Open source does not mean free

Free to Download, Costly to Run

01

On the left, a small download box labeled 'GitHub Code: ₹0' has an arrow pointing to a large cross-section iceberg diagram in water. Above the water line, a tiny visible tip is labeled 'Code License: ₹0'. Below the water line, three large descending blocks show 'EC2 & SSD Storage', 'Multi-AZ Data Egress', and '24/7 DevOps Payroll: ₹30L+'. Arrow points from the hidden base to a final total labeled 'Real Monthly Expense'.

Notes

Open-source software is free like a stray puppy, not free like a complimentary cup of cutting chai. You pay zero rupees to clone Apache Kafka from GitHub, but that ₹0 sticker price vanishes the moment you hit deploy. Running Kafka for a Swiggy-like delivery tracking system demands round-the-clock compute servers, SSD storage, backups, and network bandwidth. Assuming open source is automatically cheaper than cloud services ignores the massive hidden iceberg of operational overhead.

Kafka vs SQS: The Real Math

02

Two parallel horizontal tracks show contrasting architectures. The top track shows a single serverless cloud box labeled 'AWS SQS' taking spiky input arrows and outputting to a bill box marked '₹1,750 / month'. The bottom track shows a cluster of six interconnected server boxes (three Brokers, three Controllers) with continuous power meters, outputting to a baseline bill box marked '₹40,000+ / month fixed cost'.

Notes

AWS SQS charges roughly ₹35 per 10 lakh messages, scaling completely down to ₹0 when your store closes at midnight. A production-ready Kafka setup needs at least 3 broker servers and 3 controller nodes running 24/7 across three Availability Zones to prevent downtime. Even before processing a single grocery order on Blinkit, your baseline Kafka server bill runs at ₹35,000 to ₹50,000 every single month. For 5 crore messages a month, SQS costs barely ₹1,750, while idle Kafka infrastructure burns twenty times that amount just staying alive.

The Invisible DevOps Salary Tax

03

In the center, a large clock reads '2:15 AM' above an alert siren flashing over a cracked server cylinder labeled 'Kafka Broker 02 Full'. Two engineer figures with laptops have dashed red panic lines connecting to the cracked cylinder, flanked by salary banners reading '₹35L/yr each'. To the right, a single green box labeled 'Managed Cloud SQS' has an automated recovery loop and a shield icon labeled 'Zero Maintenance Payroll'.

Notes

When a Kafka disk fills up at 2:00 AM during an India-Pakistan IPL match on Hotstar, AWS will not step in to save you. You need at least two senior Site Reliability Engineers earning ₹35 Lakhs per year each on standby for upgrades, partition rebalancing, and crash recovery. That adds ₹70 Lakhs annually in engineering payroll just to babysit a tool that carried a ₹0 price tag on GitHub. Managed cloud services offload 99% of that operational headache to cloud providers, freeing engineers to build revenue-generating features instead.

The Licensing Trap: tldraw & BSL

04

On the left, a repository box labeled 'Library Repo (tldraw / Redis)' shows a green open-source badge that splits into two directional pathways. The top path leads to an individual student icon labeled 'Personal Projects: Permitted (₹0)'. The bottom path leads toward a commercial fintech app box, blocked by a prominent red barrier sign labeled 'BSL / Commercial License Required ($$$)' with a legal notice icon pointing to an audit invoice.

Notes

Developers often believe open-source code gives them permission to build commercial apps with zero restrictions. Canvas libraries like tldraw look free on GitHub, but their commercial terms require paid business licenses if you remove their watermark or build enterprise tools. Heavyweights like Elastic, MongoDB, and Redis abandoned traditional Apache 2.0 licenses for restrictive Business Source Licenses (BSL) to block free commercial exploitation. If a Bengaluru fintech startup embeds these libraries blindly into their UPI payment dashboard, an audit can trigger severe legal penalties and forced licensing fees.

When to Self-Host vs Buy

05

A line chart with x-axis 'Monthly Event Volume' and y-axis 'Total Cost in ₹'. A dashed orange line for 'Managed Cloud' starts at near zero and slopes upward at a steep 45-degree angle. A solid blue line for 'Self-Hosted OSS' starts high at a ₹3,00,000 monthly baseline (servers + SRE payroll) and slopes gently. A red circular marker highlights the intersection point labeled 'Break-Even Threshold (50M+ events/month)'.

Notes

For early-stage startups handling under 1 crore events monthly, serverless managed tools like SQS, DynamoDB, and managed queues are far cheaper. At massive scale—like PhonePe processing 700 crore UPI transactions a month—cloud vendor markups cross ₹50 Lakhs monthly, making dedicated SRE teams cost-effective. Self-hosting only saves money when the managed cloud markup is significantly higher than the ₹70 Lakh annual engineering salary required to maintain it. Defaulting to open-source software solely because the license is free is an expensive rookie mistake that drains company capital.