Software-First Path to AI Hardware
Sam Hall, CEO of Wafer, on AI agent form factors
The Xiaomi analogy points to a very specific startup path, start with software that creates fan demand, then use that demand to earn the right to ship hardware. Xiaomi began with MIUI in August 2010, built an enthusiast community around a custom Android layer, then launched its first phone in August 2011 after raising a $41 million Series A. In this interview, the same logic underpins Wafer’s thinking about using an AI native OS to prove demand before chasing large OEM distribution.
-
A custom ROM is a low cost way to test whether users want a different phone experience. It lets a startup ship a home screen, assistant, notifications, and workflows before paying for manufacturing, carrier deals, and retail distribution. The interview frames Wafer’s early plan the same way, first seed devices, then broader installs, then OEM talks.
-
The important Xiaomi lesson is not just that software came first. It is that software can become a distribution wedge for hardware. MIUI gave Xiaomi direct feedback loops with power users, weekly releases, and a base of people already attached to the experience before the Mi 1 phone arrived.
-
For an AI OS company, this path is attractive because app level products cannot see enough context across the phone. In the interview, Wafer argues that only the OS can watch behavior across apps closely enough to predict and automate actions, which makes the software layer itself the core product, not just a feature on top of Android.
The next step is likely a split between enthusiast adoption and focused commercial deployments. If an AI OS proves it can make a phone materially more useful, the Xiaomi playbook suggests two expansion lanes, branded hardware that showcases the software, or narrower firmware deals where a buyer trusts the system enough to deploy it across a defined fleet of phones.