OpenAI and the 'agentic' smartphone project: it could arrive sooner than expected

OpenAI would be accelerating the development of its first smartphone, according to analyst Ming Chi Kuo of TF Securities, known for his sources in the Asian supply chain. The project, still without official confirmation from the company, began to be discussed last week and aims for mass production as early as the first half of 2027, with an estimated cumulative shipment volume of around 30 million units between 2027 and 2028.

As previously reported, the smartphone would not be a "traditional" project, but a device designed to autonomously interact with the user based on context, proactively fulfilling tasks without the user having to navigate through multiple apps. In short, an "agentic" smartphone.

According to indications gathered by the analyst, MediaTek is currently the most accredited supplier for the device's processor. It would be a customized version of the Dimensity 9600, produced using TSMC's N2P process. Qualcomm is mentioned as an alternative, but for now, MediaTek seems better positioned to secure exclusive supply.

The architectural strength of the device would be its dual-NPU (Neural Processing Unit) configuration, designed to handle heterogeneous AI workloads—parallel operations of different natures, such as natural language understanding, visual recognition, and contextual inference—more efficiently than a single unit.

Among the standout technical specifications is an Image Signal Processor (ISP), described as a superior solution compared to current standards. The stated goal would be to enhance the device's ability to "see" and interpret the real world in real-time, a critical functionality for AI models that rely on camera input for visual reasoning and context awareness.

On the memory front, the adoption of LPDDR6 paired with UFS 5.0 aims to reduce bottlenecks in data transfer, a limitation that in the AI field can significantly impact the latency of on-device inference operations.

OpenAI's entry into the hardware sector is driven by the need to control the entire technological stack to ensure a coherent agentic experience that is not mediated by third-party platforms like iOS or Android. The smartphone, unlike current devices, would allow continuous access to contextual user data—location, activities, usage habits—essential for real-time AI inference.

Another strategic element concerns the company's financial positioning: according to some analyses, a smartphone could strengthen OpenAI's narrative in view of a possible initial public offering (IPO), amplifying the perception of a complete proprietary ecosystem rather than just a language model provider.

The hypothetical business model includes a combination of on-device AI for low-latency operations and cloud AI for computationally complex tasks, possibly coupled with subscription services integrated into the device. Luxshare is indicated as the main production partner for the device.

As is often the case, the information comes from indirect supply chain channels, and timelines remain speculative. Any official updates from OpenAI could significantly alter the picture.