Ai Mining Co.
The definitive source for GPU rental market data.
18 months ago, a dog attack ended the 10-year construction management career I'd built. The injury required spinal cord surgery, permanent nerve damage, and I lost the ability to drive. Being on job sites, driving daily, the physical motion of the work, none of it was possible the way it had been. Everything about how I operated, physically and professionally, had to change.
I'm not writing this for sympathy. I'm writing it because the thing I built during that recovery is now live.
Sitting at a desk became the new path. I came from a real estate background, and when I started looking into GPU rentals, I recognized something familiar: a fragmented market in the early stages of organization. Lots of activity, no clean way to understand it from the outside. So I started digging into the data and immediately hit a wall I couldn't believe hadn't been solved.
The same H100 GPU clears at under $1/hour on one marketplace and $10/hour on another. Same chip, same week. A 10x spread, with no neutral place to see it.
So instead of becoming a GPU operator, I spent six months building the data layer I couldn't find. Ai Mining Co. (aiminingco.com) tracks daily observed GPU rental prices across 30+ marketplaces, spanning North America, Europe, and Latin America.
The philosophy behind every product decision: "Do not make the system more helpful than the data allows." When a price is missing, we show an em dash. We don't estimate, we don't fill gaps to look smarter, because in institutional markets, wrong data is worse than no data.
I built this on late nights, with a body that doesn't work the way it used to, and no prior experience building software platforms. Solo, bootstrapped, no outside funding. The marketplace operators who took my cold emails seriously and the engineers who answered my questions at 11 pm proved it wasn't impossible to do alone.
If you're going through a hard stretch and wondering whether the next thing is possible - it is.
Happy to answer anything here about the data, the market, or building a platform solo through all of this.
5 Comments
That philosophy line stood out to me:
"Do not make the system more helpful than the data allows."
A lot of products say they value accuracy, but they still end up under pressure to replace uncertainty with something that feels more complete.
I'm curious whether you've run into that tension yet while building this.
Not pressure, really.
The principle decides it before the moment comes. If the data isn't there, it's an em dash, full stop. The call's already made, so there's nothing to wrestle with.
Where my head actually is. Trying to get 1% better every day and listening to the people using it. That's the part that's alive.
Feedback tells me what to build next. The principle just tells me what not to fake.
That's actually the part I'd be most interested in exploring further.
The principle sounds clear.
What I'd be curious about is what happens when the feedback starts pulling in a different direction.
Not because the principle changes.
Because some of the hardest product decisions seem to happen when two reasonable things stop pointing to the same answer.
Probably more than I'd try to unpack properly in a thread though.
What's the best email to reach you on?
Easiest to keep it here. I'd rather think out loud where people can poke holes.
On the divergence: when "what users want" and "what's true" stop agreeing, only one of them gets a vote on the data.
Feedback can reshape what I build, how I show it, and what I prioritize next. It doesn't get to decide whether a number is real. So those conflicts resolve cleaner than you'd expect, the principle isn't one input among many, it's the floor.
The hard calls all live above the floor: which real thing to build next, never whether to fake one.
Maybe.
I just wonder whether the boundary feels as clear from the user's side as it does from yours.
The principle seems straightforward.
The expectations people build around the product are usually the part that surprises me.