Davis raises €4.6M for Gaudi‑1

Architecture AI is trying to move from pretty renderings to actual building workflows. That is the gap here. Lots of tools can generate images of buildings, but real projects have to survive zoning rules, floor-plan logic, economics, and human review. Davis, a Paris startup founded in 2025, says it wants to solve that harder problem — and on May 6 it paired a €4.6 million pre-seed round with the launch of Gaudi-1, its first model for automated architectural generation. ### What is Davis actually building? Davis is not pitching itself as a generic AI design app. The company says it compresses the messy early stage of real estate development — site analysis, feasibility work, volumetrics, floor plans, and space planning — into one workflow that can deliver architect-grade outputs in hours or days instead of weeks or months. Humans still review the output before anything goes to customers. ### Why is that stage so painful? Because early real estate work is fragmented by default. A developer usually has to pull together regulatory limits, technical constraints, market assumptions, and design studies across several tools and several specialists. That coordination drag — that speed is not a nice-to-have here; it changes project economics. ### So what is Gaudi-1? Gaudi-1 is Davis’s first proprietary model for generating architectural designs under real-world constraints. The important detail is the representation. Instead of treating a building like an image made of pixels, Davis says the model works in a discrete architecture built around building parts should be easier to steer than a system built around pictures. ### Why does “discrete” matter so much? Because architects do not need a beautiful guess. They need something that can respect constraints. A pixel model is a bit like asking AI to paint a convincing building. A discrete model is closer to asking it to arrange legal, physical, and economic constraints. Gaudi-1 posted state-of-the-art results on floor-plan benchmarks including RPLAN and MSD, though those are company claims for now. ### Who backed the company? The €4.6 million round was led by Heartcore Capital and Balderton Capital, with participation from Yellow, Evantic, and Entrepreneurs First. There is also a notable angel list from teams behind SpaceMaker, Black Forest Labs, Hugging Face, Supabase, Cleo, and Spore.bio. That mix matters because it blends proptech, frontier AI, and company-building networks rather than treating this as just another vertical SaaS bet. ### Why not just sell software seats? Turns out Davis is explicitly avoiding that. The company says it uses a service model instead of a standard software licensing model, delivering finished studies and designs directly to developers and investors. That is a strong statement about where the market pain really sits. If buyers care about getting a feasibility package in 48 hours, they may prefer an outcome over another dashboard. ### Does it have real traction yet? The company is still very early, so this is the part to treat carefully. But investors and secondary coverage say Davis has already signed dozens of clients or pilots across two continents, with plans to support hundreds of projects over the next year. If that holds up, it suggests the company found a buyer need before polishing a broad software platform — which is usually the harder part in construction tech. ### Bottom line? This raise matters less because of the amount and more because of the bet behind it. Davis is arguing that architecture AI will not win by making nicer images. It will win by producing constrained, reviewable, economically useful work products that slot into real development pipelines. Gaudi-1 is its first shot at that — and now it has fresh capital to see whether the model can scale beyond early customer enthusiasm.