Cloud computing transformed how the world runs software. Instead of owning servers, companies rent compute, storage, and software from cloud providers — paying only for what they use. This shift created one of the most valuable and fastest-growing sectors in technology, generating over $700 billion in annual revenue in 2025 across infrastructure, platform, and software layers.
The Three Layers of Cloud Computing
The cloud industry is organised into three distinct service layers, often called the “cloud stack”:
Infrastructure as a Service (IaaS)
IaaS providers rent raw compute, storage, and networking. Customers provision virtual machines, databases, and storage buckets on demand. This is the most capital-intensive layer — cloud providers must build and operate massive data centres — but it generates the largest revenue pools.
Amazon Web Services (AWS), Microsoft Azure, and Google Cloud Platform (GCP) collectively control roughly 65–70% of global cloud infrastructure revenue.
Platform as a Service (PaaS)
PaaS sits above IaaS and provides managed runtimes, development tools, databases, and middleware. Customers deploy applications without managing the underlying infrastructure. This layer generates higher gross margins than IaaS because it packages proprietary services on top of commodity compute.
Software as a Service (SaaS)
SaaS delivers complete software applications over the internet — no installation, no maintenance. Users pay a subscription, the vendor handles everything else. SaaS is the highest-margin layer of the cloud stack, with gross margins of 70–85%+ for leading vendors.
Revenue Models in Cloud Computing
| Model | How Revenue is Generated | Typical Gross Margin |
|---|---|---|
| Consumption (IaaS/PaaS) | Billed per GB, CPU-hour, or API call | 60–70% |
| Subscription (SaaS) | Fixed monthly/annual seat fees | 70–85% |
| Hybrid (platform + seats) | Consumption floor + per-seat licenses | 65–80% |
| Marketplace | Commission on third-party software sales | 80%+ |
Consumption vs. Subscription
The most important distinction in cloud revenue is between consumption-based and subscription-based models:
Consumption models (Snowflake, Datadog, MongoDB) align customer costs directly with usage. Revenue can grow faster than seat counts — but can also compress quickly if customers optimise their spend. These models carry higher revenue quality signals (customers only pay when they extract value) but lower predictability.
Subscription models (most SaaS) provide predictable annual recurring revenue (ARR). Net revenue retention (NRR) above 120% is the gold standard — it means the installed base is expanding faster than churn, producing compounding revenue growth even with zero new customer acquisition.
Key Companies in Cloud Computing
Hyperscalers (infrastructure + platform + AI):
- Amazon Web Services (part of Amazon) — the market leader
- Microsoft Azure (part of Microsoft) — second place; growing fastest
- Google Cloud Platform (part of Alphabet) — third; AI-led strategy
Pure-play cloud infrastructure and SaaS:
- Snowflake — cloud data warehousing, consumption model
- Datadog — cloud observability, monitoring, and security
- MongoDB — cloud-native database platform
- Cloudflare — cloud network security and edge computing
- Atlassian — cloud collaboration and DevOps software
- Twilio — cloud communications APIs
- Arista Networks — cloud networking hardware and software
- Equinix — data centre colocation and interconnection
- IBM — hybrid cloud and enterprise AI
Key Metrics for Cloud Companies
Annual Recurring Revenue (ARR)
ARR measures the annualised value of subscription contracts. Growth rate and ARR expansion are the primary valuation drivers for SaaS companies. An ARR growth rate above 30% is considered high-growth.
Net Revenue Retention (NRR)
NRR measures how much existing customers spent this year vs. last year, including expansions, upgrades, and churn. NRR above 120% means the company grows revenue from existing customers alone — new sales are pure upside. Most elite cloud companies target NRR of 115–130%+.
Gross Margin
Infrastructure-layer companies (AWS, Azure, GCP) run at 60–70% gross margin at scale. Pure SaaS businesses target 75–85%. Gross margin expansion over time signals improving economies of scale and pricing power.
Operating Leverage
The cloud business model generates powerful operating leverage — once the software is written and infrastructure is provisioned, each additional customer dollar costs almost nothing to serve. The best cloud companies see operating margins expand 5–10 percentage points per year as they scale.
Deferred Revenue
SaaS companies collect annual or multi-year contract payments upfront. The unrecognised portion sits as deferred revenue on the balance sheet — a forward indicator of future recognised revenue. Rising deferred revenue signals strong bookings; flat or falling deferred revenue is a warning sign.
The AI Inflection Point
AI is reshaping every layer of the cloud stack. The implications differ by layer:
For hyperscalers: AWS, Azure, and GCP are absorbing unprecedented capital investment to build AI training and inference infrastructure. Microsoft has committed over $80 billion in data centre capex for 2025 alone. This spending is both a growth driver (higher revenue per server) and a margin headwind (more depreciation, more energy costs).
For SaaS vendors: Every SaaS company is racing to embed AI into its product — AI-assisted code review (GitHub Copilot), AI-generated analytics (Datadog AI), AI workflow automation (Atlassian Intelligence). Companies that successfully monetise AI features can expand ARPU (average revenue per user) without acquiring new seats.
For infrastructure companies: Cloudflare, Arista, and Equinix are critical enablers of AI inference at the edge. As AI workloads move from centralised training to distributed inference, these companies become essential chokepoints.
Competitive Dynamics
The Hyperscaler Moat
AWS, Azure, and GCP have entrenched advantages: global data centre footprints, mature managed services portfolios, and deep enterprise relationships. Switching costs are high — workloads are often tied to proprietary services that do not easily migrate. This lock-in is the core moat.
Multi-Cloud and Cloud Optimisation
Enterprise customers increasingly run workloads across multiple clouds to avoid lock-in and optimise costs. This trend benefits neutral infrastructure players (Equinix, Cloudflare) and multi-cloud data platforms (Snowflake, MongoDB) while limiting any single hyperscaler’s ability to capture 100% of a customer’s spend.
The Price Compression Risk
IaaS is inherently commoditising. Compute and storage prices fall roughly 20–30% per year as hardware improves and competition intensifies. Cloud providers must move customers up the stack (to higher-margin managed services and AI) to maintain revenue growth and margins.
Key Comparisons
Related Glossary Terms
- Annual Recurring Revenue (ARR) — the primary SaaS growth metric
- Deferred Revenue — bookings health indicator for subscription businesses
- Gross Margin — quality of cloud revenue
- Operating Leverage — why cloud margins expand with scale