A few months ago I tried to find a "best stack under 100 USD a month" article for a side project. Every result was either an enterprise list with line items at thousands per month, or a vague "use Vercel and Supabase" post with no actual numbers.

So I built one. The result is cheapstack.dev — an editorial site that publishes itemized reference stacks for indie SaaS infrastructure on 50–200 USD a month, with every line item, the date prices were verified, and what to swap when traffic crosses specific thresholds.

I'm not going to pretend this is a masterpiece. I want to write this post because the things I learned along the way will probably save someone else two weeks.

The build

Stack:

• Next.js 16 with output: "export" for static export
• Cloudflare Workers Static Assets for hosting (free, unlimited bandwidth)
• IBM Plex Sans + Plex Mono for typography
• Markdown-driven content layer (no CMS, no backend, no database)

Two-week evening build, mostly weekends. Lighthouse 100/100/100/100 across all pages.

Why Cloudflare Workers, not Vercel

Cheapstack's whole pitch is "real numbers for cheap stacks." Hosting it on Vercel — whose Hobby tier is explicitly non-commercial, and whose Pro tier starts at 20 USD a month — would have been hypocritical the moment I added an affiliate link.

Cloudflare Workers Static Assets bills 0 USD a month, even with commercial use, and edges out Vercel on first-byte latency in 60% of regions. The DX is rougher (you can't just next build and deploy — you need output: "export" and a wrangler config), but for a site that's truly static the trade is worth it.

I tried @opennextjs/cloudflare first because Cheapstack started life as a dynamic Next.js app with edge runtime OG images. Two days of debugging dynamic-route 404s later, I realized I didn't need any of that. Pre-generating the OG image as a 1200x420 PNG and switching to plain static export resolved everything in 30 minutes.

Lesson: if your site has no per-user state and no per-request logic, static export beats every adapter. Adapter complexity buys you flexibility you don't have.

The content layer

I went into this thinking I'd write everything from scratch. Then I realized the actual work was structuring the data so the prose was a thin layer on top.

Every tool on the site lives in one file:

// src/lib/tools.ts
type Tool = {
  slug: string;
  name: string;
  category: ToolCategory;
  pricing: { tier: string; usd: number; verified: string }[];
  affiliate: string | null;
  // ... more fields
};

export const TOOLS: Record<string, Tool> = {
  digitalocean: { ... },
  postmark: { ... },
  // ~70 tools
};

Stack guides, comparison pages, and tool roundups all pull from this single registry. When DigitalOcean changes their pricing, I update one entry and 11 pages reflect it. When an affiliate program approves me, I drop the tracking URL into the affiliate field and every link to that tool across the site updates.

This is mundane infrastructure but it's the difference between a site that decays in 3 months and a site that stays accurate for years.

Affiliate programs: the reality of being new

I applied to 12 affiliate programs in week one. Here's the score so far:

So far: 0 approved, 2 declined, 1 pending, 9 deferred. The "site too new" pattern across every network was demoralizing — I'd built the whole monetization architecture on the assumption that applying with a live, polished site would be enough.

It's not. Affiliate networks gate on traffic signal, not site quality. The site needs to show up in Google for 30+ days with consistent search clicks before most programs unlock.

This was a real lesson: the FTC disclosures on cheapstack are mostly aspirational at the moment. Every tool link has rel="sponsored nofollow noopener" and a clear disclosure — even though most of them don't pay me anything. I left the architecture in because (a) when programs do approve me, the wiring is already there, and (b) the credibility hit of adding affiliate disclosures later would be worse than declaring intent on day one.

What's actually on the site

For the curious:

• 6 stack guides — itemized monthly bills for SaaS-100 (~87 USD/mo), MVP-50, marketplace-150, course-100, newsletter-25, landing-15
• 8 comparisons — Firebase vs Supabase, Postmark vs Resend, Beehiiv vs Ghost, Neon vs PlanetScale, Clerk vs Auth.js, Vercel vs Cloudflare Pages, Stripe vs Lemon Squeezy, Stripe vs Paddle
• 12 tool roundups — hosting, databases, transactional email, newsletter platforms, domain registrars, SEO tools, CRM, landing builders, payments, analytics, headless CMS, auth services
• 8 framework deep-dives — Next.js, Astro, SvelteKit, Remix, Nuxt, SolidStart, Qwik, RedwoodSDK

Everything's at cheapstack.dev if you want to see how it shipped.

Early signal

Four days post-launch, before any external traffic push:

• Google Search Console: 41 impressions, position 12.1 average
• Indexed: 8 of 47 pages (normal week-1 curve)
• Most interesting query: "affordable node.js headless cms under 50/month" — exactly the kind of long-tail commercial-intent query the site was designed for

Position 12 means page 2 of Google. With backlinks and another month of indexing it could realistically move to page 1 for that query. That's the whole game.

What I'd do differently

1. Skip the OpenNext adapter entirely. Static export from day one. Two days saved.
2. Apply to affiliate programs only after 30 days of traffic. Day-one applications are a coin flip and rejection cool-down windows are 60–90 days.
3. Write less, structure more. The articles that are working are the ones with strong data tables and clean comparisons. Prose-heavy pieces sit in the middle of the rankings.
4. Start the newsletter on day one. I held off until "the site was ready." That was wrong. The newsletter is the only thing you actually own — everything else (Google rank, affiliate approvals, social reach) is rented from someone else.

What's next

• More long-tail comparison pages targeting the queries already showing up in GSC
• A weekly newsletter when I cross 100 organic visits/day
• Real receipts from working indie builds (if you're running profitable on a 50–200 USD/mo stack, I'd love to feature your build — with permission)

If you've built anything similar — editorial site, affiliate engine, programmatic SEO play — I'd love to hear what worked and what didn't. The two-week timeline only worked because I made a lot of decisions fast; some of them are probably wrong.

Site: cheapstack.dev