The Unseen Bill: Uncovering Cross‑Layer Cost Externalities in AI‑Driven AWS Rightsizing and Their Mitigation through Policy‑Based Guardrails
DOI:
https://doi.org/10.63282/3050-9416.IJAIBDCMS-V7I1P146Keywords:
Cloud Cost Optimization, Finops, AWS Rightsizing, AI‑Driven Automation, Cost Externalities, Shared Responsibility Economics, Policy‑Based Guardrails, Cross‑Layer Billing AnomaliesAbstract
We’ve all been there, you run an AI‑based rightsizing tool to trim down your EC2 fleet, and somehow next month’s bill is higher, not lower. Digging in, you realize the savings from compute got eaten up by a spike in data transfer or a sudden jump in EBS IOPS. That’s the cost‑shifting problem nobody talks about. In this paper, we analyze three months of real‑world AWS usage telemetry from a mid‑sized e‑commerce deployment. We show that automated rightsizing decisions often create hidden cross‑layer cost externalities especially between compute, storage, and network egress. These aren’t bugs; they’re structural side effects of how AI optimizers are currently scoped (single service, short horizon). The good news? You don’t have to turn off automation. We test a set of lightweight, policy‑based guardrails like interdependency budgets and service‑layer cost caps that cut these unintended cost shifts by over 40% without sacrificing the original rightsizing gains. Our findings suggest that effective FinOps isn’t just about smarter AI; it’s about constraining the AI with cross‑layer rules that mirror how real cloud bills actually add up.
References
1. N. P. A. Vo, S. Thakur, A. Kumar, and R. Singh, "FinOps agent — A use-case for IT infrastructure and cost optimization," arXiv preprint arXiv:2510.25914, Oct. 2025.
2. D. Bodra and S. Khairnar, "Machine learning-based cloud resource allocation algorithms: A comprehensive comparative review," Frontiers in Computer Science, vol. 7, p. 1678976, 2025, doi: 10.3389/fcomp.2025.1678976.
3. E. I. Zatsarinnaya, S. V. Markova, and D. A. Krymshokalova, "Impact of cloud computing on enterprise cost structure: Shift from CAPEX to OPEX," Ekonomika, Upravlenie i Pravo: Innovatsionnoe Reshenie, vol. 8, no. 5, pp. 30–38, May 2025. DOI: 10.36871/ek.up.p.r.2025.05.08.004.
4. T. Mahmood and M. Lacity, "Cloud computing economics: A review and research agenda on FinOps," Journal of Information Technology, vol. 37, no. 4, pp. 438–466, Dec. 2022.
5. K. Rajamani, "Predictive hybrid autoscaling for cloud workloads: A machine learning approach to vertical and horizontal resource optimization on AWS EC2," Journal of Information Systems Engineering and Management, vol. 10, no. 8, pp. 1–12, Nov. 2025.
6. Y. Zhang and H. Li, "Predictive auto scaling and cost optimization using machine learning in AWS cloud environments," in Proc. 2025 ACM Symp. Cloud Comput. (SoCC '25), Seattle, WA, USA, Dec. 2025, pp. 1–8. doi: 10.1145/3772326.3774726.
7. M. Xu et al., "Auto-scaling approaches for cloud-native applications: A survey and taxonomy," arXiv preprint arXiv:2507.17128, Jul. 2025.
8. S. Muthukumarasamy and P. Senthilkumar, "A review of AI-driven techniques for cost optimization in Kubernetes environments," in Proc. 2025 3rd Int. Conf. on Adv. Comput. (ICAC 2025), Colombo, Sri Lanka, Apr. 2025, pp. 1–6.
9. Y. Wang and V. Liaskos, "The non-expert tax: Quantifying the cost of auto-scaling in cloud-based data stream analytics," in Proc. Int. Workshop on Big Data in Emergent Distributed Environments (BiDEDE '22), Philadelphia, PA, USA, Jun. 2022, pp. 1–6. doi: 10.1145/3530050.3532925.
10. L. Kondrashov, B. Zhou, H. Wang, and D. Ustiugov, "The high cost of keeping warm: Characterizing overhead in serverless autoscaling policies," arXiv preprint arXiv:2509.03104, Sep. 2025.
11. J. Chamberlain, J. Zheng, Z. Zhu, Z. Liu, and D. Starobinski, "Exploiting Kubernetes autoscaling for economic denial of sustainability," Proc. ACM Meas. Anal. Comput. Syst., vol. 9, no. 2, pp. 1–25, Jun. 2025. doi: 10.1145/3727114.
12. A. Q. Khan, M. Matskin, R.-C. Prodan, C. Bussler, D. Roman, and A. Soylu, "Cost modelling and optimisation for cloud: A graph‑based approach," Journal of Cloud Computing, vol. 13, p. 147, 2024. doi: 10.1186/s13677-024-00684-6.
13. M. O. Diaz, "Mitigating the cost amplification of platform defaults in lean cloud data deployments through configuration hygiene," Transaction on Informatics and Data Science, vol. 2, no. 2, pp. 1–12, 2025. doi: 10.24090/tids.v2i2.14737.
14. S. Deochake, "ABACUS: A FinOps service for cloud cost optimization," arXiv preprint arXiv:2501.14753, Jan. 2025.
