70 people. That's the full headcount at Black Forest Labs, the German startup behind the FLUX image generation models - which have quietly become infrastructure for a substantial slice of the AI image market, competing directly with Midjourney, Stable Diffusion, and DALL-E 3.
Now the company says its next move is powering physical AI.
Physical AI refers to systems that process the real world and act in it - robots, autonomous vehicles, industrial automation - as opposed to models that generate text or images on a screen. The connection Black Forest Labs is betting on: models trained to produce high-quality visual and spatial data may be well-positioned to help machines understand and navigate physical space.
The lab hasn't detailed exactly what that looks like in practice. But the company has a track record of punching above its weight. FLUX launched in August 2024 and quickly became a default choice for developers building image generation into products - largely because the output quality rivaled or beat much more expensive alternatives, and the licensing terms were more permissive than many competitors.
Competing in physical AI is a different proposition. NVIDIA has spent years and billions positioning itself as the infrastructure layer for robotics. Boston Dynamics, Figure, and a dozen well-funded humanoid robot startups are all spending heavily. Getting traction at 70 people in that space requires a very specific and defensible angle - not just better images, but something physical AI systems specifically need that Black Forest Labs uniquely provides.
The image generation track record proves they can ship quality work with a small team. The question is whether the same approach - staying lean, moving fast, finding the technical gap that larger competitors miss - translates to a market that runs on hardware, not API calls.