What the three clouds are betting on

Thirteen jobs every cloud has to answer, and how much of its own output each one spends on them. Measured from the vendors’ own announcement feeds — what they shipped, not what anyone runs.

90-day window from 2026-08-09 · store generated 2026-11-07 · rolled-up announcements — AWS 625 · Azure 220 · Google Cloud 2,262

The job	AWS	Azure	Google Cloud
Virtual machines	EC2 7.0%44 items	Virtual Machines 2.7%6 items	Compute Engine 1.9%42 items
Object storage	S3 3.8%24 items	Azure Storage 2.7%6 items	Cloud Storage 0.6%13 items
Managed Postgres	RDS 7.5%47 items	Azure Database for PostgreSQL 5.9%13 items	Cloud SQL for PostgreSQL 1.9%43 items
Kubernetes	EKS 1.8%11 items	AKS 5.0%11 items	Google Kubernetes Engine 3.2%73 items
Containers, serverless	ECS 2.4%15 items	Azure Container Apps 1.8%4 items	Cloud Run 1.0%22 items
Serverless functions	Lambda 3.5%22 items	Azure Functions 1.4%3 items	Cloud Run functions 0.4%9 items
Data warehouse	Redshift 2.1%13 items	Microsoft Fabric 0.5%1 item	BigQuery 4.4%99 items
GenAI platform	Bedrock 7.4%46 items	Microsoft Foundry 1.8%4 items	Vertex AI 7.8%176 items
Identity	IAM 2.2%14 items	Microsoft Entra 0.5%1 item	Identity and Access Management 0.4%10 items
CDN and edge	CloudFront 1.1%7 items	Azure Front Door 3.2%7 items	Cloud CDN 0.2%5 items
Secrets	Secrets Manager 0.5%3 items	Azure Key Vault 1.4%3 items	Secret Manager 0.1%3 items
Observability	CloudWatch 2.9%18 items	Azure Monitor 2.3%5 items	Cloud Monitoring 0.4%8 items
Messaging	EventBridge 0.5%3 items	Azure Service Bus 0.5%1 item	Pub/Sub 0.0%1 item

How to read this, and how it could mislead you

The percentage is each service’s share of its own cloud’s output, never a count. Google’s combined feed is day-granular and produces several times as many rows as AWS’s, so raw totals would say Google ships four times as much. It does not; the feed is simply shaped differently.

One announcement repeated is counted once. Without that, Container-Optimized OS alone would top the Google column — a single kernel fix shipped across four releases arrives as four notes.

This is vendor attention, not adoption. A high share means that cloud spent a lot of its announcements there. It says nothing about how many companies run the service. That question needs survey data, which is annual and citable rather than measurable from a feed, so it is left off rather than guessed at.

Faded cells carry one item or none — too thin to read as a finding. The thirteen jobs, and which service answers each, are an editorial judgement: nothing in the data says S3 and Cloud Storage are the same question.

The sparkline runs from 2026-08, the first month the store holds enough to plot. A longer window would draw growth that belongs to this page, not to the services.