According to TechCrunch, founder Nick Donahue has raised $1.65 million at a $35 million valuation for his new startup, Drafted, just nine months after shutting down his previous venture, Atmos. Atmos, a Y Combinator alum, had raised $20 million from investors like Khosla Ventures and Sam Altman, grew to 40 people and $7 million in revenue, and designed $200 million worth of houses before high interest rates killed client demand. Drafted is a complete pivot, using proprietary AI to generate residential floor plans and exterior designs in minutes for between $1,000 and $2,000, with no in-house designers. The round was led by Bill Clerico and includes Stripe’s Patrick Collison and Warriors player Moses Moody. Since opening to the public, the five-month-old company is seeing about 1,000 daily users.
From operations to algorithms
Here’s the thing about Atmos: it sounds successful on paper. But Donahue calls it a “glamorized architecture firm.” It never escaped the service business model. You still needed expensive human designers in the loop, which meant it was slow, operationally heavy, and ultimately just as vulnerable to economic shifts as any traditional firm. When the Fed hiked rates, the whole thing collapsed. So Drafted is the anti-Atmos. It’s a pure software play. You input your specs—bedrooms, square footage, style—and its AI model, trained on real, permitted house plans, spits out options. Donahue says it costs a fraction of a penny per generation versus using a general-purpose AI. The goal is to sit in the neglected middle between a $20,000 architect and a $500 cookie-cutter template you find online.
The chicken-and-egg market
Now, the big question: is there even a market for this? The numbers are sobering. Of the million new homes built in the U.S. annually, only about 300,000 are custom. Most people buy existing homes or pick from a builder’s catalog. Investors like Clerico are betting on a classic disruption thesis: make something radically cheaper and faster, and you’ll expand the market itself. He compares it to Uber not just replacing taxis but creating a whole new category of frequent users. “There’s really no reason in the future why everyone shouldn’t have a totally custom designed home,” Clerico says. That’s a bold vision. But is it realistic? The housing industry is a beast of permits, local codes, entrenched builders, and cautious buyers. It’s rebuffed plenty of would-be disruptors before.
Moats and Midjourney
Then there’s the defensibility problem. What’s to stop someone else from buying similar data and training a competing model? Donahue’s answer is interesting: he points to brand and community, citing his friend David Holz at Midjourney. Despite a flood of new image AI models, Midjourney’s user base remains loyal because of the specific experience and quality. Donahue thinks Drafted can achieve the same in home design—become the trusted, go-to place. It’s a plausible theory, but it requires moving incredibly fast and nailing the user experience from day one. The early traction of 1,000 daily users is a tiny start, but it’s something.
The second-time advantage
Look, failure is often the best market research. Donahue’s real edge might not be the AI model itself, but the brutal, ground-level knowledge he gained from Atmos. He knows exactly where the operational landmines are. He’s felt the pain of client financing falling through. He understands the real constraints of what can actually be built and permitted. That’s invaluable context when you’re building a tool meant to output *practical* plans. So while the market question is huge and the competition will come, betting on a founder who’s already failed once in the exact same problem space? That’s probably the smartest part of this whole deal. He’s not guessing at the pain points; he’s lived them.
