Building software has never been cheaper. Cloud hosting, databases, AI APIs: you can run a real product for the cost of a few coffees. But if infrastructure is so cheap, why does building software still feel expensive?
Because the real cost was never the servers. It's human time.
I'm building Alydo, a grocery tracking app, as a solo founder. Three weeks in, I want to share every number: infrastructure, tools, traffic, conversions, and what I think it all means for indie development.
The thesis: AI tools are compressing the time cost of building software. This could mean more sustainable products, built at smaller scale, that actually prioritise users over growth metrics.
All figures in AUD.
The Infrastructure: $131/month
Here's what it costs to run Alydo:
| Service | Cost | Purpose |
|---|---|---|
| Vercel | $25/month | Hosting, serverless functions |
| Supabase | $25/month | Database, auth, storage |
| Gemini API | $5/month | AI receipt parsing |
| Resend | Free | Transactional emails |
| Upstash Redis | Free | Rate limiting, caching |
| Domain | $12/year | alydo.app |
| Claude Code | $75/month | AI-assisted development |
| Total | ~$131/month |
A few notes:
The AI costs surprised me. Gemini handles the core feature of Alydo (parsing receipts with AI), and it costs $5/month. Five dollars. The “expensive AI” narrative doesn't match my experience.
I've included Claude Code in this breakdown. It's not infrastructure, it's a development tool. But it's real money spent on building the product, so it belongs here.
I could run cheaper on free tiers, but I chose paid plans for reliability. Even so: $131/month. That's the cash cost of running a real, AI-powered product.
The Real Cost: Time
Infrastructure is 1.6% of the real cost. The other 98.4% is human time.
Cloud services have raced to the bottom. But someone still has to build the product, maintain it, fix bugs, talk to users. The cost hasn't disappeared. It's shifted almost entirely to human labour.
If I valued my time at what I'd earn employed, we're talking around $8,000/month in opportunity cost. For every dollar I spend on servers, I spend $61 on myself.
But it's not just about cost. It's about focus.
Every hour spent wrestling with boilerplate, debugging obscure errors, or learning a new API is an hour not spent on the actual problem I'm trying to solve. The goal isn't just “cheaper development”. It's having more time to understand users, improve the product, and solve real human problems.
Tools that reduce development time don't just save money. They let you spend more of your limited attention on what actually matters.
The AI Efficiency Multiplier
Using AI-assisted development, I estimate I'm building 3-5x faster than I would otherwise.
A concrete example: the receipt scanning feature. This required integrating external AI APIs, writing generator functions for streaming responses, and building a responsive UI that updates as data flows in.
Time taken: 1 day.
Estimated time without AI assistance: 2 weeks.
That's not a 3x speedup. On this feature, it was closer to 10x.
But here's the caveat: not all the generated code was production-ready. Some of it needed fixing, refactoring, or rethinking entirely. AI doesn't replace expertise. It amplifies it. You still need to know what good code looks like to recognise when it's not.
If anything, experience becomes more valuable. You can move faster, but only if you can steer.
The skill shifts from “writing everything yourself” to “reviewing, directing, and knowing what good looks like.”
The Traction: Full Transparency
Since launch on 5 January 2026 (about three weeks):
- 325 visitors
- 1,161 page views (~3.6 pages per visitor)
- 52% bounce rate
- 10 signed-up users
That's roughly a 3% visitor-to-signup conversion. Decent for an early product with no paid marketing.
All traffic is organic: SEO, word of mouth, posting in communities. $0 spent on ads.
Early days, but real usage from real people.
The Bigger Picture
What actually matters
Lower costs don't matter if you're not solving a real problem. AI tools let you build faster, but faster garbage is still garbage. The fundamentals haven't changed: find a real problem, solve it well, serve people who care.
The new economics of small
You don't need millions of users to build a sustainable business. With infrastructure at $131/month, a small community of paying users is enough. 100 users paying $5/month equals $500/month equals profitable.
This isn't “lifestyle business” as a consolation prize. It's a legitimate model.
Building for community, not scale
The goal isn't hockey-stick growth. It's serving a community of like-minded people well. Small means you can listen, respond, and actually know your users. VC-scale thinking isn't the only path. Sometimes it's the wrong one.
A note of caution
Lower costs don't automatically mean lower prices or better products. The opportunity exists, but only if you stay focused on real problems and real users. It's easier to build now, which means it's also easier to build the wrong thing.
The opportunity
Lower development costs mean you can be sustainable at smaller scale. Products can prioritise users over growth metrics. That's the future I'm betting on.
Follow Along
I'm building Alydo in public. If you're interested in the journey, want to share your own cost breakdown, or just want to chat about the economics of indie development, I'd love to hear from you.
Try Alydo if you're in Australia and want to track your grocery spending. It's free to start.