XPeng chairman He Xiaopeng has warned that while a new class of middleware exemplified by Openclaw could one day act like a different kind of operating system, using idle automotive AI chips to run it today is difficult. In a March 7 interview he said Openclaw’s approach is instructive, but the differences between automotive ecosystems and the smartphone or PC worlds mean broad deployment in cars will likely lag consumer devices.
Openclaw has become a talking point in Chinese tech circles as a way to aggregate and monetise idle computing resources at the edge. The metaphor of "raising little lobsters"—putting spare chip cycles to work for third‑party services—captures the idea that vehicles might host non‑driving workloads, turning cars into distributed compute nodes when they are parked or under‑utilised.
He pushed back on the notion that three of XPeng’s Turing chips could simply be repurposed to run Openclaw‑style workloads. He argued the automotive environment demands a tightly integrated ecology—hardware, system, software, security and scenario orchestration—that differs sharply from the Apple model and from typical consumer device markets. Achieving the ‘‘ant‑hive’’ network effect that brings many services and users into an ecosystem is the hard part, he said.
The remarks underscore technical and regulatory constraints unique to vehicles. Safety‑critical functions require deterministic compute, strict isolation, and certification regimes that complicate running third‑party tasks on the same silicon as driving stacks. Privacy, cybersecurity and liability concerns further limit how easily automakers can monetise spare capacity without compromising safety or regulatory compliance.
Still, He conceded Openclaw’s logic has lessons for carmakers: if the platform matures and attracts many applications over the coming year, it could begin to look and behave like an alternative operating system for connected devices. That would shift the battle over control and monetisation to software layers and developer relations as much as to raw hardware performance.
For the wider industry this is a moment of strategic choice. Automakers must weigh the potential upside of turning vehicles into distributed compute assets against the reputational and safety risks of opening up in‑car platforms. Firms that can knit together secure hardware isolation, a developer network and clear business models may find new revenue streams, but the path to scale is neither quick nor guaranteed.
He’s closing note was pragmatic: XPeng has explored many angles and will roll out experiments gradually. Partial, later‑stage adoption of Openclaw‑like features in cars is plausible, he said, but full realisation will take time and will be shaped by ecosystem dynamics rather than simply by chip availability.