Regional champions drive robot adoption
Mike Xia, CEO of Anvil Robotics, on humanoid vs. non-humanoid robots
The repeat buy signal matters because physical AI is not becoming one global supplier market, it is becoming a patchwork of regional champions that each need the same underlying robot hardware stack. In practice, local leaders want their own data, their own customer relationships, and their own deployment footprint, but they do not want to spend months wiring cameras, actuators, controls, and sensing together from scratch. That creates a reseller like pattern for Anvil, where more teams can buy the same base system as they scale data collection and proofs of concept.
-
The customer workflow is concrete. A startup or data collector buys a kit to get robots running fast, uses it to collect training data or test a task, then adds more units when more engineers, customer pilots, or collection sites come online. Anvil says 15% to 20% of customers already repeat buy, often in batches of five or ten robots.
-
The regional angle follows how robotics companies win. The stack is splitting into model builders, hardware providers, and go to market solution companies. The last group wins by owning local customer deployments and task data, not by rebuilding low level hardware infrastructure. That is why separate players can exist in Southeast Asia, Europe, and the US while still buying similar base systems.
-
This also matches the broader humanoid market structure. Even large robotics players are still early in revenue and are competing to build data flywheels from real world environments. Scale has built a Physical AI Data Engine and says open datasets are far too small, while Vingroup has formed robot subsidiaries like VinMotion and VinDynamics, showing why region specific data programs can justify their own hardware purchases.
Going forward, the winners in physical AI will look less like one dominant global robot vendor and more like many regional and vertical operators running similar base hardware with proprietary data on top. That favors suppliers that can shorten setup time, standardize the messy low level stack, and sell more units each time a customer opens a new data collection lane or deployment site.