NeuroScaler: AI-native Energy-Optimal Autoscaling for Container-Based Services System

NeuroScaler, presented by Alisson O. Chaves et al., introduces an AI-native, energy-efficient, carbon-aware orchestrator for cloud and edge containerized services.

• Main announcement:NeuroScaler aggregates multi-tier telemetry (from PDUs through bare-metal servers to containerized infrastructure managed by Kubernetes) and uses machine-learning pipelines plus a model-predictive control policy to optimize energy while meeting service-level objectives; evaluated on a real testbed and shown to reduce energy consumption by 34.68% compared to the Horizontal Pod Autoscaler (HPA) while maintaining target latency.
• Background and details: Submitted to arXiv (arXiv:2602.08191) on 9 Feb 2026 by Rodrigo Moreira et al.; includes links to PDF, HTML, and TeX source; licensed under CC BY 4.0. The system links load, performance, and power metrics across tiers and targets cloud and edge deployments (no monetary figures or deployment timelines disclosed).