Hardware

6/976 → 22/652; +2.76 percentage points of formation share.

The most striking thing about this batch of "hardware" startups is how few of them are actually building physical products at all. Diode, Axal, camfer, Adam, and Photonium are software companies whose entire pitch is compressing the design and fabrication loop for other people's hardware, while Vendra and Forge Automation do the same for manufacturing logistics, turning what used to require weeks of RFQs and mechanical-engineering back-and-forth into a four-day or automated process. That split shows up again downstream: Autumn Labs, Dartboard, and Boost Robotics aren't designing machines either, they're managing fleets of robots that already exist, and the two live investigations into fleet observability and safety-case certification point at the same emerging problem, that once robots are deployed at scale near people and inside data centers, keeping them certified and debuggable becomes its own software category, one Saphira and Inversion Semiconductor's reshoring ambitions depend on just as much as any actuator does. This formation's 5.49x growth sits below Industrial's 8.98x, suggesting the surrounding industrial base is expanding even faster than the tooling built to serve it, a gap that makes sense given that generative design and agentic CAD tools like camfer and Adam only became credible after 2023-era multimodal models could reliably reason over geometry and engineering constraints rather than just text. The result is a category that still says "hardware" on the label but increasingly means the software layer that lets a much smaller number of physical engineers move at the speed AI-era demand now requires.