FinOps Monthly AI Edition #2

FinOpsMonthly.AI
Written By Tony Mackelworth

Your AI-Curated FinOps Briefing

#2 March 2025

FinOps Monthly AI Edition delivers the essential FinOps news, trends, and best practices - intelligently curated by AI to save you time. Each month, get a concise briefing on key industry developments, emerging technologies, and strategic insights to empower your FinOps practice.

FinOps Foundation Updates

FinOps for AI Whitepaper

The FinOps Foundation has released a comprehensive whitepaper on FinOps for AI, addressing the growing need for guidance in managing artificial intelligence and machine learning costs. This publication tackles unique AI cost challenges including token-based pricing models for Large Language Models and the substantial computational resources required for training and inference.

The whitepaper emphasizes extending FinOps principles across the entire AI lifecycle, from planning and development through deployment and optimization. It explores methodologies for AI cost planning, estimation tools, and strategies for aligning AI expenditure with business value. This resource provides essential guidance for practitioners seeking to optimize their organization's AI investments while maintaining financial control.

FinOps for SaaS Whitepaper

In March 2025, the FinOps Foundation published a new paper titled "FinOps for SaaS," expanding FinOps principles beyond traditional cloud infrastructure. This publication addresses the growing complexity of managing software-as-a-service expenditures, which often constitute a significant portion of technology budgets. The paper outlines frameworks for SaaS cost visibility, optimization techniques, and governance approaches tailored to the subscription-based software model.

Sources:

State of FinOps 2025 Report

The State of FinOps 2025 Report released in February provides critical insights into industry priorities and trends. Key findings reveal that while workload optimization and cloud waste reduction remain top concerns, organizations are increasingly prioritizing governance and policy implementation at scale to establish more sustainable cost management frameworks.

The report highlights growing concern about rapidly escalating AI and machine learning expenditures, underscoring the need for specialized FinOps strategies for AI workloads. Additionally, the concept of "Cloud+" is gaining prominence, representing an expansion of FinOps practices beyond traditional public cloud to encompass SaaS, software licensing, and AI costs. Organizations are increasingly investing in dedicated FinOps tooling platforms to address these evolving challenges.

Sources:

FinOps for AI Courses and Certifications

The FinOps Foundation is developing specialized courses and certifications focused on AI cost management. These educational resources will equip practitioners with the knowledge and skills needed to navigate the unique financial challenges of AI workloads. The curriculum is expected to cover AI-specific cost metrics, token-based pricing models, compute resource optimization, and frameworks for assessing AI ROI.

These certifications will complement the Foundation's existing credentialing program while addressing the specialized knowledge required for effective AI financial governance. The certification development represents a critical step in establishing standardized approaches to AI cost management across the industry. An official announcement regarding availability and enrollment opportunities is expected in the coming months.

Sources:

FinOps Tooling Update

The FinOps tooling landscape continues to evolve rapidly as organizations seek better ways to manage cloud costs, particularly with increasing AI workloads. Below is a summary of key developments from leading vendors over the past month.

CloudZero announced its fourth consecutive year of triple-digit revenue growth in 2024, with a 124% year-over-year increase in cloud spend under management. Many customers now oversee over $50 million annually across multiple cloud and SaaS providers. The company has expanded globally with new teams in Europe, India, and Latin America, and established a new headquarters in Boston. CloudZero's consistent growth validates its approach to cloud cost management, as evidenced by its enterprise client roster including DraftKings, Expedia, and Grammarly.

Harness completed its merger with Traceable on March 4, 2025, creating what they call the world's most advanced AI-native DevSecOps platform. This strategic union integrates Harness' software delivery expertise with Traceable's API security capabilities to provide customers with a seamless solution for developing, delivering, and securing applications. The merger reflects a growing trend toward embedding security and cost considerations earlier in the software development lifecycle.

Apptio (IBM) has highlighted a potential crisis where organizations struggle to balance technological ambitions in AI with financial realities. Research shows a disconnect between expected technology budget increases and actual AI implementation costs. Apptio advocates for robust financial management systems to track AI costs in real-time, optimize spending, and make informed decisions about AI investments.

CAST AI published its 2025 Kubernetes Cost Benchmark Report, analyzing over 2,100 organizations across major cloud providers. The report reveals persistent inefficiencies in cloud resource utilization, with low average CPU utilization (10%) and memory utilization (23%). Strategic use of Spot Instances can achieve an average compute cost reduction of 59%, with clusters running exclusively on Spot Instances seeing a 77% reduction.

Datadog released its State of Cloud Costs 2024 Report, finding a 40% increase in average spending on GPU instances among organizations, driven by AI and large language model adoption. The report identified containers as a common area of wasted cloud spend, with 83% of container costs associated with idle resources.

Finout announced that its Virtual Tagging technology and MegaBill feature have been officially patented. MegaBill provides a unified view of costs across all cloud providers, while Virtual Tagging allows for instant cost allocation without infrastructure modifications.

ProsperOps continues to highlight the importance of continuous optimization for AI costs. Their platform automates the management of cloud discounts across major providers by leveraging AI to continuously optimize commitments, adapting to usage changes in real-time.

Flexera completed its acquisition of NetApp's Spot FinOps portfolio on March 3, 2025, bolstering its FinOps offerings. The expanded portfolio delivers AI-powered technologies and Kubernetes cost optimization, positioning Flexera as a leader in the evolving FinOps market.

Sources:

Cloud Provider Updates

Microsoft Azure: FOCUS Support and AI Insights

Microsoft Azure has introduced several significant updates to enhance FinOps capabilities. In February 2025, Enterprise Agreement (EA) customers received improved cost allocation functionality with the addition of an AccountId column and population of the InvoiceSectionID column with DepartmentId, simplifying organizational cost attribution.

Azure has integrated Copilot nudges within the Cost Management interface, guiding users in analyzing current expenditures, comparing costs across different periods, and forecasting future spending. A noteworthy development is Azure's support for the FinOps Open Cost and Usage Specification (FOCUS) as an open billing data format, accompanied by a "Learning FOCUS" blog series to educate users on implementation benefits.

March updates emphasized optimization for Azure Kubernetes Service (AKS) costs, providing granular visibility through Kubernetes views in cost analysis. These views enable expense examination at the namespace level and offer aggregated insights across AKS clusters. Azure also highlighted built-in tools for minimizing idle costs, including auto-scaling and node auto-provisioning.

Azure announced the retirement of the AWS connector in Microsoft Cost Management, scheduled for March 31st, 2025, recommending FOCUS for multi-cloud cost analysis. Additionally, Azure introduced the ability to exchange Azure OpenAI Service Provisioned Reservations, offering greater flexibility in managing AI capacity investments.

Sources:

AWS: AI Infrastructure Investment and Cost Management Tools

Amazon Web Services (AWS) has announced plans to allocate $100 billion in capital expenditure for 2025, primarily focused on strengthening its AI infrastructure capabilities. This substantial investment underscores the growing importance of AI workloads and their associated costs.

AWS continues to enhance its comprehensive suite of FinOps services, including AWS Cost Explorer for visualizing cloud spending, AWS Cost Anomaly Detection which leverages machine learning to identify unusual spending patterns, and AWS Trusted Advisor offering recommendations for cost optimization and performance.

The Amazon Q assistant has been integrated with AWS Cost Explorer, enhancing the user experience for cost analysis through natural language interactions. Amazon itself has developed a custom solution utilizing AWS tools to manage its internal cloud costs, validating the effectiveness of these services for enterprise-scale cloud financial management.

Sources:

Google Cloud: Reorganization with AI Focus

Google Cloud Platform (GCP) is implementing a strategic reorganization plan to increase its focus on artificial intelligence initiatives. This shift includes planned workforce adjustments within the Google Cloud division to better align resources with AI priorities.

The FinOps Foundation continues its engagement with the Google Cloud community, with a scheduled presence at Google Cloud Next 2025 where further developments in cloud financial management may be announced.

This strategic reorganization emphasizes Google Cloud's commitment to AI technologies and may lead to future developments in AI-related cost management tools and services as part of their FinOps portfolio.

Sources:

FinOps Investment Trends

Global venture capital funding decreased 26% in February 2025, though the FinOps sector maintained significant activity as organizations seek to control technology expenditures.

Finout: Secured $40 million in Series C funding (January 2025), bringing total investment to $85 million, demonstrating continued investor confidence in FinOps platforms.

Mission (CDW): Partnered with Vega Cloud to deliver enhanced cloud financial management solutions with real-time billing anomaly detection.

Crayon: Achieved the prestigious Certified FinOps Service Provider status from the FinOps Foundation, joining a select group of globally certified providers.

Vantage/CloudSisters: Formed strategic alliance to deliver integrated FinOps, governance, and cost optimization solutions to mutual customers.

Ternary: Introduced specialized features for Managed Service Providers including custom branding capabilities and enhanced billing statements.

Industry forecasts project $44.5 billion in wasted cloud infrastructure spending in 2025, with 63% of companies planning to implement AI expense management, up from 31% in 2024.

Cost Optimization for AI

The optimization of Generative AI costs is increasingly centered on model efficiency and intelligent resource allocation strategies. Industry trends show growing adoption of smaller, more efficient AI models that deliver comparable performance at lower costs, addressing the substantial computational demands of advanced AI systems.

OpenAI has focused on streamlining API usage through enhanced patterns analysis, request batching optimization, and careful token management. These techniques enable organizations to maintain performance while significantly reducing operational expenses.

Two approaches gaining particular traction include:

  1. Intelligent model selection: Organizations are becoming more strategic about choosing the most appropriate model complexity for specific tasks rather than defaulting to the most powerful option available.

  2. Parameter-efficient fine-tuning: This technique enables model customization with substantially fewer computational resources, allowing organizations to create specialized models without the full expense of comprehensive training.

Anthropic is addressing cost concerns by introducing token-saving updates for models like Claude 3.7 Sonnet, which reduce expenses associated with foundational AI technologies. For organizations developing sophisticated AI agents and training frontier models, cost management strategies increasingly include leveraging open-source tools and utilizing pre-trained models to minimize extensive training requirements.

These efficiency-focused approaches collectively make powerful AI technologies more economically viable and accessible to a broader range of organizations, democratizing access to advanced AI capabilities.

Best Practices for Managing LLM and Frontier Model Costs

Optimizing costs associated with Large Language Models (LLMs) and frontier models requires a comprehensive approach encompassing several key strategies:

First, achieving granular visibility into AI spending allows for accurate cost allocation to specific teams, projects, and use cases. Tools like Portkey offer capabilities to track expenditures down to individual prompts and model calls, providing the detailed insights necessary for effective management.

Resource management is critical for cost control. Organizations should diversify instance types based on specific AI workload requirements, selecting between CPU-intensive or GPU-accelerated instances based on actual processing need. Implementing resource rightsizing and leveraging auto-scaling capabilities ensures capacity aligns with demand, preventing both expensive over-provisioning and performance-limiting underutilization. For non-production environments, scheduling resources to operate only during active development periods can substantially reduce expenditures.

For token-based pricing models, optimizing consumption through techniques such as advanced prompt engineering (crafting efficient and concise prompts) and implementing caching systems for frequently used API responses can minimize token usage and associated costs. Continuous monitoring of AI usage patterns helps identify inefficiencies and unexpected cost increases, while implementing usage limits, throttling mechanisms, and anomaly detection tools can prevent runaway expenses.

Adopting AI-specific FinOps practices provides substantial benefits. Establishing cross-functional teams that include data scientists and ML engineers ensures cost considerations are integrated into technical decision-making. Embedding FinOps principles early in the AI development lifecycle (often referred to as "Shift Left") prevents costly architectural choices and promotes cost-conscious design approaches.

Organizations should maintain focus on business value derived from AI investments, ensuring optimization efforts align with strategic objectives. As the intersection of FinOps and sustainability grows, implementing carbon-aware cloud spending strategies and selecting energy-efficient data centers for AI workloads can contribute to both cost reductions and environmental responsibility.

These comprehensive strategies enable organizations to significantly reduce AI-related expenses while maximizing business value from their AI initiatives.

Sources:

Conclusion

The latest developments underscore the increasingly vital role of FinOps in effectively managing the rapidly expanding costs associated with the widespread adoption of artificial intelligence. A clear trend within the industry indicates a concerted effort to broaden the scope of FinOps practices to fully address the complexities inherent in AI, encompassing areas such as Generative AI, AI Agents, and the underlying infrastructure that powers these technologies.

While the current emphasis appears to be on establishing comprehensive visibility and accurate allocation of AI-related spend, the next phase of this evolution will likely involve the implementation of more sophisticated optimization strategies and the development of specialized tools tailored to the unique characteristics of AI costs. To navigate the future of AI cost management successfully, FinOps practitioners must remain well-informed about the latest offerings from cloud providers, and evolving industry best practices, and continuously adapt to the rapid developments in AI technology.

- Subscribe today for a competitive edge in the rapidly evolving world of FinOps and AI.

About Tony Mackelworth

Tony Mackelworth has a proven track record in service leadership, product management and consulting. He has built and scaled global service portfolios in Microsoft consulting and FinOps, driving innovation, efficiency, and tangible results for global organizations.

With extensive experience delivering consulting services and leading practices, Tony combines strategic vision with hands-on expertise to help organizations maximize value from their Microsoft investments.

This website serves as a resource for the licensing community and a platform to share insights, empowering businesses to navigate FinOps, AI business transformation and cloud commercial models with confidence.

He is the Head of Solutions at Codestone Group.

Learn more about his work and insights at Softspend.

Disclaimer

This article is intended for informational purposes only and does not constitute legal, financial, or licensing advice. Microsoft licensing and feature availability can vary by region, subscription type, and contract terms.

Please be aware that nothing on this website constitutes specific technical advice. Some of the material on this website may have been prepared some time ago and therefore may have been superseded. Specialist advice should be taken in relation to specific circumstances.

The contents of this website are for general information purposes only. Whilst the author(s) endeavour to ensure that the information on this website is correct, no warranty, express or implied, is given as to its accuracy and the primary author and website owner or it’s contributing Authors do not accept any liability for error or omission.

The contributing authors and owner of the website shall not be liable for any damage (including, without limitation, damage for loss of business or loss of profits) arising in contract, tort or otherwise from the use of, or inability to use, this website or any material contained in it, or from any action or decision taken as a result of using this website or any such material.

This Disclaimer is not intended to and does not create any contractual or other legal rights. This website is not run by Microsoft and the opinions are the author’s own.

All content on this website created by the author is subject to copyright with all rights reserved.

Tony Mackelworth https://www.softspend.com
Next

FinOps Monthly AI Edition #1