Chapter 10: AWS EC2 Pricing
AWS Cloud EC2 Pricing — the part that makes beginners nervous (“Will I get a surprise ₹50,000 bill?”) but is actually very flexible and beginner-friendly if you understand the rules.
Think of EC2 pricing like renting a bike/scooter in Hyderabad:
- You can pay per minute (no commitment — expensive but flexible)
- Commit to monthly/annual pass for big discount
- Grab cheap “spot” bikes when available (but they might get taken back suddenly)
- Or rent a dedicated bike just for you (more expensive, but guaranteed)
AWS gives you similar choices for EC2 instances. Pricing is region-specific (cheaper in Mumbai/Hyderabad regions for us in India), per-second for most cases (with 60-second minimum), and depends on instance type, OS (Linux cheaper than Windows), and model you pick.
Let’s break it down step-by-step like a real classroom session — models, examples, costs (approx 2026 in ap-south-1 Mumbai region), tips to avoid bills, and real Hyderabad stories.
1. Core Pricing Basics (Applies to Almost Everything)
- Billing unit: Per second (minimum 60 seconds per instance start/stop cycle) for Linux, most Windows, RHEL, etc. → Great for short tests!
- What you pay for: Instance running time + attached EBS storage + data transfer out (in usually free up to limits).
- Free Tier reminder: First 12 months — 750 hours/month of t3.micro or t4g.micro (enough for one small instance 24/7) → usually ₹0 if you stay within.
- Region matters: ap-south-1 (Mumbai) and ap-south-2 (Hyderabad) are among the cheaper regions globally (~20–40% less than US/Europe for many types).
- Currency: Billed in USD, converted to INR on your card (watch exchange rate + taxes ~18% GST).
- No surprises tip: Set a billing alarm at ₹500/month right after signup (Billing console → Budgets).
2. The Main EC2 Pricing Models (6 Key Ones in 2026)
| Model | Commitment | Discount vs On-Demand | Best For (Beginner View) | Risk / Flexibility | Approx Savings |
|---|---|---|---|---|---|
| On-Demand | None | 0% (base price) | Testing, learning, unpredictable traffic, short projects | Full flexibility, no lock-in | Baseline |
| Spot Instances | None | Up to 90% | Fault-tolerant jobs (batch, CI/CD, rendering, data analysis) | Can be interrupted (with 2-min notice) | Highest |
| Savings Plans | 1 or 3 years (USD/hour commit) | Up to 72% | Steady workloads (production apps, always-on servers) | Flexible across instance types/families | Very high |
| Reserved Instances (RIs) | 1 or 3 years | Up to 72% | Predictable, long-term in specific AZ/region | Less flexible than Savings Plans | High |
| EC2 Capacity Blocks for ML | Time-block reservation | Varies (recent hikes) | Planned GPU/ML training (P5, Trn2 etc.) | Reserved for specific hours | For ML only |
| Dedicated Hosts / Instances | On-Demand or committed | Varies | License-bound software (Windows Server, Oracle) | Dedicated physical server | For compliance |
Big news 2026: Savings Plans are now the most popular (more flexible than old RIs). Spot still king for savings if your app can handle interruptions. GPU/ML Capacity Blocks saw ~15% price increase in Jan 2026 due to high demand (e.g., p5e.48xlarge from ~$34–$40/hour).
3. On-Demand Pricing (The One Beginners Start With)
This is pay-as-you-go — no upfront, no commitment.
- Charged from launch to stop/terminate.
- Per-second billing (60-sec min).
- OS matters: Linux/Amazon Linux cheapest → Windows + SQL adds licensing fee.
Approximate 2026 On-Demand prices in ap-south-1 Mumbai (USD/hour — convert ~₹83–85/USD):
| Instance Type | vCPU / RAM | Linux/Unix (~USD/hr) | Windows (~USD/hr) | Monthly 24/7 (Linux, approx INR) | Use Case Example |
|---|---|---|---|---|---|
| t4g.micro (free tier favorite) | 2 / 1 GiB | ~$0.0028–0.004 | N/A | ₹150–300 (but free first year) | Learning, small script |
| t3.medium | 2 / 4 GiB | ~$0.03–0.04 | ~$0.08–0.10 | ₹1,800–3,000 | Small web app backend |
| m8g.large (Graviton) | 2 / 8 GiB | ~$0.04–0.06 | N/A | ₹2,500–4,000 | Balanced modern app |
| c8g.2xlarge | 8 / 16 GiB | ~$0.25–0.35 | N/A | ₹15,000–25,000 | Video processing |
| g6.xlarge (GPU) | 4 / 16 GiB + GPU | ~$0.8–1.2 | N/A | ₹50,000+ | Small AI inference |
Hyderabad example: Student launches t4g.micro for project → runs 100 hours testing → bill ~₹20–50 (or ₹0 in free tier). Leaves running 1 month → ~₹200–400.
4. Spot Instances (Cheapest – But Interruptible)
- AWS sells spare capacity — up to 90% off On-Demand.
- You set max price (or “persistent” request).
- AWS can reclaim with 2-min warning if demand rises.
- 2026 tip: Use Spot Fleet or Spot Blocks for more control.
Example: Batch job encoding 100 videos → Spot c8g.2xlarge at ~70–90% off → cost drops from ₹20,000 to ₹2,000–5,000/month.
Hyderabad startup story: A gaming company runs ML model training overnight on Spot p5 instances → saves 80% vs On-Demand → affordable even for small teams.
5. Savings Plans & Reserved Instances (For Steady Workloads)
- Commit to $X/hour for 1/3 years → up to 72% off.
- Compute Savings Plans — most flexible (any instance family, region, OS).
- EC2 Instance Savings Plans — deeper discount but tied to family/region.
Example: Commit $10/hour for 1 year → run m8g.large equivalents → effective rate ~30–50% lower than On-Demand. Many Hyderabad SaaS companies use this for prod servers.
6. Other Costs to Watch (Not Just Instance)
- EBS storage: ~$0.08–0.10/GB-month (gp3 SSD).
- Data transfer out: First 100 GB/month free → then ~$0.08–0.09/GB.
- Public IPv4: ~$0.005/hour if used.
- Avoid bill shock: Use AWS Cost Explorer + Budgets + Trusted Advisor.
7. Quick Tips for Hyderabad Beginners (2026)
- Start On-Demand in free tier (t4g.micro Graviton — cheapest + efficient).
- Always choose ap-south-1 or ap-south-2 (lowest latency + cheaper).
- Set budget ₹500–1,000/month + alerts at 80%.
- Use AWS Pricing Calculator (calculator.aws) — add EC2 → select region → play with types.
- Terminate instances when not using (stop ≠ terminate — stop still charges for EBS!).
- For production → move to Savings Plans after 1–2 months.
Real story: Friend in Gachibowli left EC2 running 3 months unused → ₹8,000 bill. Lesson: Set alarms + auto-stop scripts!
Got it? EC2 pricing is flexible pay-what-you-use — not scary if you plan.
Next?
- Walk through Pricing Calculator for your instance?
- Spot vs Savings Plans deep dive?
- How to set billing alerts step-by-step?
Tell me — next class ready! 🚀💰
