AWS Lambda: When Serverless Actually Costs More

💰 The Promise

Serverless is supposed to be cheaper. Pay only for what you use. No idle costs. Perfect for APIs with variable traffic.

So I moved my API to AWS Lambda. And my costs went up.

The reality: Serverless isn't always cheaper. Here's when it costs more.

📊 The Numbers

My API stats:

• 10 million requests/month
• Average response time: 200ms
• Peak traffic: 100 requests/second
• Average memory: 256MB

Traditional Server Cost

EC2 t3.medium instance:

• Instance: $30/month
• Data transfer: $10/month
• Total: $40/month

AWS Lambda Cost

Lambda pricing:

• Requests: 10M × $0.20 per 1M = $2.00
• Compute: 10M × 0.2s × 0.256GB × $0.0000166667 = $85.33
• Data transfer: $10/month
• Total: $97.33/month

Lambda is 2.4x more expensive!

❌ When Lambda Is Expensive

1. High Request Volume

At 10M+ requests/month, Lambda compute costs add up quickly. Traditional servers have fixed costs.

2. Long Execution Times

Lambda charges per 100ms. If your functions take 200ms+, costs multiply.

3. Consistent Traffic

If traffic is steady, you're always paying. A server is cheaper at constant load.

4. Cold Starts

Cold starts add latency and cost. Frequent cold starts make Lambda inefficient.

✅ When Lambda Is Worth It

• Sporadic traffic: Low, unpredictable load
• Short functions: < 100ms execution time
• Event-driven: Triggered by events, not constant requests
• Low volume: < 1M requests/month

💡 Cost Optimization Tips

1. Use Provisioned Concurrency

Eliminates cold starts, but adds fixed costs. Only worth it if you have steady traffic.

2. Optimize Memory

Right-size memory allocation. More memory = faster execution = lower cost (if it reduces execution time).

3. Use Lambda@Edge

For simple responses, Edge functions are cheaper than regular Lambda.

4. Consider Fargate

For consistent traffic, Fargate can be cheaper than Lambda.

📊 Cost Comparison Table

🎯 My Decision

I moved back to EC2. For my traffic pattern (10M+ requests/month, steady load), it's 2.4x cheaper.

Lambda is great for some use cases, but it's not always cheaper. Do the math before you switch.

💡 Key Takeaways

• Serverless isn't always cheaper
• High volume + steady traffic = EC2 is cheaper
• Low volume + sporadic traffic = Lambda is cheaper
• Do the math before you decide
• Consider all costs: compute, requests, data transfer

Don't assume serverless is cheaper. Calculate the actual costs for your use case.