GreenOps: AI to Green IT Under Certain Conditions

The application of AI to IT operations promises more efficient resource management and cost savings. However, its deployment comes with financial and environmental costs, necessitating a solid use case and reserving AI for problems where it is essential, according to expert Anubha Jain.

The Cost of AI Requires Strict and Measurable Framing

The implementation of GreenAIOps approaches first encounters the cost of AI. Developing, training, and testing models for sustainability is expensive, and there are also execution and maintenance costs. These costs are both financial and environmental, as AI consumes significant amounts of electricity and water. To avoid overly resource-intensive models, Anubha Jain recommends involving end-users and clearly defining performance benchmarks, data requirements, and computational constraints. She advises building a business case that quantifies savings from reduced energy and waste, incorporating benefits related to reputation, competitive positioning, and the ability to meet potential regulatory obligations. Before any launch, she urges checking whether AI is essential or if less resource-intensive solutions can achieve the same results.

AI Levers to Manage Load, Maintenance, and Energy

The operational value of AI in GreenOps lies in real-time management of data center loads and capacities. Through predictive analytics, unused components can be shut down, improving server utilization while reducing electricity and cooling needs. In the cloud, dynamic allocation anticipates priority needs and assigns the most efficient resources, leading to expected reductions in consumption and waste. AI can also analyze sensor data to predict maintenance before failures occur and closely monitor energy usage by producing immediately actionable insights, effective for managing power and cooling while minimizing costs. Finally, carbon-aware computing seeks to align computing with energy needs and availability.

Environmental Impacts of IT Directly Targeted by GreenOps

Traditional IT operations have a heavy environmental footprint. They heavily mobilize electricity and water and emit gases, increasing pressure on the grid and raising costs for all users. The water needed for cooling reduces the supply available for agriculture, while carbon emissions rise. These effects are exacerbated by the addition of servers and the expansion of data centers, especially when inefficient legacy systems run continuously. Beyond operations, the manufacturing of hardware depletes rare minerals, and rapid replacement cycles increase end-of-life waste.

Potential and Limits: A More Proactive AI to Use Sparingly

Anubha Jain believes that AI can shift sustainability from a reactive logic to a proactive, predictive, and intelligent approach. Among the promises mentioned are the identification of inefficient hardware to be removed and the prevention of failures that could increase waste and consumption. While AI has contributed to an overall increase in carbon footprint, it also has the capacity to reduce it. For the researcher, the goal is not to apply AI everywhere but to use it responsibly and judiciously, with environmental benefits outweighing the costs. This debate occurs in a context where sustainability is among the major challenges for IT leadership.

Operational Goals of GreenOps in the Enterprise

At the organizational level, GreenOps mobilizes AI to optimize IT operations for better sustainability. The ambition focuses on reducing the environmental footprint, improving resource management, and seeking operational efficiency, with cost savings when technical and governance conditions are met.