INDIA – August 5, 2024 – A new study commissioned by Amazon Web Services (AWS) and completed by Accenture shows that an effective way to minimise the environmental footprint of leveraging Artificial Intelligence (AI) is by moving IT workloads from on-premises infrastructure to AWS cloud data centres in India and around the globe. Accenture estimates that AWS’s global infrastructure is up to 4.1 times more efficient than on-premises. For Indian organisations, the total potential carbon reduction opportunity for AI workloads optimised on AWS is up to 99% compared to on-premises data centres.
The research states that simply utilizing AWS data centres for compute-heavy, or AI, workloads in India yields a 98% reduction in carbon emissions compared to on-premises data centres. This is credited to AWS’s utilisation of more efficient hardware (32%), improvements in power and cooling efficiency (35%), and additional carbon-free energy procurement (31%). Further optimising on AWS by leveraging purpose-built silicon can increase the total carbon reduction potential of AI workloads to up to 99% for Indian organisations that migrate to and optimise on AWS.
“Considering 85% of global IT spend by organisations remains on-premises, a carbon reduction of up to 99% for AI workloads optimised on AWS in India is a meaningful sustainability opportunity for Indian organisations,” said Jenna Leiner, Head of Environment Social Governance (ESG) and External Engagement, AWS Global. “As India accelerates towards its US$1 trillion-dollar digital opportunity and encourages investments into digital infrastructure, sustainability innovations and minimising IT related carbon emissions will be critical in also helping India meet its net-zero emissions by 2070 goal. This is particularly important given the rising adoption of AI. AWS is constantly innovating for sustainability across our data centres —optimising our data centre design, investing in purpose-built chips, and innovating with new cooling technologies – so that we continuously increase energy efficiency to serve customer compute demands.”
“This research shows that AWS’s focus on hardware and cooling efficiency, carbon-free energy, purpose-built silicon, and optimized storage can help organizations reduce the carbon footprint of AI and machine learning workloads,” said Sanjay Podder, global lead for Technology Sustainability Innovation at Accenture. “As the demand for AI continues to grow, sustainability through technology can play a crucial role in helping businesses meet environmental goals while driving innovation.”
Sustainable chip technology innovation – purpose-built silicon
One of the most visible ways AWS is innovating for energy efficiency is through the company’s investment in AWS chips. Launched in 2018, the custom AWS-engineered general purpose processor, Graviton, was the first-of-its-kind to be deployed at scale by a major cloud provider. The latest Graviton4 offers four times the performance of Graviton, and while Graviton3 uses 60% less energy for the same performance as comparable Amazon EC2 instances (where the compute happens in a data centre), Graviton4 is even more energy efficient.
AWS customers are also benefiting from the carbon reduction potential of Graviton. Paytm, India’s leading payments and financial services distribution platform, witnessed a reduction in workload carbon intensity by adopting Graviton processors, reporting up to 47% estimated decrease in carbon emissions per transaction. Similarly, IBS Software, a leading SaaS solutions provider to the global travel industry, reported that other than improving performance and reducing cost by adopting Graviton processors, the company saw a 40% reduction in carbon emissions per instance hour.
Running generative AI applications in a more sustainable way requires innovation at the silicon level with energy efficient hardware. To optimise performance and energy consumption, AWS developed purpose-built silicon like the AWS Trainium chip and AWS Inferentia chip to achieve significantly higher throughput than comparable accelerated compute instances. AWS Trainium cuts the time taken to train generative AI models—in some cases from months to hours. This means building new models requires less money and power, with energy-consumption reductions of almost one third/up to 29%. AWS Inferentia is AWS’s most power-efficient machine learning inference chip. AWS Inferentia2 machine learning accelerator delivers up to 50% higher performance per watt and can reduce costs by up to 40% against comparable instances. These purpose-built accelerators enable AWS to efficiently execute AI models at scale. This translates to a reduced infrastructure footprint for similar workloads, resulting in enhanced performance per watt of power consumption.
Improving energy efficiency across AWS infrastructure
Through innovations in engineering—from electrical distribution to cooling techniques, AWS’s infrastructure is able to operate closer to peak energy efficiency. AWS optimises resource utilisation to minimise idle capacity, and continuously improves the efficiency of its infrastructure. For example, AWS removed the large central Uninterruptible Power Supply (UPS) from its data centre design to instead use small battery packs and custom power supplies that AWS integrates into every rack, which has improved power efficiency and has further increased availability. Every time power is converted from one voltage to another, or from AC to DC and vice versa, some power is lost in the process. By eliminating the central UPS, AWS are able to reduce these conversions. Additionally, AWS have optimised rack power supplies to reduce energy loss in that final conversion. Combined, these changes reduce energy conversion loss by about 35%.
After powering AWS’s server equipment, cooling is one of the largest sources of energy use in AWS data centres. To increase efficiency, AWS uses different cooling techniques, including free air cooling depending on the location and time of year, as well as real-time data to adapt to weather conditions. Implementing these innovative cooling strategies is more challenging on a smaller scale at a typical on-premises data centre. AWS’s latest data centre design seamlessly integrates optimised air-cooling solutions alongside liquid cooling capabilities for the most powerful AI chipsets, like the NVIDIA Grace Blackwell Superchips. This flexible, multimodal cooling design allows AWS to extract maximum performance and efficiency whether running traditional workloads or AI models.
According to the study, AWS’s additional carbon-free energy procurement in India contributes 31% in carbon emissions reduction for compute-heavy workloads and 44% for storage-heavy workloads. Aligning with Amazon’s commitment to achieving net-zero carbon emissions across all operations by 2040, AWS is rapidly transitioning its global infrastructure to match electricity use with 100% carbon-free energy.