AI’s Growing Demand For Resources Is Unsustainable: NTT Data Study
Researchers predict AI workloads will drive more than 50% of data centre power consumption by 2028.
There is an urgent need to embed sustainability into every layer of AI development and deployment to counteract the technology’s environmental impact, a new study from NTT Data shows. The study illustrates the growing environmental impact of AI and outlines a path to sustainable innovation.
The technology requires enormous volumes of electricity to support surging computational demands to train large language models, run inference pipelines, and maintain always-on services. Researchers predict AI workloads will drive more than 50% of data centre power consumption by 2028.
Other primary environmental impacts include water consumption for data centre cooling systems, e-waste, and rare earth mineral extraction for hardware production.
Key Insights
Key study insights include:
NTT Data’s AI experts and sustainability consultants urge the use of holistic sustainability goals, not just conventional AI performance metrics such as accuracy and speed. Efficiency must be a core design principle.
AI’s energy consumption, carbon emissions, and water footprint need standard and verifiable metrics. Industry benchmarks such as the “AI Energy Score” and “Software Carbon Intensity for AI” offer ways to embed sustainability into governance, procurement, and compliance protocols.
Sustainable AI requires lifecycle thinking, from raw material extraction and hardware production to system deployment and ultimate disposal. Important steps include lengthening hardware lifespans, optimising cooling systems and applying circular-economy principles.
The responsibility for sustainability should be taken up by hardware manufacturers, data centre operators, software developers, cloud providers, policymakers, investors, and consumers, the study advocates.
Barriers And Best Practices
Many organisations today focus narrowly on energy or emissions without considering water usage, rare material depletion, and e-waste. These and other factors must be addressed comprehensively. Even when environmental goals are set, organisations often lack actionable methods to apply sustainability in the AI lifecycle.
To address these concerns, the report outlines best practices:
Applying green software engineering patterns to reduce resource consumption.
Running AI workloads in locations and at times that align with renewable energy availability.
Leveraging remote GPU services and on-premises AI.
Reducing e-waste by prioritising modular and upgradable components, and extending hardware lifespans through refurbishment, reuse and responsible recycling.
