April 10, 2024

Avoid Cloud Overruns: The FinOps & AI Blueprint

In a catch-22 scenario, while cloud computing promises cost-effective scalability, inadequate management can result in financial waste, prompting a shift back to on-premises solutions for some companies

6 min read

Meet our Editor-in-chief

Paul Estes

For 20 years, Paul struggled to balance his home life with fast-moving leadership roles at Dell, Amazon, and Microsoft, where he led a team of progressive HR, procurement, and legal trailblazers to launch Microsoft’s Gig Economy freelance program

Gig Economy
  • Embracing FinOps and AI can reduce cloud costs by up to 20%.

  • Cloud cost forecasting is a significant challenge, with 60% of IT leaders experiencing budget overruns, leading some to revert to on-premises solutions.

  • AI-powered forecasting tools offer speed, accuracy, and agility, which are crucial for optimizing cloud costs and achieving business outcomes.

Staff writer

From AI to FinOps, our team's collective brainpower fuels this blog.

According to Gartner, 60% of leaders in the IT infrastructure and operations division see cost overruns, affecting their on-premises budget.

Because of the cost overruns, many companies investing in cloud infrastructure are increasingly returning to on-premises. This dents the 'glamorized' picture of cloud computing being a cost savior for IT operations.

If that paints a gloomy picture, thankfully, there’s a solution. It is FinOps and AI. Our company has leveraged the FinOps principle (popularized by the FinOps Foundation) to lower cloud costs. In this article, learn how you can replicate the same.

Managing Cloud Costs: The FinOps Approach


Cloud computing has firmly established itself as the foundation of modern digital infrastructure. In a survey conducted by O'Reilly, nearly half of the respondents said they plan to migrate 50% or more of their operations to the cloud. 

While there are several benefits to cloud computing, the one that stands out is the inherent flexibility and price elasticity. Users get to pay as much as they use when they use it. This is better than the traditional in-house server deployment, which requires a high upfront cost.

So why are companies turning back to on-premises?

That’s because flexibility can be a double-edged sword. Without proper monitoring and cost controls, applications or teams might inadvertently overprovision resources, leading to wasted spending on unused capacity.

Conversely, if resources are provisioned at a higher level than necessary for average workloads, they may need to be more utilized during periods of low demand. While elasticity allows for scaling down, if it’s not done promptly or efficiently, you’re still paying for resources that aren’t fully utilized.

Hidden Costs

There are some unaccounted hidden costs associated with cloud computing.

Cloud providers often charge fees for data leaving their platforms. This is accounted as egress fees and is referred to by some as “stupid pill.” Data leaving the platform also includes downloading data to on-premises environments or accessing data stored in the cloud from external networks. In a survey conducted by IDC, 99% of respondents admitted to incurring unfair egress fees.

Likewise, storage, network bandwidth, licensing, and manpower costs can all be hidden costs.

This unpredictability associated with cloud computing costs makes budgeting challenging and leads to problems like missed budgets, strained cash flow, and reactive decision-making.

The FinOps Mindset

Solving the ‘cloud cost overrun’ problem requires a shift in mindset, and that mindset change requires embracing the FinOps methodology.

FinOps, short for financial operations, is a methodology and cultural shift that helps organizations get the most value from their cloud investments. It requires a mindset shift in which all stakeholders, not just the cloud computing team, are responsible for managing cloud costs effectively.

It promotes collaboration between finance, IT, and development teams. This fosters a shared understanding of cloud costs and encourages responsible resource usage.

So, it’s not just about cutting costs but optimizing spending to achieve the best possible business outcomes.

AI to the Rescue: How it Works

Once you’ve laid out the FinOps principle, there’s the execution step. And when executing your plans, you need AI at the core, especially regarding forecasting.

AI-powered forecasting involves using machine learning algorithms to make forecasts. The specialized algorithms are fed rich data, and the models find valuable patterns and insights to help them make forecasts.

Traditional cloud cost management often relies on historical data and manual analysis, which can be time-consuming and lead to inaccurate predictions. Alvaro Torroba, technical account manager at Google, agrees. In one of his articles, he mentions that “accurate cloud forecasting is one of the most difficult things to get right. In The State of Finops survey, advanced practitioners reported variances of about +/- 5% from their predictions - whilst less advanced reported variances as +/- 20%.

In contrast, AI-powered cloud cost forecasting offers a more sophisticated approach, providing businesses with valuable insights for better financial planning and cost optimization.

AI algorithms can analyze vast amounts of historical data, including usage patterns, resource trends, and external factors like seasonality – something that requires a human data analyst. This comprehensive analysis leads to more accurate cost forecasts compared to manual estimations.

As a real-life example, AI models can look at past data and predict when a spike in traffic is expected. This way, teams can better prepare for high-traffic scenarios and avoid outages and downtime.

In a recent survey conducted by Forbes Advisor, 59% of businesses expect AI to help them save costs. And with cloud cost, it’s a real possibility.

Enhanced Visibility and Control

Besides prediction, AI empowers teams with enhanced cloud or hybrid infrastructure visibility.

AI algorithms can analyze data generated by cloud resources, including CPU utilization, network traffic, and application performance metrics, in real-time. This comprehensive view allows for a deeper understanding of cloud health and performance.

These benefits aren't limited to tech-first companies like Microsoft or Google. Even offline-heavy businesses like McDonald's are saving as much as $20 million with enhanced visibility enabled by FinOps. Likewise, Capital One, a financial giant, is saving in the range of $100 million.

According to a McKinsey report, organizations that deploy FinOps can effectively reduce their cloud costs by 20-30%

The AI models can aid proactive planning and optimize resources when combined with current market trends and historical data.

Enhanced visibility also helps teams identify and eliminate zombie infrastructure. Keeping zombie infrastructure running or provisioned incurs ongoing costs, including compute charges, storage fees, and data transfer costs. Over time, these unnecessary expenses can accumulate and significantly impact the organization’s cloud budget, eroding cost savings and ROI. Moreover, they pose a significant security risk.

Adopting AI-driven cloud management puts control back in your hands.

The Role of AI in Cloud Cost Management

Accurate cloud cost forecasting is still one of the most challenging tasks to get right. But specific AI tools can make your life easier. The question is which ones.

The answer is simple: if you have a small team with limited configurations on a single cloud, you can use your cloud provider's stock management and forecasting tool.

However, once your cloud usage gets more extensive and you utilize more than one cloud provider, you should consider investing in a standalone management tool. These tools offer multi-cloud support, extensive analytics capabilities, and deep personalization.

Here’s a checklist to help you find the right AI cloud forecasting tool:

  • Cloud compatibility: Ensure the tool integrates with your specific cloud provider(s) (AWS, Azure, GCP, etc.).
  • Data sources: Verify the tool can ingest data from your existing cloud cost management tools and billing platforms.
  • Forecasting capabilities: Evaluate the level of detail provided in forecasts (granular cost breakdowns by resource, department, etc.).
  • AI and Machine Learning (ML) Features: As must-have features, understand the AI/ML techniques employed by the tool and their effectiveness in predicting costs.
  • Alerting and reporting: Assess the tool’s ability to generate alerts for potential cost anomalies and provide customizable reports.
  • Scalability and automation: Consider the tool’s ability to scale with your cloud usage and automate cost optimization tasks.
  • Security and compliance: Ensure the tool meets relevant security and data privacy regulations.
  • Ease of use: Evaluate the user interface and ensure it aligns with the technical expertise of your team.
  • Pricing and ROI: Compare pricing models and assess the expected return on investment the tool can deliver.

Getting Started with AI Forecasting

Once you have zeroed in on the forecasting tool, it’s time to start forecasting.


Before you forecast cost, it's better to understand the cost complexity for each AI function. Some functions cost more to perform than others. And you're billed accordingly. For example, building and deploying an AI model from scratch will incur the highest cost, while requesting structured or raw unstructured data costs the lowest.

The AI forecasting model needs to get this right. Thus, you may have to manually set data for the model to calculate the cost.

Once that's set, you must feed the right amount and data quality to the AI forecasting model. This includes both historical trends and real-time data.

But to get to that level, you must feed the correct data and ensure its quality. This includes both historical trends and real-time data.

Among other data types, you’d need cloud usage data like CPU utilization, memory usage, disk I/O, network traffic, API requests, cost data, resource metadata, and key business metrics.

ML models then process those data to create forecasts. The good part is that the self-learning models self-optimize the forecast by corroborating it with real-life results. Thus, as time passes, the estimates get better and better.

Once the algorithm is in place, you must set up alerts and notifications. The alerts inform key stakeholders of important events, like the likelihood of a spike in cloud cost or overconsumption of cloud resources. The team can then take proactive measures to address those issues.

Embracing AI for Cloud Cost Optimization

The future of cloud cost management is pretty straightforward: it will be driven by AI.

From now on, the goal should be to get started with FinOps and have AI at its core. A study by Foundry concluded that businesses activating the FinOps model with AI-powered solutions see a 20% reduction in overall cloud cost.

Many exciting things are happening in space, with concepts like Generative AI and Explainable AI coming to life. The former helps generate new data from existing data, while the latter helps understand the reasoning behind an action. Both of these hold the potential to create higher savings soon.

Business leaders who embrace AI, starting with cloud cost forecasting, can position themselves better than their peers for future success.

Cut through the AI hype and join the thousands of business leaders getting practical enterprise insights delivered to their inbox

Welcome to the community! We'll be in touch soon.