Nubia's NaviX Ultra, the second-generation Doubao Phone, arrives at RMB 5,999 for the 12GB/512GB tier, with a government-subsidized entry near RMB 5,499 and higher-memory variants topping out at RMB 7,499. The pricing is unremarkable; the strategy is not.

The real story is architectural. This is a handset designed around a single foundation model—ByteDance's Doubao—rather than around a grid of app icons. That is a direct wager that the assistant becomes the operating layer, intercepting user intent before it ever reaches a discrete application. If that thesis holds, it compresses the value of the app store, the browser, and the notification stream into one conversational surface. Apple and Google have spent fifteen years monetizing the app layer; an AI-native device that routes tasks through a model quietly threatens the toll booths built on top of it. The presence of a state subsidy is equally telling. Beijing is not just supporting a phone—it is underwriting a vertically integrated stack of domestic silicon, domestic model, and domestic distribution, insulated from US export friction.

For global executives, the signal is that AI-native hardware is moving from concept to shipping product, and the differentiation is shifting from cameras and chips to which model owns the interaction. Whoever controls the on-device assistant controls the data exhaust and the default choices behind it—a more durable moat than raw hardware specs.

For Japan, the implications are sharper because the gap is structural. Japan has world-class device engineering and carrier reach through NTT Docomo, KDDI, and SoftBank, but no domestic frontier model with the scale of Doubao or GPT-class systems. An AI-native phone imported into Japan means the assistant layer—and the customer relationship—is owned offshore. That reframes the opportunity for Japanese enterprises and SIers: the winning play is not building a rival model but owning the integration and governance layer. Expect demand for on-device model orchestration, enterprise policy controls, and Japanese-language grounding to sit on top of foreign models.

For local development and RPA teams, the shift is concrete. When the assistant becomes the front door, automation moves from screen-scraping fixed UIs toward intent-driven API and agent orchestration. RPA vendors and SIers that pivot early—wrapping legacy Japanese enterprise systems in agent-callable interfaces—stand to convert this platform transition into billable modernization work rather than watching it erode the app-centric integration business they know.