Compute Pricing: EC2 vs Azure VMs vs Compute Engine
| Instance Type (equiv) | AWS EC2 | Azure VM | GCP Compute Engine |
|---|---|---|---|
| 2 vCPU / 8 GB RAM (on-demand) | $0.0928/hr (m6i.large) | $0.096/hr (D2s v5) | $0.0850/hr (n2-standard-2) |
| 8 vCPU / 32 GB RAM (on-demand) | $0.3712/hr (m6i.2xlarge) | $0.384/hr (D8s v5) | $0.3400/hr (n2-standard-8) |
| 32 vCPU / 128 GB RAM (on-demand) | $1.4848/hr (m6i.8xlarge) | $1.536/hr (D32s v5) | $1.3600/hr (n2-standard-32) |
| GPU instance (A100 equiv, on-demand) | $3.26/hr (p3.2xlarge) | $3.40/hr (NC24s v3) | $2.93/hr (a2-highgpu-1g) |
| 1-yr Reserved / Committed | ~35% off on-demand (Savings Plans) | ~36% off (1-yr Reserved) | ~37% off (1-yr CUD) |
| 3-yr Reserved / Committed | ~52% off (Convertible) | ~56% off (3-yr Reserved) | ~57% off (3-yr CUD) |
| Spot / Preemptible | 60–90% off (variable price) | 60–80% off (Spot VMs) | 60–91% off (Spot VMs, 30-day guarantee) |
On-demand compute pricing is within 5–10% across providers. The real differences emerge in storage, egress, support, and commitment program flexibility.
The Hidden Costs: Where Your Bill Actually Diverges
| Cost Category | AWS | Azure | GCP |
|---|---|---|---|
| Egress to internet (per GB, US) | $0.09 | $0.087 | $0.08 (waived first 200GB/mo free) |
| Egress between regions | $0.02/GB | $0.02/GB | $0.01/GB |
| Object storage (per GB/mo) | $0.023 (S3 Standard) | $0.018 (Blob Hot) | $0.020 (Standard) |
| Managed Kubernetes | $0.10/hr cluster fee + node cost | Free cluster mgmt (AKS) | Free 1 cluster (GKE Autopilot) |
| Enterprise Support (mid-tier) | $15K+/mo Business Support (10% of spend) | $300/mo Developer; $29K+/mo Professional Direct | $150/mo Standard; $1,500/mo Enhanced |
| Sustained use discount (auto) | None (manual Savings Plans only) | None (manual Reservations) | Up to 30% automatic for consistent use |
| M365 / Active Directory integration | Extra config required | Native Azure AD — free for M365 users | Extra config required |
| BigQuery / Redshift / Synapse | Redshift: $0.25/hr + storage | Synapse: $0.20/hr serverless | BigQuery: $6.25/TB queried (serverless) |
Cloud Provider Profiles: Who Each One Is Best For
AWS
Mature, broadest services
- 200+ services — broadest catalog
- Best global region coverage (32 regions)
- Deepest marketplace & ISV ecosystem
- Best Spot instance availability
- Lambda leads serverless market
- Strong compliance (FedRAMP, HIPAA)
- Egress costs among highest
- Complexity — easy to over-provision
- Support expensive without EA discount
- Savings Plans require annual commitment upfront
Azure
Best for Microsoft shops
- Native M365/Active Directory integration
- Hybrid Benefit: Windows/SQL licenses transfer
- AKS cluster management free
- Dev/Test pricing (60% off)
- Best enterprise agreement pricing
- Strong compliance for regulated industries
- Egress pricing similar to AWS
- Fewer services than AWS
- Steeper learning curve for Linux workloads
- Azure OpenAI can lock into Azure dependency
Google Cloud
Lowest unit compute cost
- Cheapest on-demand compute (~8% below AWS)
- Sustained use discounts — automatic, no commitment
- BigQuery: best serverless analytics pricing
- TPUs for ML workloads (20–80% cheaper than GPU)
- Free GKE cluster management
- Best egress pricing
- Smaller service catalog than AWS/Azure
- Smaller marketplace / ISV ecosystem
- Enterprise support pricing can spike
- Fewer compliance certifications
Real Workload Cost Scenarios (Annual Estimates)
| Workload | AWS Annual | Azure Annual | GCP Annual | Cheapest |
|---|---|---|---|---|
| Web app: 4 m-size VMs + 5TB storage + 10TB egress | $38,400 | $34,200 | $31,800 | GCP (17% below AWS) |
| SaaS startup: 10 VMs + RDS/managed DB + K8s | $84,000 | $72,000 (AKS free) | $76,000 | Azure (14% below AWS) |
| Enterprise: 100 VMs + Windows Server + SQL Server | $420,000 | $195,000 (Hybrid Benefit) | $380,000 | Azure (54% below AWS) |
| Data analytics: 10TB/day BigQuery / Redshift queries | $240,000 (Redshift + storage) | $180,000 (Synapse) | $62,500 (BigQuery on-demand) | GCP (74% below AWS) |
| ML training: GPU cluster, 1,000 hrs/mo A100-equiv | $39,120 | $40,800 | $35,160 (TPU v4 equiv) | GCP (10% below AWS) |
Key insight: Azure wins decisively for Windows/SQL Server workloads via Hybrid Benefit. GCP wins on analytics and raw compute. AWS wins on ecosystem, compliance, and availability.
The Egress Problem: Why Multi-Cloud Is Expensive
AWS Egress Trap
$0.09/GB
AWS charges $0.09/GB for egress to the internet and $0.02/GB between regions. A SaaS app serving 100TB/month pays $9,000/month in egress alone. The AWS Free Tier only covers the first 100GB. CloudFront can reduce egress by 60–80% via edge caching.
Azure + AWS Multi-Cloud Cost
$0.02–$0.05/GB cross-cloud
Running services across AWS and Azure? Data transfer between clouds is charged on both sides. A 10TB/month cross-cloud data pipeline costs $2,000–$5,000/month in pure egress. Avoid cross-cloud data movement wherever possible.
GCP Egress Advantage
$0.08/GB + 200GB free
GCP offers the first 200GB/month of internet egress free and charges $0.08/GB thereafter — 11% less than AWS. For high-egress workloads (media streaming, CDN bypass, API-heavy apps), GCP saves $1,000–$5,000/month per 50TB of traffic vs AWS.
Support Cost Shock
$15K–$100K/yr
AWS Business Support costs 10% of your monthly bill (min $100/mo). At $500K/yr cloud spend, that's $50,000/yr in support alone. GCP Enhanced Support starts at $1,500/mo flat. Azure Professional Direct runs $29K+/mo. Factor support tier into your TCO — it's often invisible until bill shock arrives.
Windows/SQL License Arbitrage
40–60% off on Azure
Azure Hybrid Benefit allows existing Windows Server and SQL Server licenses to transfer to Azure VMs — saving 40–60% vs AWS for identical Windows workloads. A 100-VM Windows deployment costs $420K/yr on AWS vs $195K/yr on Azure with Hybrid Benefit applied.
Commitment Program Complexity
30–72% off if used correctly
AWS Savings Plans require upfront commitment to a $/hr spend level. Azure Reservations require specific VM family. GCP CUDs apply automatically with sustained use — no reservation required for the base 30% sustained use discount. Over-committing on AWS or Azure wastes capital; GCP's automatic discounts reduce that risk.
5 Cloud Cost Optimization Tactics (Works on All Three)
1. Right-size before committing (saves 20–40%)
20–40% savings
The biggest cloud waste is over-provisioned compute. AWS Compute Optimizer, Azure Advisor, and GCP Recommender all provide free right-sizing recommendations. Teams typically find 30–50% of VMs running at <20% CPU utilization. Right-size first — then commit to reservations. Never buy reserved capacity on over-sized instances.
2. Savings Plans / Reserved Instances / CUDs (saves 30–57%)
30–57% savings
All three clouds offer 30–57% discount for 1–3 year commitments. AWS Compute Savings Plans are the most flexible (apply to any EC2, Lambda, Fargate). Azure Reservations require VM family selection. GCP CUDs can be resource-based (locked to instance type) or spend-based (flexible). Target 60–70% coverage of baseline workloads — keep 30–40% on-demand for flexibility.
3. CDN for egress reduction (saves 60–80% on egress)
60–80% egress savings
CloudFront (AWS), Azure CDN, and Cloud CDN (GCP) all reduce egress from compute by caching at edge. At $0.08–$0.09/GB from origin vs $0.01–$0.02/GB from CDN edge, putting 70% of traffic on CDN saves $50K–$300K/yr for high-traffic apps. CDN costs are significantly lower than origin egress.
4. Dev/test auto-stop (saves 15–25% on dev environments)
15–25% savings
Dev and staging environments running 24/7 waste 65% of their cost on nights and weekends. Use AWS Instance Scheduler, Azure Start/Stop VMs, or GCP Cloud Scheduler to stop non-prod instances outside business hours (Mon–Fri, 8am–6pm). This alone saves 15–25% on total cloud bill for companies with large dev environments.
5. Enterprise Discount Program negotiation (saves 15–30% on large spend)
15–30% savings
AWS EDP (Enterprise Discount Program), Azure EA (Enterprise Agreement), and GCP CSA (Cloud Savings Agreement) all provide volume-based discounts of 15–30% for multi-year spend commitments. AWS EDP requires $1M+ annual commitment. Azure EA starts at $100K/yr. GCP CSA starts at $250K/yr. Use competitive bids from all three clouds as leverage — switching threats alone often unlock 10–15% additional discount.
Track Cloud Price Changes Automatically
AWS, Azure, and GCP update pricing constantly — reserved instance rates, spot prices, storage tiers. PricePulse monitors 90+ SaaS and cloud tools and alerts you when costs change.
Get Lifetime Access for $93 Real Case Studies
Series B SaaS — E-Commerce Platform
AWS → GCP migration, 200 VMs
$126K/yr saved
Running all-AWS with 80TB/month egress for image/video serving. Migrated CDN to GCP CDN + moved 40 analytics VMs to BigQuery serverless. Retained AWS for primary application. Annual savings: $72K egress reduction + $54K on analytics compute. Migration took 6 months with zero downtime.
Mid-Market Enterprise — Insurance Firm
AWS → Azure migration, 150 Windows VMs
$225K/yr saved
All workloads were Windows Server + SQL Server. AWS was charging full on-demand rates for licenses. Migrated to Azure with Hybrid Benefit (existing MSDN + EA licenses). Applied 3-year Azure Reservations on all production VMs. Total savings 54% on compute/storage. Migration covered by Azure Migration credits.
Growth-Stage Startup — ML / AI Platform
Multi-cloud consolidation to GCP
$480K/yr saved
Was running GPU training on AWS (A100 instances), storage on S3, inference on GCP. Cross-cloud egress was $40K/month. Consolidated all training and inference to GCP: TPU v4 for training (60% cheaper than AWS A100), GCS for storage, and GKE for inference serving. Sustained use discounts kicked in automatically at 25% usage. Total savings: $480K/year on $1.2M cloud budget.
The Decision Framework: Which Cloud to Choose
| Your Situation | Best Choice | Why |
|---|---|---|
| Greenfield SaaS startup | AWS | Broadest services, best startup credits ($100K+), largest talent pool, deepest marketplace |
| Windows Server / SQL Server workloads | Azure | Hybrid Benefit transfers existing licenses — saves 40–60% vs AWS for same workloads |
| Already using M365 / Active Directory | Azure | Native integration, single vendor, SSO, conditional access — no extra identity costs |
| Heavy analytics / data warehouse | GCP | BigQuery serverless analytics is 50–74% cheaper than Redshift or Synapse at scale |
| ML/AI training workloads | GCP | TPU v4/v5 access, cheapest A100 equivalents, best ML infrastructure (TensorFlow native) |
| High egress / CDN-heavy workloads | GCP | Cheapest egress pricing + free CDN tier; sustained use discounts reduce compute base |
| Enterprise with existing AWS footprint | AWS + optimization | Switching costs outweigh savings. Focus on Savings Plans, right-sizing, CDN instead |
| Regulated industry (FedRAMP, HIPAA) | AWS or Azure | Both have comprehensive compliance portfolios; GCP growing but has fewer certifications |
Related Cloud Cost Guides
- → AWS Cost Optimization Guide — 10 tactics, $72K–$630K savings
- → Azure Cost Optimization Guide — 9 tactics including Hybrid Benefit
- → GCP Cost Optimization Guide — CUDs, BigQuery, Cloud CDN
- → Kubernetes Cost Optimization — Works on EKS, AKS, GKE
- → Snowflake Cost Optimization — Multi-cloud warehouse savings
Stop Overpaying for Cloud Infrastructure
PricePulse tracks cloud pricing changes across AWS, Azure, GCP, and 90+ SaaS tools. Get alerted when prices change — before your renewal locks you in.
Get Lifetime Access for $9