SECTION 01
Bottom Line: Free Tiers Can Handle Thousands of Monthly Users
Here's the short answer: free tiers can comfortably handle a few thousand monthly page views with dozens of concurrent users. This isn't theoretical — it's a range that has worked reliably across multiple shipped products.
The specific combination is Vercel (frontend + API), Neon (database), and Auth.js (authentication) — all at zero monthly cost. With these three, you can ship a login-enabled web app without paying anything.
After years of trial and error, this starting stack has become the default launch configuration for every new service. The priority is shipping fast and validating, and zero upfront cost fits that perfectly.
There's no reason to spend on infrastructure before validating demand. Proving the concept should come before paying for scale. If a service can't work within free-tier constraints, the design itself needs rethinking.
SECTION 02
Free Tier Limits by Service Category
Not all free tiers are created equal. The bottleneck varies by service — bandwidth, storage, compute time, or request count — so you need to understand what each one limits.
Frontend + API options include:
- Vercel Hobby: 100GB bandwidth/month, 10-second function limit
- Cloudflare Pages: effectively unlimited bandwidth, Workers capped at 100K requests/day
- Netlify: 100GB bandwidth/month, 125K function executions/month
Database options to consider:
- Neon Free: 0.5GB storage, 190 compute hours/month
- Supabase Free: 500MB storage, pauses after 1 week of inactivity
- Turso Free: 9GB storage, edge-native SQLite
Authentication has generous free options:
- Auth.js: fully free and open-source, self-hosted
- Supabase Auth: free up to 50K MAU
- Clerk: free up to 10K MAU, then per-user charges
Storage and payments round out the stack:
- Cloudflare R2: 10GB free, zero egress fees
- Vercel Blob: 250MB free
- Stripe: no monthly fee, charges only on successful transactions
SECTION 03
The Sleep Problem — and How Serverless Solves It
The single most damaging free-tier issue is server sleep — where your app goes dormant after a period of inactivity. Cold starts can take tens of seconds, and users often assume the service is broken.
Traditional always-on servers on free plans will almost always have this problem. Serverless platforms like Vercel and Cloudflare Workers don't sleep because they spin up on demand — there's no persistent server to shut down.
This is a major reason Vercel sits at the center of the default stack. A free tier that doesn't sleep is ideal for indie projects where traffic is small and irregular.
If you must use a traditional server, you can set up external health-check pings to keep it warm. But that's a workaround, not a solution. Moving to serverless is the cleaner path.
SECTION 04
Stack Example 1: Web App (Next.js + Neon + R2 + Stripe)
Here's the actual configuration in use. Next.js on Vercel combines the frontend and API Routes into a single deployment. No separate backend server to manage.
The database is Neon (PostgreSQL) — a serverless Postgres that pairs well with Vercel, including built-in connection pooling. Auth.js handles authentication on top of this.
For services with file uploads, offloading media to Cloudflare R2 is the key move. R2's zero egress fees mean you're not burning through Vercel's 100GB bandwidth cap on images. This optimization extends free-tier life significantly.
If the product needs payments, Stripe plugs in with zero fixed costs. You only pay when transactions happen. This aligns perfectly with the zero-cost-until-revenue philosophy.
This stack comfortably handles a few thousand monthly page views and a few hundred active users within free limits. Paid plans only enter the picture when traffic clearly exceeds this range.
SECTION 05
Stack Example 2: Mobile App (React Native Expo)
For mobile apps, the backend reuses the same Neon + Vercel API Routes from the web stack. The frontend is built with React Native via Expo, calling the shared API. No duplicate infrastructure needed.
Expo's free plan includes 30 builds per month and EAS Update delivery. For the release cadence of a solo developer, 30 builds per month is more than enough. OTA updates let you push fixes without going through store review.
However, mobile has unavoidable fixed costs:
- Apple Developer Program: ~$99/year
- Google Play Developer: ~$25 one-time registration fee
Even with server costs at zero, store fees are non-negotiable. "Completely free" isn't possible on mobile. That said, keeping everything else at zero means annual costs are limited to store maintenance alone.
SECTION 06
Three Ways Free Tiers Can Burn You
Free hosting works, but there are three traps that catch indie developers off guard. Each of these has caused real scrambles in past projects.
Trap 1: Sudden free-tier removal. Major platforms have eliminated free plans with little notice in the past, forcing mass migrations on short timelines. Never go all-in on one provider — always have one alternative in mind.
Trap 2: Uncapped usage billing. When you upgrade to a paid plan, some services auto-charge for overages with no ceiling. Always set a Spend Limit on platforms that offer it. Skipping this step can turn a viral moment into a billing shock.
Trap 3: Sleep-induced user abandonment. As covered earlier, cold starts of 30+ seconds make users think your service is dead. Serverless architectures eliminate this entirely — it's the cleanest fix available.
SECTION 07
When to Switch from Free to Paid
Hitting a free-tier limit doesn't automatically mean it's time to pay. The real question is whether the service is generating revenue, not whether it's generating traffic.
A product that was eventually sold via M&A ran on near-zero infrastructure costs during its growth phase. Paying for infrastructure made sense only after monthly revenue was consistently flowing. Reversing that order just extends the burn period.
If a service is barely running within free limits and has no revenue path, the problem isn't infrastructure — it's product-market fit. Upgrading won't fix that. Validate the business before scaling the server.
When the time does come, the first paid tier looks like this:
- Vercel Pro: $20/month
- Neon Launch: $19/month
- Total: roughly $40–80/month
That's an amount that makes sense only with stable monthly revenue behind it. Committing to this cost at zero revenue is a bet that rarely pays off for solo developers.
SECTION 08
Keep Cloudflare Tunnel as a Backup Option
Locking into Vercel exclusively creates concentration risk. Cloudflare Tunnel offers an alternative path — it lets you expose a self-hosted server through Cloudflare's network without opening ports directly.
Cloudflare's free tier is exceptionally generous across CDN, DNS, and security. For media-heavy services, the effectively unlimited bandwidth is a significant advantage over other providers.
Concentrating everything on one platform means any policy change hits your entire stack. Keeping Cloudflare as a secondary option — even without actively using it — builds resilience against free-tier removal risks.
The principle is don't lock into a single provider. Staying familiar with at least one alternative makes the inevitable free-tier changes far less disruptive.
SECTION 09
Design Tips for Maximizing Free-Tier Longevity
The most effective way to stay within free limits isn't optimizing infrastructure — it's designing the product around free-tier constraints from the start. Retrofitting is always harder than building it in.
One deliberate design choice: make the product usable without requiring login. Forcing sign-up on a low-traffic service increases drop-off. A no-login entry point maximizes learning from limited traffic.
On the infrastructure side, route all media through Cloudflare R2 to preserve Vercel's bandwidth for HTML and API responses. This single optimization can extend free-tier viability by months.
The deeper mindset shift is this: don't design around constraints — ask whether the product works within them. Instead of cutting features to fit limits, ask whether the core value proposition can be delivered within those boundaries.
SECTION 10
Free Tiers Are a Validation Framework, Not a Limitation
With the right stack, free hosting handles thousands of monthly users without breaking. This is a tested baseline, not an optimistic guess.
The default starting stack in summary:
- Frontend + API: Vercel (Next.js)
- Database: Neon (PostgreSQL)
- Auth: Auth.js
- Storage: Cloudflare R2
- Payments: Stripe
Switch to paid only after revenue is established. Paying for infrastructure before validating the business means investing in scale before confirming there's something worth scaling.
Think of free tiers not as ceilings but as a validation framework — a bounded space to prove the product works. Ship, validate, then invest. That sequence is the foundation of sustainable indie infrastructure strategy.
