3 months in: 85 users, $47 revenue, 71% job success — the honest numbers

#buildinpublic #startup #api #solofounder

Hey IH

I am Om, building PixelAPI from my home office in Hyderabad, India. Three months in. Here is the unfiltered state of things.

The Numbers

85 total users (acquired organically, zero paid ads)
9 paying customers
$47.64 total revenue
1,085 API jobs processed in the last 30 days
71.6% job success rate (up from 65% two weeks ago)

The Hardware Angle

I am running this on bare metal GPU servers in my office — no AWS, no cloud GPU bills. That is the whole pricing moat: RTX 4070 + RTX 6000 Ada hardware I already own, running 24/7. Means I can charge 2x below PhotoRoom, Claid.io, and similar services and still have margin.

Zero cloud GPU bills = structural cost advantage that is very hard for cloud-based competitors to match.

The Failure Rate Problem

28% of jobs were failing silently. Users submitting API calls, getting errors, churning without a word.

Dug into the logs and fixed the three root causes: GPU OOMs on remove-bg, upstream API timeouts, and an I/O bug in remove-text. Job success climbed from 65% to 71.6%.

What I Learned

The pricing moat is real. Nobody can match my cost structure while running on cloud GPUs. But cloud providers will keep dropping prices, so the window to build a user base before they catch up is limited.

The failure rate mattered more than I thought. Every silent failure is a user who does not come back.