Why is it important to optimize cloud costs?

Cloud costs directly impact net revenue and, therefore, company valuation – as among the main factors that determine COGS for software companies, cloud costs are one of the only expenses that directly affect net income for tech companies. Since gross margin influences the exit multiple, any decrease in COGS increases the potential value of the company.

High cloud costs can have critical cash flow implications – if cloud costs aren’t paid on time, your systems go offline and paying customers can’t use your product. Because cloud payments can never be delayed, higher cloud costs give companies less flexibility in how they use their cash and can pose an operational risk for companies that are cash-poor.

Well-managed cloud costs indicate a well-operated business – effectively managing cloud costs requires strong collaboration between Engineering and Finance. The sooner a company’s leadership can implement effective solutions to complex problems, the better equipped they’ll be to operate more efficiently in the future. This is especially relevant for companies that are in the early part of a PE hold period.

What do typical savings look like for a company that takes a systematic approach to optimizing cloud costs?

Companies can often expect to reduce cloud costs by over 50% – companies that haven’t yet taken a comprehensive approach to cloud cost optimization can save over 80% on cloud costs.

Savings opportunities exist for all levels of technical complexity – because cloud costs are often determined at the level of the individual engineer, organizations with sophisticated technological architecture can still benefit from this process.

Wide-scale cloud cost optimization should be a one-time cost saving exercise – while cloud costs must be monitored on an ongoing basis, a successful optimization (including its implementation) shouldn’t become obsolete as a company continues to grow.

Why do companies overspend on cloud costs?

Product and engineering leaders often have incentives to spend more than they need – it’s in the Product team’s best interest to make their cloud structure bulletproof, which encourages CTOs to err on the side of overestimating the need and overpaying for cloud services.

The skillset needed to optimize cloud costs (finance and engineering) often isn’t involved – strong communication and trust between the Finance and Engineering teams result in better cloud costs and higher profitability. However, many small companies lack the bandwidth or awareness to realize that the wrong teams are weighing in on cloud architecture decisions. Partnering with experts who have engineering and financial expertise can help both parties better understand how to think about cloud costs going forward.

When should companies evaluate their cloud costs? Is there a particular milestone or time when reevaluating your cloud costs can have the greatest impact?

Common milestones for evaluating cloud costs include:

• After you’ve established a viable growth strategy – early-stage companies should focus on perfecting their product-market fit, building new features, and not getting in the way of their own growth. Once they’ve reached a more stable stage—often under PE ownership—they have the bandwidth to reframe their operations and build a foundation for long-term profitability.
• If your cloud costs reach 20% of revenue – on average, cloud costs equate to 12% of ARR; highly optimized cloud strategies can reduce cloud cost as a percentage of revenue to 5% or less of ARR.
• The beginning of the PE hold period – PE-owned companies often have greater cash needs than they did pre-acquisition. Unoptimized cloud costs are a consistent opportunity to unlock additional cash.
• After conducting an acquisition – acquisitions are the exception to the rule. You should evaluate the cloud costs of any acquisitions that will continue to use their own cloud infrastructure.

Cloud costs should always be monitored – though intensive cloud cost evaluations aren’t a cyclical exercise. If cloud costs are a recurring issue, you might have a foundational operational or pricing problem.

Who should own cloud spend in your organization?

Engineering teams should own day-to-day cloud cost management – there’s no practical way to restrict how line-level engineers utilize or provision cloud resources. Engineering, Product, or Technology leaders should own the overall strategy and educate their teams on how to use resources accordingly.

CFOs should provide oversight and education around financial concerns – engineering teams lack both the financial context and accountability that could help them make more strategic decisions about cloud spend. Finance should also proactively flag issues to the engineering team if cloud costs are approaching the 20% mark.

Distribution of responsibilities between Finance and Engineering
	Finance	Engineering
Priority	Reduce costs to increase profitability	Prepare product to withstand worst case scenario
Most common problem that contributes to conflict around high cloud costs	Requests dramatic operational overhauls without understanding the practical implications	Accidentally creates cost center due to a lack of understanding about how cloud operations relate to company performance
Ideal role	• Helps Engineering understand the financial implications of different cloud strategies • Provides political support for operational overhauls that can be long or complicated	• Creates cloud strategy that supports technical needs and is informed by financial concerns • Individual engineers understand how their activities drive costs for the company

Measuring Costs

How does cloud billing work? How do you determine whether or not your spend is efficient? 

Cloud billing works as a metered system – cloud bills are similar to utility bills, where usage is collected after the fact, totaled, then billed to you. The different types of cloud meters (server disk, data transfer in and out, server usage, operating system license, and RAM) are directly influenced by the activities of each line-order engineer.

Cloud cost management depends on how costs relate to revenue, not metrics like utilization rate – each cloud infrastructure structure has its own cost structure. If the cost of offering your product is too high (e.g., 20% of revenue) and you have already optimized pricing from a customer standpoint, you need to explore a more cost-efficient cloud infrastructure. Average utilization rate might not factor into this decision because it can be heavily influenced by customer behavior, which often varies throughout the day.

Optimized cloud operations are scaled to your company’s workload – workload should determine the quantity and size of resources you use. Significant discrepancies between what you actually use and what you pay for are opportunities to improve efficiency.

How do you audit your current cloud costs? 

There are three pieces which you have to triangulate:

• Utilization metrics – review your actual usage patterns (e.g., number of gigabytes, IOs, utilization rate, and other usage metrics). Note that “utilization metrics” is not another term for “utilization rate”, one of many metrics.
• Billing data – the resulting bills from your utilization.
• Configuration options – quantify the savings that another configuration could provide, you can determine what level of savings you can achieve from an operational standpoint.

Where do companies typically find cloud cost savings? 

Look for basic configuration changes that can yield significant savings – usage pattern analysis can inform configuration optimizations that save $10,000s per year. For example, AWS Aurora has 2 pricing models: 1 with an hourly rate with additional charges for exceeding certain IO limits, and 1 with unlimited IO operations charged at a highly hourly rate. Understanding realistic usage patterns can let you commit to the lower priced model while reducing the risk of exceeding IO limits.

Only consider committing to contracts after you’ve identified your ideal configuration – companies that commit to contracts first risk locking in contracts that make sense for much higher usage levels, which results in wasted spending. One of the largest sources of savings is right-sizing commitment levels.

Don’t rely on vendor-provided optimization reports – cloud vendor reports often emphasize information that doesn’t help users understand ROI. The large quantities of information included at the beginning of these reports can distract from more meaningful analysis that can create more value.

Which tools can help you measure cloud costs?

Native cloud provider tools generally provide all the information you need – AWS, Azure, and Google Cloud allow you to tag, categorize, and set budgets within their platforms. These functionalities create accountability and give engineering teams the ability to dive deeper when costs increase.

If you buy AWS services through a reseller, you might need to use a separate system like CloudHealth – AWS is not built for resellers, so companies that don’t purchase their services directly can’t monitor cloud costs without using a third-party tool. This limitation will likely be addressed in the future.

Build a report that connects cloud costs to overall financial reporting – track cloud costs against a goal or budget that evolves with your company’s growth. Cloud costs should increase in parallel with revenue growth.

What are the main cloud hosting cost structures?

All major cloud providers use metered billing – Azure’s “pay-as-you-go”, AWS’s “on-demand”, and Google Cloud’s “subscription” models all charge based on usage.

Pre-purchased resources provide discounts on meter-based bills – you can pre-purchase dollars per hour or actual resources in exchange for a discount. These arrangements might be referred to as contracts, reserved instances, capacity reservations, or savings plans.

Enterprise agreements offer volume discounts with long-term spending commitments – AWS and Azure enterprise agreements typically span 3 years and 5 years, respectively.

Systems like Prosper Ops offer convertible contracts – there is a cottage industry of automatic contract managers, which claim to offer zero risk by providing access to contracts that can change as your needs increase.

Cloud Cost Structure Comparison
Agreement type	Typical metered billing	Pre-purchased discounts	Enterprise contracts
Who it makes sense for	Early-stage companies with unoptimized cloud infrastructure	Companies with optimized cloud infrastructure	Larger companies that can project long-term infrastructure needs
Benefit	Flexibility	Discount	Discount
Risk	Minimal, as savings typically manifest at a larger scale and/or from the optimization itself	Paying for resources you don’t need	Paying for resources you don’t need for several years

Optimization Strategy

Where should companies look first when optimizing their cloud costs?

Compare maximum server workload to capacity – examine the CPU percentage and available memory on each server at peak usage to determine if they’re scaled too large for their maximum workload, which is a common source of waste.

Configure your system to scale to meet different levels of need – there are various solutions for scaling depending on your software architecture’s sophistication. Most teams can reduce cost by implementing scaling, even to a limited degree. Common types of scaling include:

• Vertical scaling – making your database larger during “on” hours and smaller during “off” hours
• Horizontal scaling – increasing the number of instances during “on” hours and decreasing them during “off” hours
• Dynamic Scaling – scaling in response to customer workloads

Note: advanced scaling environments use tools like Karpenter to vertically expand nodes in Kubernetes clusters – these highly optimized environments fine-tune scaling algorithms to create the most efficient systems possible.

Adjust database settings to only run when in use – configure your system to turn off environments when they aren’t being used. Companies often forget to use this setting.

Look for waste in data transfer or storage patterns, particularly with audio or video content – systems often accidentally send media out to the internet and back into their own environment, creating 2 sets of unnecessary data transfer costs. Tools like Cloud Custodian can also clean up redundant logs, archives, and copies to reduce overall storage costs.

Always check the most common savings opportunities for your cloud services provider:

• AWS – ARM processors can yield significant savings with minimal effort, especially because tools like Claude make library updates easier to implement.
• Azure – zonal, local, and global redundancies easily create waste when engineers are tempted to opt in to protection programs that are unnecessary due to Azure’s default disaster readiness program.

Overall

How do you know if your organization is doing a good job managing cloud costs? How do you empower your team to make good cloud cost management decisions?

Incentives are aligned with financial outcomes – Engineering’s motivation to support the financial health of the company shouldn’t be overshadowed by their desire to avoid “annoying” restructuring projects or challenging conversations.

You meet or exceed industry benchmarks for cloud cost as a percentage of revenue – the primary indicator of good cloud cost management is maintaining cloud costs below a certain target. Best-in-class companies can reach cloud costs that are as low as 1-2% of revenue.

Every team understands the costs they’re responsible for – transparency and accountability empower teams to make decisions that benefit the business. Engineering teams should have access to the information they need in order to manage the costs they’re responsible for.

What processes need to be in place to monitor cloud costs on an ongoing basis?

Set up continuous monitoring and automated alerts – Engineering should be automatically notified if usage or costs reach a certain threshold, especially if you haven’t yet optimized your cloud infrastructure.

Engineering and Finance should have clear goals for cloud cost as a percentage of revenue – if your team knows what to aim for, they can track progress, hold team members accountable, and make optimizations to achieve that goal.

The 3 Frequencies of Cloud Cost Monitoring
Cadence	Tool/ Report	Purpose
Hourly	Native tools	The most accurate reporting provides information by the hour
Daily	Secondary tools	Nightly reports provide a comprehensive view of the previous day’s usage and billing
Monthly	Financial reports	Cloud cost as a percentage of revenue can only be calculated after the month end close

How should growing companies think about strategies to reduce cloud costs as they scale?

Align cloud cost strategy with your operating model/ hold period strategy – every PE-owned company’s operating model is informed by assumptions about cost reduction and revenue growth. Cost control becomes an important determinant in future valuations, and reducing cloud costs could be preferable to reducing costs in other ways, such as large layoffs.

Educate technical leadership about high-level company priorities and financial basics – CTOs often aren’t included in strategic discussions about operating model and how financial metrics affect future valuations. Empower CTOs to contribute to the company’s success by explaining how their activities directly impact the business’ costs and profitability.