Backend Pricing Reality Check: Supabase vs Firebase vs AWS Amplify

Supabase: The "Please Don't Bankrupt Me" Option

Supabase is the only one that won't surprise you with a bill that makes you question your life choices. Free tier gives you 500MB database, 50k monthly active users, and 5GB egress. Projects pause after a week of inactivity (annoying but understandable). Their real-time benchmarks show they can handle up to 10,000 concurrent clients.

Real talk on Supabase costs:

• Free tier: Actually usable for real apps, unlike others that tease you
• Pro at $25/month: Gets you 8GB database, 100k MAU, 250GB egress, plus $10 compute credits
• Additional MAU: $0.00325 each after 100k (this actually matters)
• Database storage: $0.125/GB after the included 8GB
• Spend caps: They actually work - service pauses instead of charging you into bankruptcy

The kicker: Everything's included. Auth, real-time, storage, APIs. No surprise "oh you wanted authentication to work? That's $50/month extra" bullshit like other platforms pull.

Built a SaaS with around 30k users, Supabase was costing me like $75/month including compute. Same exact app on Firebase? Bill was over $250 the first month because I forgot to add .limit(100) to a query and it was pulling entire collections. Spent 3 hours at 2am figuring out why the bill exploded. Firebase dashboard just shows "document reads: 2.3M" without telling you which fucking query caused it.

Firebase: Where Bills Go to Die (Your Wallet)

Firebase pricing is fucking brutal - looks cheap until you actually use it. Two plans: free Spark and paid Blaze. The official pricing documentation reveals the brutal truth behind their "simple" model.

Spark plan reality check:

• 50k Firestore reads per day sounds generous until you realize a simple user dashboard burns through that in hours
• 20k writes per day? One social media post with reactions will eat that alive
• 1GB storage? Good for maybe a few thousand user profiles
• Cost optimization patterns become essential from day one

Blaze plan gotchas that hurt:

• Firestore operations pile up fast - every query, every listener, every single document read charges you
• Cloud Functions look free (2M invocations) until you hit the GB-seconds cost
• No base fee means death by a thousand cuts - each service nickels and dimes you
• Best practices help but can't eliminate the fundamental per-operation billing model
• High traffic cost management becomes critical for scaling apps

Client's Firebase bill exploded when their workout app hit Product Hunt - went from like $40 to over $800 in a day. Real-time listeners in the leaderboard component were triggering on every single user update. The fucking onSnapshot() was listening to the entire /users collection instead of just the top 10. Spent two weeks debugging this while the bill kept climbing. Firebase console showed millions of document reads but didn't tell us which query was the culprit.

After the left-pad incident, you'd think I'd learn to be more careful with third-party dependencies.

AWS Amplify: "Surprise, You're Broke!"

AWS Amplify pricing is what happens when enterprise architects design pricing for indie developers. Every service has its own billing model, and they all add up faster than you can say "serverless." The official Lambda pricing alone has multiple dimensions that most developers underestimate.

Current Amplify reality (last time I checked):

• Frontend hosting: Free tier includes 100 GB-hours/month, then $0.20/hour
• Build minutes: First 1,000 free, then $0.01 per minute
• Data transfer: 15GB free, then $0.15/GB (this one kills you)
• Lambda cold starts now cost money for init time - not sure when that started but it's fucking annoying

The AWS ecosystem tax:

• DynamoDB pricing: Looks cheap until you need consistent reads and global tables
• Lambda: $0.20 per million requests, but memory pricing and cold starts add up quickly
• Cognito: $0.0055 per monthly active user after 50k
• S3: Storage is cheap, data transfer isn't
• Lambda cost optimization strategies become mandatory for production apps

One client's "simple" employee directory was costing him over $300/month on AWS because he deployed Lambda in us-east-1 and RDS in us-west-2. Data transfer charges murdered him - $0.02/GB adds up fast when your Lambda makes thousands of DB calls per execution. Most of the bill was data transfer costs, not the actual services.

AWS has 267 different services and each one has its own billing dimension that you'll discover when your boss asks why the "free" app costs $300/month. The great Log4j panic of 2021 taught us nothing about dependency complexity.

The Brutal Truth About Free Tiers

Supabase: Free tier works for side projects and early development. You can actually launch on it. Projects pause after a week of inactivity, but that's fair for free hosting.

Firebase: Free tier is a trap designed to get you hooked. You'll outgrow it in days with real usage. The quotas look generous until you build something people actually use.

AWS: Free tier expires after 12 months, then you hit the pricing cliff. It's like a gym membership - great deal for the trial, then reality hits your credit card.

What Nobody Tells You About Scaling Costs

Supabase at scale: Scaling costs are pretty predictable. One client went from like $75 to maybe $200 over 6 months when their users exploded. No surprises.

Firebase at scale: Costs become unpredictable. Operations-heavy apps get murdered by Firestore pricing.

AWS at scale: If you know what you're doing, it's cheapest. If you don't, it's most expensive. There's no middle ground.

The Decision Matrix Nobody Talks About

Here's what really happens when you're lying awake at 3am wondering if you made the right choice:

Supabase: You sleep well knowing your bill is predictable. Worst case, service pauses instead of bankrupting you.

Firebase: You wake up to check if your bill exploded overnight. Sometimes it did, sometimes it didn't.

AWS: You gave up sleeping months ago because managing costs has become a full-time job.

The Honest Recommendation by Team Size

Solo developer/indie team:
Start with Supabase. Seriously. The free tier works, the paid tier is predictable, and you won't need a second mortgage to scale.

Small team (2-5 devs):
Still Supabase, unless you're building something that needs Firebase's offline sync or AWS's specific services. Your time is better spent building features than optimizing cloud costs.

Mid-size team (6-20 devs):
Now we're talking. Firebase works if someone on your team actually understands Google Cloud (hint: the certification they got last year doesn't count). AWS works if someone speaks fluent CloudFormation and can debug IAM policies at 3am (spoiler: they probably can't). Since v3.2.0 broke the API, even "experts" are googling basic shit. Technology decision frameworks emphasize team expertise over theoretical performance.

Large team (20+ devs):
AWS probably wins for pure cost optimization, but you'll pay for that in engineering time. Firebase if you want Google to handle the infrastructure headaches. Supabase if you want to focus on product instead of devops. Enterprise backend comparisons show PostgreSQL's transparent pricing advantage at scale.

Real Talk Based on What I've Seen

"We're bootstrapping": Supabase. Don't even think about it. Predictable costs beat fancy features when you're counting pennies.

"We just raised Series A": Still Supabase until you prove people actually want your product. Then maybe Firebase if you need to scale fast.

"We're enterprise": AWS if you have someone who speaks fluent AWS. Otherwise Supabase and save yourself the headache.

"We already use [Google Cloud/AWS]": Stick with your ecosystem, but budget for the reality that your existing discount won't apply to new services. Found out the hard way that our enterprise discount only covers certain services. Bill shock when marketing counted on that discount for everything. Multi-cloud cost analysis reveals hidden integration expenses.

The Migration Reality

Nobody talks about this, but switching platforms sucks. A lot. Database migration cost comparisons show DynamoDB can be 50-70% cheaper than Firebase but migration complexity matters more.

From Supabase: PostgreSQL dump makes data migration easier, but you'll rebuild auth and real-time features. PostgreSQL migration tools help with the database layer.

From Firebase: You're stuck for a while. Firestore data export is painful and Cloud Functions don't port cleanly. Firestore vs DynamoDB comparisons highlight the architectural differences.

From AWS: Depends on how deep you went. Lambda functions migrate okay, but if you used 20 different AWS services, good luck. Multi-cloud strategies require planning from day one.

Bottom line: Pick the platform you can live with for at least 2 years.

The Costs You're Not Thinking About

Supabase: Your biggest risk is outgrowing their scale. Team plan at $599 is expensive for smaller teams. Supabase scaling limits define when you need to consider alternatives.

Firebase: Your biggest risk is a viral spike destroying your runway. I've seen bills jump to over $2k overnight. The Firebase community is full of horror stories about unexpected costs from traffic spikes.

AWS: Your biggest risk is complexity. You'll spend more on DevOps expertise than you save on infrastructure. The learning curve and ongoing management overhead often outweigh the theoretical cost savings.

My Actual Advice

Just start building.

Seriously. You can bikeshed platform decisions for weeks while your competitor ships their MVP and steals your market. I've seen teams spend 3 months evaluating backends and then run out of runway before launching. Pick Supabase (safest bet), ship your MVP in a week, and make money. Perfect is the enemy of shipped.

If you outgrow Supabase, congratulations - you have a real business and can afford to hire someone to migrate to AWS. That's a good problem to have.

If Firebase surprises you with a huge bill, at least you validated product-market fit. Now you know people actually use your product.

If AWS complexity kills your momentum... well, you learned an expensive lesson about premature optimization. Sometimes the "wrong" choice that ships beats the "right" choice that never launches.

Pick whatever lets you ship fastest. Your users don't give a shit about your architecture diagram - they want their problem solved.

Bottom line: Stop overthinking this. Pick Supabase, ship your MVP, make money. You can always migrate later when you're rich.