build from there.
Not sure where to start? Get in touch with our cloud experts to explore smart choices for sustainable cloud usage!