OpenAI is building an AI smartphone set to launch in mid-2027, using processors from MediaTek and Qualcomm with manufacturing by Luxshare. Analyst Ming-Chi Kuo projects production of up to 30 million units in the first two years. The device represents OpenAI's first major hardware venture and signals a strategic pivot toward consumer electronics.
The phone's software strategy differs fundamentally from traditional Android or iOS experiences. Rather than an app grid, OpenAI plans to replace the home screen with an agent task stream, where AI handles user requests autonomously across applications. This approach positions the device as an AI-first product rather than a smartphone that happens to run intelligent software.
The choice of smartphone form factor is telling. OpenAI has experimented with more radical hardware concepts, including specialized AI devices and displays. Moving forward with a conventional phone suggests those experimental form factors lack mainstream appeal or aren't mature enough for commercial release. A smartphone leverages existing supply chains, consumer familiarity, and established distribution networks.
The timing matters. By 2027, large language models will have evolved significantly from today's versions. OpenAI expects its AI agents to handle complex tasks reliably enough to justify replacing traditional navigation paradigms. This assumes users will trust AI intermediaries to manage their digital workflows rather than directly controlling applications.
Competition for AI hardware intensifies. Google, Apple, and Samsung all integrate AI into existing devices. OpenAI's approach attempts to differentiate through software rather than hardware innovation. If the agent task stream works smoothly, it could establish a new interface paradigm. If execution falters, users accustomed to traditional app interfaces may resist the shift.
The 30-million-unit projection reflects confidence in mainstream adoption. That volume places OpenAI among serious hardware players, not experiment status. Success depends entirely on whether AI agents can reliably perform tasks without user supervision