All blogs
11 mins
Cloud Unit Economics: Cloud Costs & SaaS Gross Margin
Piyush-Kalra
A growing AWS bill isn't necessarily a problem. The real question is whether cloud costs are growing faster than the value they're supporting.
That's where cloud unit economics comes in. Instead of looking only at total cloud spend, it measures the cost of serving each customer, transaction, or feature and connects infrastructure costs directly to SaaS gross margin.
In this article, I will explain how cloud unit economics works, which metrics matter, and how to connect cloud costs to gross margin.
What Is Cloud Unit Economics?
Cloud unit economics consists of tracking costs of cloud services against a particular business unit, such as a customer, transaction, API call, or AI workflow, to better define the cost-to-serve instead of only looking at total spend. The FinOps Foundation has adopted unit economics as a formal capability within its FinOps Framework and defines technology costs relative to business outcomes.
Cloud Unit Economics vs. Total Cloud Spend
Total Cloud Spend indicates how much money was spent. In isolation, it does not account for the changing efficiency of the spending as the business grows. A cloud bill that increases by 20 percent month-over-month could be indicative of healthy business growth, unusually high customer use of the infrastructure, or an inefficiently scaled infrastructure.
Unit economics provides context for this missing piece. For example, does the cost for providing each unit of business value decrease as the business grows?
What Counts as a "Unit"?
The answer depends on how your business earns revenue. Some examples are listed below:
Customer: The default starting point for subscription and seat-based SaaS
Transaction: Natural for payments, marketplaces, and e-commerce platforms
API request: Fits developer platforms and usage-based products
Feature: Useful when specific product areas drive disproportionate infrastructure load
Workspace or tenant: Common in multi-tenant collaboration tools
AI workflow or inference call: Increasingly necessary as inference and token costs climb
Focus on one unit that best aligns with your business model and use data that is readily available. Two reliable data points are better than six assumptions.
How Cloud Costs Affect SaaS Gross Margin
When it comes to SaaS, your service runs on top of cloud infrastructure. This cloud infrastructure is typically categorized as Cost of Goods Sold (COGS). However, the exact treatment will be determined by your specific costs and your company’s accounting policies. Expenses that fall into the COGS category are directly related to gross margin and, therefore, include the following equation:
Gross Margin = (Revenue − COGS) ÷ Revenue × 100
When cloud costs increase at a higher rate than revenue, gross margin shrinks even if revenue is increasing, which means that during a review that only looks at ARR, people can miss that cloud costs are a major driver of gross margin.
When Growing Cloud Spend Is Healthy
If your spending on cloud services increases 15% to align with a 30% increase in revenue, then you have decreased your cost per customer. It is important that your total spending on cloud services increases because that indicates that your business is scaling.
When Cloud Spend Starts Compressing Margins
Cloud costs that increase faster than revenue erode gross margin, no matter how good your metrics look. The Benchmark 2026 SaaS Performance Metrics survey puts the median software gross margin at 80%, but that can change significantly by ARR and revenue mix. Use your cohort instead of a blanket approach when setting your benchmarks.
Why AI Workloads Make This More Important
AI features impose cost volatility that traditional SaaS infrastructure doesn’t. Inference costs vary based on model size, context length, and usage patterns in ways that flat per-seat costs never could. Should you have integrated AI features into your product, then the cost per inference or cost for a specific AI workflow should have its own line item in your unit economics model and should not be lumped under general “cloud costs," which hides the volatility you should strive to understand.
Cloud Unit Economics Metrics Companies Should Track
A handful of metrics cover most of what a SaaS finance or engineering team needs:
Cloud cost per customer: attributed cloud costs divided by the defined customer unit, such as paying accounts, tenants, or active customers
Cloud cost as a percentage of revenue: shows how much of each revenue dollar is consumed by cloud infrastructure
Cloud cost per transaction or API request: useful for usage-based, transaction-driven, and marketplace businesses
Cloud cost per feature: helps identify which product areas require the most infrastructure to operate
AI cost per workflow or inference: tracks a rapidly evolving cost category for AI-enabled products
Gross margin per customer: combines customer-level revenue with attributable COGS to show customer-level profitability
Monitoring these metrics on a consistent basis is critical. As the business grows, your definitions of units will need to evolve along with it. The metric that best captures your economics at a $2M ARR may not be the best metric at a $20M ARR.
How to Calculate Cloud Cost per Customer
This cloud-based metrics exercise is highly beneficial for SaaS teams, as it aids in relating cloud costs to the economics of each customer served.
Step 1: Identify Relevant Cloud Costs
Collect the infrastructure costs to deliver the product, including compute, storage, databases, networking, data transfer, and, if applicable, AI or GPU workloads.
For the purposes of the analysis and your company’s accounting practices, classify costs for customer-facing production versus development, test, and other environments. Not all production costs will be directly customer-related, so be consistent.
Step 2: Define the Customer Unit
For this exercise, determine what a customer is. Some common definitions are "active account," "paying account," "seat," or "tenant."
Be consistent in the definition of the customer and use the same unit of measure over time. Otherwise, changes in the metric may signify a difference in the definition and not a difference in the efficiency of the cloud.
Step 3: Allocate Shared Cloud Costs
Most cloud services are shared across several customers, and thus, some allocation methodology will be necessary.
Whereas reliable usage metrics are available, costs should be allocated based on measurable consumption metrics such as compute usage, requests, storage, etc. Where other costs are truly fixed or shared, a consistent allocation method over time should be used.
Step 4: Calculate Cloud Cost per Customer
Cloud Cost per Customer = Attributed Cloud Costs ÷ Customers Served
As an example, let's say you incur $150,000 of attributable costs on your cloud to serve 6,000 customers. It would follow that:
$150,000 ÷ 6,000 = $25 per customer
Step 5: Compare Cloud Cost per Customer With Revenue
Understanding cloud cost per customer allows you to analyze the infrastructure component of customer-level economics relative to revenue per customer.
For a complete customer gross-margin analysis, consider incorporating relevant customer-level COGS and revenue. This may help identify customers or segments with high costs to serve, as well as pricing, usage, or infrastructure issues.
Example: How Cloud Unit Economics Changes the Gross Margin Conversation
Cloud unit economics illustrates the importance of infrastructure economics within the context of gross margin.
Metric | Before | After |
Revenue | $1M | $1M |
Cloud COGS | $250K | $175K |
Cloud cost/customer | $25 | $17.50 |
Gross margin* | 75% | 82.5% |
Notice what didn't change: revenue. Cloud COGS fell by $75,000, increasing gross margin from 75% to 82.5%.
The goal is not to reach a zero cloud bill. The goal is to have an efficient cloud cost of revenue. In other words, the term refers to the cost of revenue associated with delivering services to customers.
*For simplicity, this example assumes cloud COGS is the only COGS included in the gross margin calculation.
Why Total Cloud Spend Can Be a Misleading Metric
Cloud Spend Can Rise While Unit Economics Improve
If revenue grows by 40% and cloud costs grow by 20%, your cloud bill will increase, but your unit economics improve. This is not a paradox but can sometimes be the case if the business is optimally scaling.
Cloud Spend Can Fall While Economics Get Worse
While aggressively cutting costs can result in a smaller cloud bill, that can mean that the performance, reliability, and customer experience of the product have deteriorated. A smaller bill doesn't necessarily mean a better outcome for the business.
Questions Finance and Engineering Should Ask Instead
Instead of asking, “Why did cloud spend increase?" ask better questions:
Why did cost per customer increase?
Is revenue growing faster than the cost to serve?
Which customer segment, product, or feature is driving the change?
These questions provide finance and engineering with a way to discuss cloud costs that stays focused on business growth.
What Causes Cloud Unit Economics to Get Worse?
A handful of patterns show up again and again:
Infrastructure costs grow faster than revenue: If cloud costs grow at a rate faster than the supported business value, cloud unit economics can deteriorate and put margins under pressure.
Pricing that doesn't reflect usage: Fixed pricing can pressure margins when customers with very different usage patterns consume very different infrastructure. Usage-based or hybrid pricing can better align revenue with the cost to serve.
Overprovisioned or underutilized infrastructure: Scaling resources to meet peak request demand can lead to underutilized resources most of the time, increasing costs and diminishing returns on investment for the business.
High-cost customers: Some customers may use a disproportionate amount of compute, storage, network, and AI resources from your company. Without visibility into customer-level costs, these discrepancies may go unnoticed.
Poor cost allocation: Failure to properly allocate shared infrastructure to products, teams, or customers results in distorted unit costs and makes it difficult to analyze where margins and profits are.
AI and data-intensive workloads: AI services have the potential to become more expensive and have less visible cost structures. Additionally, data-centric workloads may become more expensive to provision and maintain as demands for compute and network resources grow. According to the 2026 State of FinOps report, AI cost visibility, allocation, and valuation remain key pain areas for FinOps teams.
How to Improve Cloud Unit Economics
While it is important to reduce the overall costs of the cloud, we want to focus on improving unit economics when it comes to revenue delivery.
Reduce waste and improve utilization: Eliminate idle resources and right-size overprovisioned infrastructure so you're not paying for capacity you don't need.
Optimize cloud rates and commitments: Use Reserved Instances, Savings Plans, and Committed Use Discounts to cover predictable workloads. Cloud providers estimate up to 72% savings on sustainable compute workloads, but the actual savings will vary by the provider, the workload, the contract, and how well the workload is utilized.
Control usage growth: Set usage management controls on workloads that are most likely to have unmanaged usage growth, especially compute-intensive, data-intensive, or AI-related workloads.
Allocate costs to customers, products, and teams: Some companies cannot identify the real costs of cloud services until they can attribute the costs to the various areas of their business.
Align pricing with infrastructure consumption: Usage-based or hybrid pricing can help protect margins when certain customers or workloads consume significantly more infrastructure.
Monitor unit economics continuously: You are probably not going to get value from one analysis. To improve productivity, you need to measure usage, customer engagement, and pricing along with changes to workload consumption efficiency.
How Pump Helps Improve Cloud Unit Economics
This is the workflow Pump is built around:
cloud spend → usage visibility → cost allocation → optimization → financial impact.
Pump View gives you detailed cost and usage visibility across supported dimensions, so you can understand where spending is going before trying to allocate it to a customer or feature. Pump Save then addresses the rate side of the equation by automatically managing Reserved Instances, Savings Plans, and Committed Use Discounts across AWS, GCP, and Azure, helping lower the cost of the underlying infrastructure.
Pump can be connected with read-only access for visibility and takes under 5 minutes to set up. It is also free to use, with no platform fee or percentage of savings.
The result is a clearer path from cloud spend to the metrics that matter: understand the cost, identify the drivers, optimize the infrastructure, and measure the financial impact.
Conclusion
The shift from total cloud spend to cost per unit is the same shift that separates SaaS companies with clear visibility into gross margins from those guessing at profitability.
Total cloud spend tells you what you paid. Cloud cost per customer tells you what it costs to serve each customer and whether that cost improves as you scale.
Start with one meaningful unit. Build a baseline, connect it to revenue and gross margin, and track it over time. That gives your finance and engineering teams a better answer the next time someone asks whether a rising cloud bill is actually a problem.
Frequently Asked Questions
How do you calculate cloud cost per customer?
Cloud Cost per Customer = Attributed Cloud Costs ÷ Customers Served. Attributed costs should include directly attributable infrastructure costs plus a reasonable allocation of shared costs.
What is a good SaaS gross margin?
There is no single universal target. Gross margins vary by business model, infrastructure intensity, pricing model, and product architecture. Benchmarkit reported a 77% median gross margin for total revenue and an 81% median for subscription revenue in its 2025 benchmark data. Its 2026 benchmark reports 80% median software gross margin, highlighting the importance of evaluating margins in the context of the company's business model and product category.
What is the difference between cloud cost and cloud unit economics?
Cloud cost is the total cost of spending on the cloud. Cloud unit economics is the allocation of spending to a given business unit, which can be a customer, a transaction, an API call, or a single AI instruction flow. This enables the assessment of the efficiency of a unit, not the growing cloud bill.
Similar Blog Posts
Azure Metrics Advisor: AI Monitoring & How to Use It
